diff options
Diffstat (limited to 'thirdparty')
103 files changed, 10359 insertions, 2692 deletions
diff --git a/thirdparty/README.md b/thirdparty/README.md index 71053de016..e63e1fc109 100644 --- a/thirdparty/README.md +++ b/thirdparty/README.md @@ -128,7 +128,7 @@ Files extracted from upstream source: ## glad - Upstream: https://github.com/Dav1dde/glad -- Version: 0.1.25 +- Version: 0.1.28 - License: MIT The files we package are automatically generated. @@ -233,7 +233,7 @@ Godot-made change marked with `// -- GODOT --` comments. ## libwebp - Upstream: https://chromium.googlesource.com/webm/libwebp/ -- Version: 1.0.0 +- Version: 1.0.1 - License: BSD-3-Clause Files extracted from upstream source: @@ -264,9 +264,7 @@ File extracted from upstream source: - From `roles/ws` exclude `ext` folder. - From `tls` exclude `openssl` folder. - Also copy `win32helpers/` from `win32port/` inside `thirdparty/libwebsockets` -- A small fix has been added in `libwebsockets/libwebsockets.h` to `#include <sys/socket.h>` for the BSD family. - This change has been PRed upstream, and should be merged before the next update. Remember to check and remove this line. -- Another fix has been added to allow building for 32-bits UWP, replacing `GetFileSize[Ex]` and `CreateFileW` with supported functions. +- A fix has been added to allow building for 32-bits UWP, replacing `GetFileSize[Ex]` and `CreateFileW` with supported functions. There is a diff for this change in `thirdparty/libwebsockets/uwp_fixes.diff` Important: `lws_config.h` and `lws_config_private.h` contains custom @@ -375,6 +373,10 @@ Collection of single-file libraries used in Godot components. ### scene +- `easing_equations.cpp` + * Upstream: http://robertpenner.com/easing/ via https://github.com/jesusgollonet/ofpennereasing (modified to fit Godot types) + * Version: git (af72c14, 2008) + Godot types and style changes + * License: BSD-3-Clause - `mikktspace.{c,h}` * Upstream: https://wiki.blender.org/index.php/Dev:Shading/Tangent_Space_Normal_Maps * Version: 1.0 @@ -501,7 +503,7 @@ changes are marked with `// -- GODOT --` comments. ## tinyexr - Upstream: https://github.com/syoyo/tinyexr -- Version: git (2d5375f, 2018) +- Version: git (5ae30aa, 2018) - License: BSD-3-Clause Files extracted from upstream source: diff --git a/thirdparty/enet/godot.cpp b/thirdparty/enet/godot.cpp index 6ba7cf0000..73a09f9b1d 100644 --- a/thirdparty/enet/godot.cpp +++ b/thirdparty/enet/godot.cpp @@ -33,7 +33,7 @@ */ #include "core/io/ip.h" -#include "core/io/packet_peer_udp.h" +#include "core/io/net_socket.h" #include "core/os/os.h" // This must be last for windows to compile (tested with MinGW) @@ -90,6 +90,16 @@ int enet_address_get_host(const ENetAddress *address, char *name, size_t nameLen return -1; } +ENetSocket enet_socket_create(ENetSocketType type) { + + NetSocket *socket = NetSocket::create(); + IP::Type ip_type = IP::TYPE_ANY; + socket->open(NetSocket::TYPE_UDP, ip_type); + socket->set_blocking_enabled(false); + + return socket; +} + int enet_socket_bind(ENetSocket socket, const ENetAddress *address) { IP_Address ip; @@ -99,23 +109,15 @@ int enet_socket_bind(ENetSocket socket, const ENetAddress *address) { ip.set_ipv6(address->host); } - PacketPeerUDP *sock = (PacketPeerUDP *)socket; - if (sock->listen(address->port, ip) != OK) { + NetSocket *sock = (NetSocket *)socket; + if (sock->bind(ip, address->port) != OK) { return -1; } return 0; } -ENetSocket enet_socket_create(ENetSocketType type) { - - PacketPeerUDP *socket = memnew(PacketPeerUDP); - socket->set_blocking_mode(false); - - return socket; -} - void enet_socket_destroy(ENetSocket socket) { - PacketPeerUDP *sock = (PacketPeerUDP *)socket; + NetSocket *sock = (NetSocket *)socket; sock->close(); memdelete(sock); } @@ -124,13 +126,12 @@ int enet_socket_send(ENetSocket socket, const ENetAddress *address, const ENetBu ERR_FAIL_COND_V(address == NULL, -1); - PacketPeerUDP *sock = (PacketPeerUDP *)socket; + NetSocket *sock = (NetSocket *)socket; IP_Address dest; Error err; size_t i = 0; dest.set_ipv6(address->host); - sock->set_dest_address(dest, address->port); // Create a single packet. PoolVector<uint8_t> out; @@ -148,7 +149,8 @@ int enet_socket_send(ENetSocket socket, const ENetAddress *address, const ENetBu pos += buffers[i].dataLength; } - err = sock->put_packet((const uint8_t *)&w[0], size); + int sent = 0; + err = sock->sendto((const uint8_t *)&w[0], size, sent, dest, address->port); if (err != OK) { if (err == ERR_BUSY) { // Blocking call @@ -159,32 +161,36 @@ int enet_socket_send(ENetSocket socket, const ENetAddress *address, const ENetBu return -1; } - return size; + return sent; } int enet_socket_receive(ENetSocket socket, ENetAddress *address, ENetBuffer *buffers, size_t bufferCount) { ERR_FAIL_COND_V(bufferCount != 1, -1); - PacketPeerUDP *sock = (PacketPeerUDP *)socket; + NetSocket *sock = (NetSocket *)socket; - int pc = sock->get_available_packet_count(); - if (pc < 1) { - return pc; - } + Error ret = sock->poll(NetSocket::POLL_TYPE_IN, 0); - const uint8_t *buffer; - int buffer_size; - Error err = sock->get_packet(&buffer, buffer_size); - if (err) + if (ret == ERR_BUSY) + return 0; + + if (ret != OK) return -1; - copymem(buffers[0].data, buffer, buffer_size); + int read; + IP_Address ip; - enet_address_set_ip(address, sock->get_packet_address().get_ipv6(), 16); - address->port = sock->get_packet_port(); + Error err = sock->recvfrom((uint8_t *)buffers[0].data, buffers[0].dataLength, read, ip, address->port); + if (err == ERR_BUSY) + return 0; + + if (err != OK) + return -1; + + enet_address_set_ip(address, ip.get_ipv6(), 16); - return buffer_size; + return read; } // Not implemented diff --git a/thirdparty/glad/glad.c b/thirdparty/glad/glad.c index 35469e9031..8cc09e46e1 100644 --- a/thirdparty/glad/glad.c +++ b/thirdparty/glad/glad.c @@ -1,6 +1,6 @@ /* - OpenGL loader generated by glad 0.1.25 on Sat Jul 28 10:59:43 2018. + OpenGL loader generated by glad 0.1.28 on Thu Nov 22 16:50:04 2018. Language/Generator: C/C++ Specification: gl @@ -13,11 +13,12 @@ Loader: True Local files: False Omit khrplatform: False + Reproducible: False Commandline: --profile="compatibility" --api="gl=3.3" --generator="c" --spec="gl" --extensions="GL_ARB_debug_output,GL_ARB_framebuffer_object,GL_EXT_framebuffer_object" Online: - http://glad.dav1d.de/#profile=compatibility&language=c&specification=gl&loader=on&api=gl%3D3.3&extensions=GL_ARB_debug_output&extensions=GL_ARB_framebuffer_object&extensions=GL_EXT_framebuffer_object + https://glad.dav1d.de/#profile=compatibility&language=c&specification=gl&loader=on&api=gl%3D3.3&extensions=GL_ARB_debug_output&extensions=GL_ARB_framebuffer_object&extensions=GL_EXT_framebuffer_object */ #include <stdio.h> @@ -154,7 +155,7 @@ int gladLoadGL(void) { return status; } -struct gladGLversionStruct GLVersion; +struct gladGLversionStruct GLVersion = { 0, 0 }; #if defined(GL_ES_VERSION_3_0) || defined(GL_VERSION_3_0) #define _GLAD_IS_SOME_NEW_VERSION 1 @@ -179,7 +180,11 @@ static int get_exts(void) { num_exts_i = 0; glGetIntegerv(GL_NUM_EXTENSIONS, &num_exts_i); if (num_exts_i > 0) { - exts_i = (char **)realloc((void *)exts_i, (size_t)num_exts_i * (sizeof *exts_i)); + char **tmp_exts_i = (char **)realloc((void *)exts_i, (size_t)num_exts_i * (sizeof *exts_i)); + if (tmp_exts_i == NULL) { + return 0; + } + exts_i = tmp_exts_i; } if (exts_i == NULL) { @@ -253,766 +258,766 @@ static int has_ext(const char *ext) { return 0; } -int GLAD_GL_VERSION_1_0; -int GLAD_GL_VERSION_1_1; -int GLAD_GL_VERSION_1_2; -int GLAD_GL_VERSION_1_3; -int GLAD_GL_VERSION_1_4; -int GLAD_GL_VERSION_1_5; -int GLAD_GL_VERSION_2_0; -int GLAD_GL_VERSION_2_1; -int GLAD_GL_VERSION_3_0; -int GLAD_GL_VERSION_3_1; -int GLAD_GL_VERSION_3_2; -int GLAD_GL_VERSION_3_3; -PFNGLCOPYTEXIMAGE1DPROC glad_glCopyTexImage1D; -PFNGLVERTEXATTRIBI3UIPROC glad_glVertexAttribI3ui; -PFNGLWINDOWPOS2SPROC glad_glWindowPos2s; -PFNGLWINDOWPOS2IPROC glad_glWindowPos2i; -PFNGLWINDOWPOS2FPROC glad_glWindowPos2f; -PFNGLWINDOWPOS2DPROC glad_glWindowPos2d; -PFNGLVERTEX2FVPROC glad_glVertex2fv; -PFNGLINDEXIPROC glad_glIndexi; -PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer; -PFNGLRECTDVPROC glad_glRectdv; -PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC glad_glCompressedTexSubImage3D; -PFNGLEVALCOORD2DPROC glad_glEvalCoord2d; -PFNGLEVALCOORD2FPROC glad_glEvalCoord2f; -PFNGLINDEXDPROC glad_glIndexd; -PFNGLVERTEXATTRIB1SVPROC glad_glVertexAttrib1sv; -PFNGLINDEXFPROC glad_glIndexf; -PFNGLBINDSAMPLERPROC glad_glBindSampler; -PFNGLLINEWIDTHPROC glad_glLineWidth; -PFNGLCOLORP3UIVPROC glad_glColorP3uiv; -PFNGLGETINTEGERI_VPROC glad_glGetIntegeri_v; -PFNGLGETMAPFVPROC glad_glGetMapfv; -PFNGLINDEXSPROC glad_glIndexs; -PFNGLCOMPILESHADERPROC glad_glCompileShader; -PFNGLGETTRANSFORMFEEDBACKVARYINGPROC glad_glGetTransformFeedbackVarying; -PFNGLWINDOWPOS2IVPROC glad_glWindowPos2iv; -PFNGLINDEXFVPROC glad_glIndexfv; -PFNGLFOGIVPROC glad_glFogiv; -PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate; -PFNGLRASTERPOS2FVPROC glad_glRasterPos2fv; -PFNGLLIGHTMODELIVPROC glad_glLightModeliv; -PFNGLCOLOR4UIPROC glad_glColor4ui; -PFNGLSECONDARYCOLOR3FVPROC glad_glSecondaryColor3fv; -PFNGLMULTITEXCOORDP3UIPROC glad_glMultiTexCoordP3ui; -PFNGLFOGFVPROC glad_glFogfv; -PFNGLVERTEXP4UIPROC glad_glVertexP4ui; -PFNGLENABLEIPROC glad_glEnablei; -PFNGLVERTEX4IVPROC glad_glVertex4iv; -PFNGLEVALCOORD1FVPROC glad_glEvalCoord1fv; -PFNGLWINDOWPOS2SVPROC glad_glWindowPos2sv; -PFNGLVERTEXATTRIBP4UIPROC glad_glVertexAttribP4ui; -PFNGLCREATESHADERPROC glad_glCreateShader; -PFNGLISBUFFERPROC glad_glIsBuffer; -PFNGLGETMULTISAMPLEFVPROC glad_glGetMultisamplefv; -PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers; -PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D; -PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D; -PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f; -PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate; -PFNGLVERTEX4FVPROC glad_glVertex4fv; -PFNGLBINDTEXTUREPROC glad_glBindTexture; -PFNGLVERTEXATTRIB1SPROC glad_glVertexAttrib1s; -PFNGLTEXCOORD2FVPROC glad_glTexCoord2fv; -PFNGLSAMPLEMASKIPROC glad_glSampleMaski; -PFNGLVERTEXP2UIPROC glad_glVertexP2ui; -PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC glad_glDrawRangeElementsBaseVertex; -PFNGLTEXCOORD4FVPROC glad_glTexCoord4fv; -PFNGLUNIFORMMATRIX3X2FVPROC glad_glUniformMatrix3x2fv; -PFNGLPOINTSIZEPROC glad_glPointSize; -PFNGLVERTEXATTRIB2DVPROC glad_glVertexAttrib2dv; -PFNGLDELETEPROGRAMPROC glad_glDeleteProgram; -PFNGLCOLOR4BVPROC glad_glColor4bv; -PFNGLRASTERPOS2FPROC glad_glRasterPos2f; -PFNGLRASTERPOS2DPROC glad_glRasterPos2d; -PFNGLLOADIDENTITYPROC glad_glLoadIdentity; -PFNGLRASTERPOS2IPROC glad_glRasterPos2i; -PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage; -PFNGLUNIFORMMATRIX4X3FVPROC glad_glUniformMatrix4x3fv; -PFNGLCOLOR3BPROC glad_glColor3b; -PFNGLCLEARBUFFERFVPROC glad_glClearBufferfv; -PFNGLEDGEFLAGPROC glad_glEdgeFlag; -PFNGLDELETESAMPLERSPROC glad_glDeleteSamplers; -PFNGLVERTEX3DPROC glad_glVertex3d; -PFNGLVERTEX3FPROC glad_glVertex3f; -PFNGLVERTEX3IPROC glad_glVertex3i; -PFNGLCOLOR3IPROC glad_glColor3i; -PFNGLUNIFORM3FPROC glad_glUniform3f; -PFNGLVERTEXATTRIB4UBVPROC glad_glVertexAttrib4ubv; -PFNGLCOLOR3SPROC glad_glColor3s; -PFNGLVERTEX3SPROC glad_glVertex3s; -PFNGLTEXCOORDP2UIPROC glad_glTexCoordP2ui; -PFNGLCOLORMASKIPROC glad_glColorMaski; -PFNGLCLEARBUFFERFIPROC glad_glClearBufferfi; -PFNGLTEXCOORD1IVPROC glad_glTexCoord1iv; -PFNGLBLITFRAMEBUFFERPROC glad_glBlitFramebuffer; -PFNGLMULTITEXCOORDP2UIPROC glad_glMultiTexCoordP2ui; -PFNGLGETSAMPLERPARAMETERIIVPROC glad_glGetSamplerParameterIiv; -PFNGLGETFRAGDATAINDEXPROC glad_glGetFragDataIndex; -PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f; -PFNGLVERTEX2IVPROC glad_glVertex2iv; -PFNGLCOLOR3SVPROC glad_glColor3sv; -PFNGLGETVERTEXATTRIBDVPROC glad_glGetVertexAttribdv; -PFNGLUNIFORMMATRIX3X4FVPROC glad_glUniformMatrix3x4fv; -PFNGLNORMALPOINTERPROC glad_glNormalPointer; -PFNGLTEXCOORDP3UIVPROC glad_glTexCoordP3uiv; -PFNGLVERTEX4SVPROC glad_glVertex4sv; -PFNGLPASSTHROUGHPROC glad_glPassThrough; -PFNGLMULTITEXCOORDP4UIPROC glad_glMultiTexCoordP4ui; -PFNGLFOGIPROC glad_glFogi; -PFNGLBEGINPROC glad_glBegin; -PFNGLEVALCOORD2DVPROC glad_glEvalCoord2dv; -PFNGLCOLOR3UBVPROC glad_glColor3ubv; -PFNGLVERTEXPOINTERPROC glad_glVertexPointer; -PFNGLSECONDARYCOLOR3UIVPROC glad_glSecondaryColor3uiv; -PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers; -PFNGLDRAWARRAYSPROC glad_glDrawArrays; -PFNGLUNIFORM1UIPROC glad_glUniform1ui; -PFNGLMULTITEXCOORD1DPROC glad_glMultiTexCoord1d; -PFNGLMULTITEXCOORD1FPROC glad_glMultiTexCoord1f; -PFNGLLIGHTFVPROC glad_glLightfv; -PFNGLTEXCOORDP3UIPROC glad_glTexCoordP3ui; -PFNGLVERTEXATTRIB3DPROC glad_glVertexAttrib3d; -PFNGLCLEARPROC glad_glClear; -PFNGLMULTITEXCOORD1IPROC glad_glMultiTexCoord1i; -PFNGLGETACTIVEUNIFORMNAMEPROC glad_glGetActiveUniformName; -PFNGLMULTITEXCOORD1SPROC glad_glMultiTexCoord1s; -PFNGLISENABLEDPROC glad_glIsEnabled; -PFNGLSTENCILOPPROC glad_glStencilOp; -PFNGLGETQUERYOBJECTUIVPROC glad_glGetQueryObjectuiv; -PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D; -PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv; -PFNGLTRANSLATEFPROC glad_glTranslatef; -PFNGLVERTEXATTRIB4NUBPROC glad_glVertexAttrib4Nub; -PFNGLTRANSLATEDPROC glad_glTranslated; -PFNGLTEXCOORD3SVPROC glad_glTexCoord3sv; -PFNGLGETFRAGDATALOCATIONPROC glad_glGetFragDataLocation; -PFNGLTEXIMAGE1DPROC glad_glTexImage1D; -PFNGLVERTEXP3UIVPROC glad_glVertexP3uiv; -PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv; -PFNGLSECONDARYCOLOR3BVPROC glad_glSecondaryColor3bv; -PFNGLGETMATERIALFVPROC glad_glGetMaterialfv; -PFNGLGETTEXIMAGEPROC glad_glGetTexImage; -PFNGLFOGCOORDFVPROC glad_glFogCoordfv; -PFNGLPIXELMAPUIVPROC glad_glPixelMapuiv; -PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog; -PFNGLGETQUERYOBJECTI64VPROC glad_glGetQueryObjecti64v; -PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers; -PFNGLINDEXSVPROC glad_glIndexsv; -PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders; -PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer; -PFNGLVERTEX3IVPROC glad_glVertex3iv; -PFNGLBITMAPPROC glad_glBitmap; -PFNGLMATERIALIPROC glad_glMateriali; -PFNGLISVERTEXARRAYPROC glad_glIsVertexArray; -PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray; -PFNGLGETQUERYIVPROC glad_glGetQueryiv; -PFNGLTEXCOORD4FPROC glad_glTexCoord4f; -PFNGLTEXCOORD4DPROC glad_glTexCoord4d; -PFNGLGETSAMPLERPARAMETERFVPROC glad_glGetSamplerParameterfv; -PFNGLTEXCOORD4IPROC glad_glTexCoord4i; -PFNGLMATERIALFPROC glad_glMaterialf; -PFNGLTEXCOORD4SPROC glad_glTexCoord4s; -PFNGLGETUNIFORMINDICESPROC glad_glGetUniformIndices; -PFNGLISSHADERPROC glad_glIsShader; -PFNGLMULTITEXCOORD2SPROC glad_glMultiTexCoord2s; -PFNGLVERTEXATTRIBI4UBVPROC glad_glVertexAttribI4ubv; -PFNGLVERTEX3DVPROC glad_glVertex3dv; -PFNGLGETINTEGER64VPROC glad_glGetInteger64v; -PFNGLPOINTPARAMETERIVPROC glad_glPointParameteriv; -PFNGLENABLEPROC glad_glEnable; -PFNGLGETACTIVEUNIFORMSIVPROC glad_glGetActiveUniformsiv; -PFNGLCOLOR4FVPROC glad_glColor4fv; -PFNGLTEXCOORD1FVPROC glad_glTexCoord1fv; -PFNGLTEXCOORD2SVPROC glad_glTexCoord2sv; -PFNGLVERTEXATTRIB4DVPROC glad_glVertexAttrib4dv; -PFNGLMULTITEXCOORD1DVPROC glad_glMultiTexCoord1dv; -PFNGLMULTITEXCOORD2IPROC glad_glMultiTexCoord2i; -PFNGLTEXCOORD3FVPROC glad_glTexCoord3fv; -PFNGLSECONDARYCOLOR3USVPROC glad_glSecondaryColor3usv; -PFNGLTEXGENFPROC glad_glTexGenf; -PFNGLMULTITEXCOORDP3UIVPROC glad_glMultiTexCoordP3uiv; -PFNGLVERTEXATTRIBP3UIPROC glad_glVertexAttribP3ui; -PFNGLMULTITEXCOORDP1UIPROC glad_glMultiTexCoordP1ui; -PFNGLGETPOINTERVPROC glad_glGetPointerv; -PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset; -PFNGLGETUNIFORMUIVPROC glad_glGetUniformuiv; -PFNGLNORMAL3FVPROC glad_glNormal3fv; -PFNGLSECONDARYCOLOR3SPROC glad_glSecondaryColor3s; -PFNGLDEPTHRANGEPROC glad_glDepthRange; -PFNGLFRUSTUMPROC glad_glFrustum; -PFNGLMULTITEXCOORD4SVPROC glad_glMultiTexCoord4sv; -PFNGLDRAWBUFFERPROC glad_glDrawBuffer; -PFNGLPUSHMATRIXPROC glad_glPushMatrix; -PFNGLRASTERPOS3FVPROC glad_glRasterPos3fv; -PFNGLORTHOPROC glad_glOrtho; -PFNGLDRAWELEMENTSINSTANCEDPROC glad_glDrawElementsInstanced; -PFNGLWINDOWPOS3SVPROC glad_glWindowPos3sv; -PFNGLCLEARINDEXPROC glad_glClearIndex; -PFNGLMAP1DPROC glad_glMap1d; -PFNGLMAP1FPROC glad_glMap1f; -PFNGLFLUSHPROC glad_glFlush; -PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv; -PFNGLINDEXIVPROC glad_glIndexiv; -PFNGLRASTERPOS3SVPROC glad_glRasterPos3sv; -PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv; -PFNGLPIXELZOOMPROC glad_glPixelZoom; -PFNGLFENCESYNCPROC glad_glFenceSync; -PFNGLDELETEVERTEXARRAYSPROC glad_glDeleteVertexArrays; -PFNGLCOLORP3UIPROC glad_glColorP3ui; -PFNGLVERTEXATTRIB3SVPROC glad_glVertexAttrib3sv; -PFNGLBEGINCONDITIONALRENDERPROC glad_glBeginConditionalRender; -PFNGLDRAWELEMENTSBASEVERTEXPROC glad_glDrawElementsBaseVertex; -PFNGLGETTEXLEVELPARAMETERIVPROC glad_glGetTexLevelParameteriv; -PFNGLLIGHTIPROC glad_glLighti; -PFNGLMULTITEXCOORDP4UIVPROC glad_glMultiTexCoordP4uiv; -PFNGLLIGHTFPROC glad_glLightf; -PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation; -PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate; -PFNGLGENSAMPLERSPROC glad_glGenSamplers; -PFNGLCLAMPCOLORPROC glad_glClampColor; -PFNGLUNIFORM4IVPROC glad_glUniform4iv; -PFNGLCLEARSTENCILPROC glad_glClearStencil; -PFNGLTEXCOORDP1UIVPROC glad_glTexCoordP1uiv; -PFNGLMULTITEXCOORD3FVPROC glad_glMultiTexCoord3fv; -PFNGLGETPIXELMAPUIVPROC glad_glGetPixelMapuiv; -PFNGLGENTEXTURESPROC glad_glGenTextures; -PFNGLTEXCOORD4IVPROC glad_glTexCoord4iv; -PFNGLGETTEXPARAMETERIUIVPROC glad_glGetTexParameterIuiv; -PFNGLINDEXPOINTERPROC glad_glIndexPointer; -PFNGLVERTEXATTRIB4NBVPROC glad_glVertexAttrib4Nbv; -PFNGLISSYNCPROC glad_glIsSync; -PFNGLVERTEX2FPROC glad_glVertex2f; -PFNGLVERTEX2DPROC glad_glVertex2d; -PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers; -PFNGLUNIFORM2IPROC glad_glUniform2i; -PFNGLMAPGRID2DPROC glad_glMapGrid2d; -PFNGLMAPGRID2FPROC glad_glMapGrid2f; -PFNGLTEXCOORDP4UIPROC glad_glTexCoordP4ui; -PFNGLVERTEX2IPROC glad_glVertex2i; -PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer; -PFNGLFRAMEBUFFERTEXTURELAYERPROC glad_glFramebufferTextureLayer; -PFNGLVERTEX2SPROC glad_glVertex2s; -PFNGLNORMAL3BVPROC glad_glNormal3bv; -PFNGLVERTEXATTRIB4NUIVPROC glad_glVertexAttrib4Nuiv; -PFNGLFLUSHMAPPEDBUFFERRANGEPROC glad_glFlushMappedBufferRange; -PFNGLSECONDARYCOLOR3SVPROC glad_glSecondaryColor3sv; -PFNGLVERTEX3SVPROC glad_glVertex3sv; -PFNGLGENQUERIESPROC glad_glGenQueries; -PFNGLGETPIXELMAPFVPROC glad_glGetPixelMapfv; -PFNGLTEXENVFPROC glad_glTexEnvf; -PFNGLVERTEXATTRIBP1UIPROC glad_glVertexAttribP1ui; -PFNGLTEXSUBIMAGE3DPROC glad_glTexSubImage3D; -PFNGLGETINTEGER64I_VPROC glad_glGetInteger64i_v; -PFNGLFOGCOORDDPROC glad_glFogCoordd; -PFNGLFOGCOORDFPROC glad_glFogCoordf; -PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D; -PFNGLTEXENVIPROC glad_glTexEnvi; -PFNGLMULTITEXCOORD1IVPROC glad_glMultiTexCoord1iv; -PFNGLISENABLEDIPROC glad_glIsEnabledi; -PFNGLSECONDARYCOLORP3UIPROC glad_glSecondaryColorP3ui; -PFNGLVERTEXATTRIBI2IPROC glad_glVertexAttribI2i; -PFNGLBINDFRAGDATALOCATIONINDEXEDPROC glad_glBindFragDataLocationIndexed; -PFNGLMULTITEXCOORD2DVPROC glad_glMultiTexCoord2dv; -PFNGLUNIFORM2IVPROC glad_glUniform2iv; -PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv; -PFNGLUNIFORM4UIVPROC glad_glUniform4uiv; -PFNGLMATRIXMODEPROC glad_glMatrixMode; -PFNGLFEEDBACKBUFFERPROC glad_glFeedbackBuffer; -PFNGLGETMAPIVPROC glad_glGetMapiv; -PFNGLFRAMEBUFFERTEXTURE1DPROC glad_glFramebufferTexture1D; -PFNGLGETSHADERIVPROC glad_glGetShaderiv; -PFNGLMULTITEXCOORD2DPROC glad_glMultiTexCoord2d; -PFNGLMULTITEXCOORD2FPROC glad_glMultiTexCoord2f; -PFNGLBINDFRAGDATALOCATIONPROC glad_glBindFragDataLocation; -PFNGLPRIORITIZETEXTURESPROC glad_glPrioritizeTextures; -PFNGLCALLLISTPROC glad_glCallList; -PFNGLSECONDARYCOLOR3UBVPROC glad_glSecondaryColor3ubv; -PFNGLGETDOUBLEVPROC glad_glGetDoublev; -PFNGLMULTITEXCOORD3IVPROC glad_glMultiTexCoord3iv; -PFNGLVERTEXATTRIB1DPROC glad_glVertexAttrib1d; -PFNGLLIGHTMODELFPROC glad_glLightModelf; -PFNGLGETUNIFORMIVPROC glad_glGetUniformiv; -PFNGLVERTEX2SVPROC glad_glVertex2sv; -PFNGLLIGHTMODELIPROC glad_glLightModeli; -PFNGLWINDOWPOS3IVPROC glad_glWindowPos3iv; -PFNGLMULTITEXCOORDP1UIVPROC glad_glMultiTexCoordP1uiv; -PFNGLUNIFORM3FVPROC glad_glUniform3fv; -PFNGLPIXELSTOREIPROC glad_glPixelStorei; -PFNGLCALLLISTSPROC glad_glCallLists; -PFNGLMAPBUFFERPROC glad_glMapBuffer; -PFNGLSECONDARYCOLOR3DPROC glad_glSecondaryColor3d; -PFNGLTEXCOORD3IPROC glad_glTexCoord3i; -PFNGLMULTITEXCOORD4FVPROC glad_glMultiTexCoord4fv; -PFNGLRASTERPOS3IPROC glad_glRasterPos3i; -PFNGLSECONDARYCOLOR3BPROC glad_glSecondaryColor3b; -PFNGLRASTERPOS3DPROC glad_glRasterPos3d; -PFNGLRASTERPOS3FPROC glad_glRasterPos3f; -PFNGLCOMPRESSEDTEXIMAGE3DPROC glad_glCompressedTexImage3D; -PFNGLTEXCOORD3FPROC glad_glTexCoord3f; -PFNGLDELETESYNCPROC glad_glDeleteSync; -PFNGLTEXCOORD3DPROC glad_glTexCoord3d; -PFNGLTEXIMAGE2DMULTISAMPLEPROC glad_glTexImage2DMultisample; -PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv; -PFNGLMULTIDRAWELEMENTSPROC glad_glMultiDrawElements; -PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv; -PFNGLTEXCOORD3SPROC glad_glTexCoord3s; -PFNGLUNIFORM3IVPROC glad_glUniform3iv; -PFNGLRASTERPOS3SPROC glad_glRasterPos3s; -PFNGLPOLYGONMODEPROC glad_glPolygonMode; -PFNGLDRAWBUFFERSPROC glad_glDrawBuffers; -PFNGLGETACTIVEUNIFORMBLOCKIVPROC glad_glGetActiveUniformBlockiv; -PFNGLARETEXTURESRESIDENTPROC glad_glAreTexturesResident; -PFNGLISLISTPROC glad_glIsList; -PFNGLRASTERPOS2SVPROC glad_glRasterPos2sv; -PFNGLRASTERPOS4SVPROC glad_glRasterPos4sv; -PFNGLCOLOR4SPROC glad_glColor4s; -PFNGLUSEPROGRAMPROC glad_glUseProgram; -PFNGLLINESTIPPLEPROC glad_glLineStipple; -PFNGLMULTITEXCOORD1SVPROC glad_glMultiTexCoord1sv; -PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog; -PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv; -PFNGLMULTITEXCOORD2IVPROC glad_glMultiTexCoord2iv; -PFNGLUNIFORMMATRIX2X4FVPROC glad_glUniformMatrix2x4fv; -PFNGLBINDVERTEXARRAYPROC glad_glBindVertexArray; -PFNGLCOLOR4BPROC glad_glColor4b; -PFNGLSECONDARYCOLOR3FPROC glad_glSecondaryColor3f; -PFNGLCOLOR4FPROC glad_glColor4f; -PFNGLCOLOR4DPROC glad_glColor4d; -PFNGLCOLOR4IPROC glad_glColor4i; -PFNGLSAMPLERPARAMETERIIVPROC glad_glSamplerParameterIiv; -PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC glad_glMultiDrawElementsBaseVertex; -PFNGLRASTERPOS3IVPROC glad_glRasterPos3iv; -PFNGLVERTEX2DVPROC glad_glVertex2dv; -PFNGLTEXCOORD4SVPROC glad_glTexCoord4sv; -PFNGLUNIFORM2UIVPROC glad_glUniform2uiv; -PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC glad_glCompressedTexSubImage1D; -PFNGLFINISHPROC glad_glFinish; -PFNGLGETBOOLEANVPROC glad_glGetBooleanv; -PFNGLDELETESHADERPROC glad_glDeleteShader; -PFNGLDRAWELEMENTSPROC glad_glDrawElements; -PFNGLRASTERPOS2SPROC glad_glRasterPos2s; -PFNGLGETMAPDVPROC glad_glGetMapdv; -PFNGLVERTEXATTRIB4NSVPROC glad_glVertexAttrib4Nsv; -PFNGLMATERIALFVPROC glad_glMaterialfv; -PFNGLVIEWPORTPROC glad_glViewport; -PFNGLUNIFORM1UIVPROC glad_glUniform1uiv; -PFNGLTRANSFORMFEEDBACKVARYINGSPROC glad_glTransformFeedbackVaryings; -PFNGLINDEXDVPROC glad_glIndexdv; -PFNGLCOPYTEXSUBIMAGE3DPROC glad_glCopyTexSubImage3D; -PFNGLTEXCOORD3IVPROC glad_glTexCoord3iv; -PFNGLVERTEXATTRIBI3IPROC glad_glVertexAttribI3i; -PFNGLCLEARDEPTHPROC glad_glClearDepth; -PFNGLVERTEXATTRIBI4USVPROC glad_glVertexAttribI4usv; -PFNGLTEXPARAMETERFPROC glad_glTexParameterf; -PFNGLTEXPARAMETERIPROC glad_glTexParameteri; -PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource; -PFNGLTEXBUFFERPROC glad_glTexBuffer; -PFNGLPOPNAMEPROC glad_glPopName; -PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram; -PFNGLPIXELSTOREFPROC glad_glPixelStoref; -PFNGLUNIFORM3UIVPROC glad_glUniform3uiv; -PFNGLRASTERPOS4FVPROC glad_glRasterPos4fv; -PFNGLEVALCOORD1DVPROC glad_glEvalCoord1dv; -PFNGLMULTITEXCOORDP2UIVPROC glad_glMultiTexCoordP2uiv; -PFNGLRECTIPROC glad_glRecti; -PFNGLCOLOR4UBPROC glad_glColor4ub; -PFNGLMULTTRANSPOSEMATRIXFPROC glad_glMultTransposeMatrixf; -PFNGLRECTFPROC glad_glRectf; -PFNGLRECTDPROC glad_glRectd; -PFNGLNORMAL3SVPROC glad_glNormal3sv; -PFNGLNEWLISTPROC glad_glNewList; -PFNGLCOLOR4USPROC glad_glColor4us; -PFNGLVERTEXATTRIBP1UIVPROC glad_glVertexAttribP1uiv; -PFNGLLINKPROGRAMPROC glad_glLinkProgram; -PFNGLHINTPROC glad_glHint; -PFNGLRECTSPROC glad_glRects; -PFNGLTEXCOORD2DVPROC glad_glTexCoord2dv; -PFNGLRASTERPOS4IVPROC glad_glRasterPos4iv; -PFNGLGETSTRINGPROC glad_glGetString; -PFNGLVERTEXATTRIBP2UIVPROC glad_glVertexAttribP2uiv; -PFNGLEDGEFLAGVPROC glad_glEdgeFlagv; -PFNGLDETACHSHADERPROC glad_glDetachShader; -PFNGLSCALEFPROC glad_glScalef; -PFNGLENDQUERYPROC glad_glEndQuery; -PFNGLSCALEDPROC glad_glScaled; -PFNGLEDGEFLAGPOINTERPROC glad_glEdgeFlagPointer; -PFNGLCOPYPIXELSPROC glad_glCopyPixels; -PFNGLVERTEXATTRIBI2UIPROC glad_glVertexAttribI2ui; -PFNGLPOPATTRIBPROC glad_glPopAttrib; -PFNGLDELETETEXTURESPROC glad_glDeleteTextures; -PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate; -PFNGLDELETEQUERIESPROC glad_glDeleteQueries; -PFNGLNORMALP3UIVPROC glad_glNormalP3uiv; -PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f; -PFNGLVERTEXATTRIB4DPROC glad_glVertexAttrib4d; -PFNGLINITNAMESPROC glad_glInitNames; -PFNGLGETBUFFERPARAMETERI64VPROC glad_glGetBufferParameteri64v; -PFNGLCOLOR3DVPROC glad_glColor3dv; -PFNGLVERTEXATTRIBI1IPROC glad_glVertexAttribI1i; -PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv; -PFNGLWAITSYNCPROC glad_glWaitSync; -PFNGLVERTEXATTRIB4SPROC glad_glVertexAttrib4s; -PFNGLCOLORMATERIALPROC glad_glColorMaterial; -PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage; -PFNGLSAMPLERPARAMETERIPROC glad_glSamplerParameteri; -PFNGLSAMPLERPARAMETERFPROC glad_glSamplerParameterf; -PFNGLUNIFORM1FPROC glad_glUniform1f; -PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv; -PFNGLRENDERMODEPROC glad_glRenderMode; -PFNGLGETCOMPRESSEDTEXIMAGEPROC glad_glGetCompressedTexImage; -PFNGLWINDOWPOS2DVPROC glad_glWindowPos2dv; -PFNGLUNIFORM1IPROC glad_glUniform1i; -PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib; -PFNGLUNIFORM3IPROC glad_glUniform3i; -PFNGLPIXELTRANSFERIPROC glad_glPixelTransferi; -PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D; -PFNGLDISABLEPROC glad_glDisable; -PFNGLLOGICOPPROC glad_glLogicOp; -PFNGLEVALPOINT2PROC glad_glEvalPoint2; -PFNGLPIXELTRANSFERFPROC glad_glPixelTransferf; -PFNGLSECONDARYCOLOR3IPROC glad_glSecondaryColor3i; -PFNGLUNIFORM4UIPROC glad_glUniform4ui; -PFNGLCOLOR3FPROC glad_glColor3f; -PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer; -PFNGLGETTEXENVFVPROC glad_glGetTexEnvfv; -PFNGLRECTFVPROC glad_glRectfv; -PFNGLCULLFACEPROC glad_glCullFace; -PFNGLGETLIGHTFVPROC glad_glGetLightfv; -PFNGLCOLOR3DPROC glad_glColor3d; -PFNGLTEXGENDPROC glad_glTexGend; -PFNGLTEXGENIPROC glad_glTexGeni; -PFNGLMULTITEXCOORD3SPROC glad_glMultiTexCoord3s; -PFNGLGETSTRINGIPROC glad_glGetStringi; -PFNGLMULTITEXCOORD3IPROC glad_glMultiTexCoord3i; -PFNGLMULTITEXCOORD3FPROC glad_glMultiTexCoord3f; -PFNGLMULTITEXCOORD3DPROC glad_glMultiTexCoord3d; -PFNGLATTACHSHADERPROC glad_glAttachShader; -PFNGLFOGCOORDDVPROC glad_glFogCoorddv; -PFNGLUNIFORMMATRIX2X3FVPROC glad_glUniformMatrix2x3fv; -PFNGLGETTEXGENFVPROC glad_glGetTexGenfv; -PFNGLQUERYCOUNTERPROC glad_glQueryCounter; -PFNGLFOGCOORDPOINTERPROC glad_glFogCoordPointer; -PFNGLPROVOKINGVERTEXPROC glad_glProvokingVertex; -PFNGLFRAMEBUFFERTEXTURE3DPROC glad_glFramebufferTexture3D; -PFNGLTEXGENIVPROC glad_glTexGeniv; -PFNGLRASTERPOS2DVPROC glad_glRasterPos2dv; -PFNGLSECONDARYCOLOR3DVPROC glad_glSecondaryColor3dv; -PFNGLCLIENTACTIVETEXTUREPROC glad_glClientActiveTexture; -PFNGLVERTEXATTRIBI4SVPROC glad_glVertexAttribI4sv; -PFNGLSECONDARYCOLOR3USPROC glad_glSecondaryColor3us; -PFNGLNORMALP3UIPROC glad_glNormalP3ui; -PFNGLTEXENVFVPROC glad_glTexEnvfv; -PFNGLREADBUFFERPROC glad_glReadBuffer; -PFNGLTEXPARAMETERIUIVPROC glad_glTexParameterIuiv; -PFNGLDRAWARRAYSINSTANCEDPROC glad_glDrawArraysInstanced; -PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap; -PFNGLWINDOWPOS3FVPROC glad_glWindowPos3fv; -PFNGLLIGHTMODELFVPROC glad_glLightModelfv; -PFNGLSAMPLERPARAMETERIVPROC glad_glSamplerParameteriv; -PFNGLDELETELISTSPROC glad_glDeleteLists; -PFNGLGETCLIPPLANEPROC glad_glGetClipPlane; -PFNGLVERTEX4DVPROC glad_glVertex4dv; -PFNGLTEXCOORD2DPROC glad_glTexCoord2d; -PFNGLPOPMATRIXPROC glad_glPopMatrix; -PFNGLTEXCOORD2FPROC glad_glTexCoord2f; -PFNGLCOLOR4IVPROC glad_glColor4iv; -PFNGLINDEXUBVPROC glad_glIndexubv; -PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer; -PFNGLTEXCOORD2IPROC glad_glTexCoord2i; -PFNGLRASTERPOS4DPROC glad_glRasterPos4d; -PFNGLRASTERPOS4FPROC glad_glRasterPos4f; -PFNGLVERTEXATTRIB3SPROC glad_glVertexAttrib3s; -PFNGLTEXCOORD2SPROC glad_glTexCoord2s; -PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer; -PFNGLVERTEX3FVPROC glad_glVertex3fv; -PFNGLTEXCOORD4DVPROC glad_glTexCoord4dv; -PFNGLMATERIALIVPROC glad_glMaterialiv; -PFNGLVERTEXATTRIBP4UIVPROC glad_glVertexAttribP4uiv; -PFNGLISPROGRAMPROC glad_glIsProgram; -PFNGLVERTEXATTRIB4BVPROC glad_glVertexAttrib4bv; -PFNGLVERTEX4SPROC glad_glVertex4s; -PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv; -PFNGLNORMAL3DVPROC glad_glNormal3dv; -PFNGLUNIFORM4IPROC glad_glUniform4i; -PFNGLACTIVETEXTUREPROC glad_glActiveTexture; -PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray; -PFNGLROTATEDPROC glad_glRotated; -PFNGLROTATEFPROC glad_glRotatef; -PFNGLVERTEX4IPROC glad_glVertex4i; -PFNGLREADPIXELSPROC glad_glReadPixels; -PFNGLVERTEXATTRIBI3IVPROC glad_glVertexAttribI3iv; -PFNGLLOADNAMEPROC glad_glLoadName; -PFNGLUNIFORM4FPROC glad_glUniform4f; -PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glad_glRenderbufferStorageMultisample; -PFNGLGENVERTEXARRAYSPROC glad_glGenVertexArrays; -PFNGLSHADEMODELPROC glad_glShadeModel; -PFNGLMAPGRID1DPROC glad_glMapGrid1d; -PFNGLGETUNIFORMFVPROC glad_glGetUniformfv; -PFNGLMAPGRID1FPROC glad_glMapGrid1f; -PFNGLSAMPLERPARAMETERFVPROC glad_glSamplerParameterfv; -PFNGLDISABLECLIENTSTATEPROC glad_glDisableClientState; -PFNGLMULTITEXCOORD3SVPROC glad_glMultiTexCoord3sv; -PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC glad_glDrawElementsInstancedBaseVertex; -PFNGLSECONDARYCOLORPOINTERPROC glad_glSecondaryColorPointer; -PFNGLALPHAFUNCPROC glad_glAlphaFunc; -PFNGLUNIFORM1IVPROC glad_glUniform1iv; -PFNGLMULTITEXCOORD4IVPROC glad_glMultiTexCoord4iv; -PFNGLGETQUERYOBJECTIVPROC glad_glGetQueryObjectiv; -PFNGLSTENCILFUNCPROC glad_glStencilFunc; -PFNGLMULTITEXCOORD1FVPROC glad_glMultiTexCoord1fv; -PFNGLUNIFORMBLOCKBINDINGPROC glad_glUniformBlockBinding; -PFNGLCOLOR4UIVPROC glad_glColor4uiv; -PFNGLRECTIVPROC glad_glRectiv; -PFNGLCOLORP4UIPROC glad_glColorP4ui; -PFNGLRASTERPOS3DVPROC glad_glRasterPos3dv; -PFNGLEVALMESH2PROC glad_glEvalMesh2; -PFNGLEVALMESH1PROC glad_glEvalMesh1; -PFNGLTEXCOORDPOINTERPROC glad_glTexCoordPointer; -PFNGLVERTEXATTRIB4NUBVPROC glad_glVertexAttrib4Nubv; -PFNGLVERTEXATTRIBI4IVPROC glad_glVertexAttribI4iv; -PFNGLEVALCOORD2FVPROC glad_glEvalCoord2fv; -PFNGLCOLOR4UBVPROC glad_glColor4ubv; -PFNGLLOADTRANSPOSEMATRIXDPROC glad_glLoadTransposeMatrixd; -PFNGLLOADTRANSPOSEMATRIXFPROC glad_glLoadTransposeMatrixf; -PFNGLVERTEXATTRIBI4IPROC glad_glVertexAttribI4i; -PFNGLRASTERPOS2IVPROC glad_glRasterPos2iv; -PFNGLGETBUFFERSUBDATAPROC glad_glGetBufferSubData; -PFNGLTEXENVIVPROC glad_glTexEnviv; -PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate; -PFNGLVERTEXATTRIBI1UIPROC glad_glVertexAttribI1ui; -PFNGLGENBUFFERSPROC glad_glGenBuffers; -PFNGLSELECTBUFFERPROC glad_glSelectBuffer; -PFNGLVERTEXATTRIB2SVPROC glad_glVertexAttrib2sv; -PFNGLPUSHATTRIBPROC glad_glPushAttrib; -PFNGLVERTEXATTRIBIPOINTERPROC glad_glVertexAttribIPointer; -PFNGLBLENDFUNCPROC glad_glBlendFunc; -PFNGLCREATEPROGRAMPROC glad_glCreateProgram; -PFNGLTEXIMAGE3DPROC glad_glTexImage3D; -PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer; -PFNGLLIGHTIVPROC glad_glLightiv; -PFNGLPRIMITIVERESTARTINDEXPROC glad_glPrimitiveRestartIndex; -PFNGLTEXGENFVPROC glad_glTexGenfv; -PFNGLENDPROC glad_glEnd; -PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers; -PFNGLSCISSORPROC glad_glScissor; -PFNGLTEXCOORDP4UIVPROC glad_glTexCoordP4uiv; -PFNGLCLIPPLANEPROC glad_glClipPlane; -PFNGLPUSHNAMEPROC glad_glPushName; -PFNGLTEXGENDVPROC glad_glTexGendv; -PFNGLINDEXUBPROC glad_glIndexub; -PFNGLVERTEXP2UIVPROC glad_glVertexP2uiv; -PFNGLSECONDARYCOLOR3IVPROC glad_glSecondaryColor3iv; -PFNGLRASTERPOS4IPROC glad_glRasterPos4i; -PFNGLMULTTRANSPOSEMATRIXDPROC glad_glMultTransposeMatrixd; -PFNGLCLEARCOLORPROC glad_glClearColor; -PFNGLVERTEXATTRIB4UIVPROC glad_glVertexAttrib4uiv; -PFNGLNORMAL3SPROC glad_glNormal3s; -PFNGLVERTEXATTRIB4NIVPROC glad_glVertexAttrib4Niv; -PFNGLCLEARBUFFERIVPROC glad_glClearBufferiv; -PFNGLPOINTPARAMETERIPROC glad_glPointParameteri; -PFNGLCOLORP4UIVPROC glad_glColorP4uiv; -PFNGLBLENDCOLORPROC glad_glBlendColor; -PFNGLWINDOWPOS3DPROC glad_glWindowPos3d; -PFNGLVERTEXATTRIBI2UIVPROC glad_glVertexAttribI2uiv; -PFNGLSAMPLERPARAMETERIUIVPROC glad_glSamplerParameterIuiv; -PFNGLUNIFORM3UIPROC glad_glUniform3ui; -PFNGLCOLOR4DVPROC glad_glColor4dv; -PFNGLVERTEXATTRIBI4UIVPROC glad_glVertexAttribI4uiv; -PFNGLPOINTPARAMETERFVPROC glad_glPointParameterfv; -PFNGLUNIFORM2FVPROC glad_glUniform2fv; -PFNGLSECONDARYCOLOR3UBPROC glad_glSecondaryColor3ub; -PFNGLSECONDARYCOLOR3UIPROC glad_glSecondaryColor3ui; -PFNGLTEXCOORD3DVPROC glad_glTexCoord3dv; -PFNGLGETSAMPLERPARAMETERIUIVPROC glad_glGetSamplerParameterIuiv; -PFNGLBINDBUFFERRANGEPROC glad_glBindBufferRange; -PFNGLNORMAL3IVPROC glad_glNormal3iv; -PFNGLWINDOWPOS3SPROC glad_glWindowPos3s; -PFNGLPOINTPARAMETERFPROC glad_glPointParameterf; -PFNGLGETVERTEXATTRIBIUIVPROC glad_glGetVertexAttribIuiv; -PFNGLWINDOWPOS3IPROC glad_glWindowPos3i; -PFNGLMULTITEXCOORD4SPROC glad_glMultiTexCoord4s; -PFNGLWINDOWPOS3FPROC glad_glWindowPos3f; -PFNGLCOLOR3USPROC glad_glColor3us; -PFNGLCOLOR3UIVPROC glad_glColor3uiv; -PFNGLVERTEXATTRIB4NUSVPROC glad_glVertexAttrib4Nusv; -PFNGLGETLIGHTIVPROC glad_glGetLightiv; -PFNGLDEPTHFUNCPROC glad_glDepthFunc; -PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D; -PFNGLLISTBASEPROC glad_glListBase; -PFNGLMULTITEXCOORD4FPROC glad_glMultiTexCoord4f; -PFNGLCOLOR3UBPROC glad_glColor3ub; -PFNGLMULTITEXCOORD4DPROC glad_glMultiTexCoord4d; -PFNGLVERTEXATTRIBI4BVPROC glad_glVertexAttribI4bv; -PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv; -PFNGLCOLOR3UIPROC glad_glColor3ui; -PFNGLMULTITEXCOORD4IPROC glad_glMultiTexCoord4i; -PFNGLGETPOLYGONSTIPPLEPROC glad_glGetPolygonStipple; -PFNGLCLIENTWAITSYNCPROC glad_glClientWaitSync; -PFNGLVERTEXATTRIBI4UIPROC glad_glVertexAttribI4ui; -PFNGLMULTITEXCOORD4DVPROC glad_glMultiTexCoord4dv; -PFNGLCOLORMASKPROC glad_glColorMask; -PFNGLTEXPARAMETERIIVPROC glad_glTexParameterIiv; -PFNGLBLENDEQUATIONPROC glad_glBlendEquation; -PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation; -PFNGLGETSAMPLERPARAMETERIVPROC glad_glGetSamplerParameteriv; -PFNGLRASTERPOS4SPROC glad_glRasterPos4s; -PFNGLENDTRANSFORMFEEDBACKPROC glad_glEndTransformFeedback; -PFNGLVERTEXATTRIB4USVPROC glad_glVertexAttrib4usv; -PFNGLMULTITEXCOORD3DVPROC glad_glMultiTexCoord3dv; -PFNGLCOLOR4SVPROC glad_glColor4sv; -PFNGLPOPCLIENTATTRIBPROC glad_glPopClientAttrib; -PFNGLBEGINTRANSFORMFEEDBACKPROC glad_glBeginTransformFeedback; -PFNGLFOGFPROC glad_glFogf; -PFNGLVERTEXATTRIBI1IVPROC glad_glVertexAttribI1iv; -PFNGLISSAMPLERPROC glad_glIsSampler; -PFNGLVERTEXP3UIPROC glad_glVertexP3ui; -PFNGLVERTEXATTRIBDIVISORPROC glad_glVertexAttribDivisor; -PFNGLCOLOR3IVPROC glad_glColor3iv; -PFNGLCOMPRESSEDTEXIMAGE1DPROC glad_glCompressedTexImage1D; -PFNGLCOPYTEXSUBIMAGE1DPROC glad_glCopyTexSubImage1D; -PFNGLTEXCOORD1IPROC glad_glTexCoord1i; -PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus; -PFNGLTEXCOORD1DPROC glad_glTexCoord1d; -PFNGLTEXCOORD1FPROC glad_glTexCoord1f; -PFNGLENDCONDITIONALRENDERPROC glad_glEndConditionalRender; -PFNGLENABLECLIENTSTATEPROC glad_glEnableClientState; -PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation; -PFNGLUNIFORMMATRIX4X2FVPROC glad_glUniformMatrix4x2fv; -PFNGLMULTITEXCOORD2SVPROC glad_glMultiTexCoord2sv; -PFNGLVERTEXATTRIB1DVPROC glad_glVertexAttrib1dv; -PFNGLDRAWRANGEELEMENTSPROC glad_glDrawRangeElements; -PFNGLTEXCOORD1SPROC glad_glTexCoord1s; -PFNGLBINDBUFFERBASEPROC glad_glBindBufferBase; -PFNGLBUFFERSUBDATAPROC glad_glBufferSubData; -PFNGLVERTEXATTRIB4IVPROC glad_glVertexAttrib4iv; -PFNGLGENLISTSPROC glad_glGenLists; -PFNGLCOLOR3BVPROC glad_glColor3bv; -PFNGLMAPBUFFERRANGEPROC glad_glMapBufferRange; -PFNGLFRAMEBUFFERTEXTUREPROC glad_glFramebufferTexture; -PFNGLGETTEXGENDVPROC glad_glGetTexGendv; -PFNGLMULTIDRAWARRAYSPROC glad_glMultiDrawArrays; -PFNGLENDLISTPROC glad_glEndList; -PFNGLVERTEXP4UIVPROC glad_glVertexP4uiv; -PFNGLUNIFORM2UIPROC glad_glUniform2ui; -PFNGLVERTEXATTRIBI2IVPROC glad_glVertexAttribI2iv; -PFNGLCOLOR3USVPROC glad_glColor3usv; -PFNGLWINDOWPOS2FVPROC glad_glWindowPos2fv; -PFNGLDISABLEIPROC glad_glDisablei; -PFNGLINDEXMASKPROC glad_glIndexMask; -PFNGLPUSHCLIENTATTRIBPROC glad_glPushClientAttrib; -PFNGLSHADERSOURCEPROC glad_glShaderSource; -PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC glad_glGetActiveUniformBlockName; -PFNGLVERTEXATTRIBI3UIVPROC glad_glVertexAttribI3uiv; -PFNGLCLEARACCUMPROC glad_glClearAccum; -PFNGLGETSYNCIVPROC glad_glGetSynciv; -PFNGLTEXCOORDP2UIVPROC glad_glTexCoordP2uiv; -PFNGLUNIFORM2FPROC glad_glUniform2f; -PFNGLBEGINQUERYPROC glad_glBeginQuery; -PFNGLGETUNIFORMBLOCKINDEXPROC glad_glGetUniformBlockIndex; -PFNGLBINDBUFFERPROC glad_glBindBuffer; -PFNGLMAP2DPROC glad_glMap2d; -PFNGLMAP2FPROC glad_glMap2f; -PFNGLVERTEX4DPROC glad_glVertex4d; -PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv; -PFNGLTEXCOORD1SVPROC glad_glTexCoord1sv; -PFNGLBUFFERDATAPROC glad_glBufferData; -PFNGLEVALPOINT1PROC glad_glEvalPoint1; -PFNGLGETTEXPARAMETERIIVPROC glad_glGetTexParameterIiv; -PFNGLTEXCOORD1DVPROC glad_glTexCoord1dv; -PFNGLTEXCOORDP1UIPROC glad_glTexCoordP1ui; -PFNGLGETERRORPROC glad_glGetError; -PFNGLGETTEXENVIVPROC glad_glGetTexEnviv; -PFNGLGETPROGRAMIVPROC glad_glGetProgramiv; -PFNGLVERTEXATTRIBP2UIPROC glad_glVertexAttribP2ui; -PFNGLGETFLOATVPROC glad_glGetFloatv; -PFNGLTEXSUBIMAGE1DPROC glad_glTexSubImage1D; -PFNGLMULTITEXCOORD2FVPROC glad_glMultiTexCoord2fv; -PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv; -PFNGLEVALCOORD1DPROC glad_glEvalCoord1d; -PFNGLGETTEXLEVELPARAMETERFVPROC glad_glGetTexLevelParameterfv; -PFNGLEVALCOORD1FPROC glad_glEvalCoord1f; -PFNGLPIXELMAPFVPROC glad_glPixelMapfv; -PFNGLVERTEXATTRIBP3UIVPROC glad_glVertexAttribP3uiv; -PFNGLGETPIXELMAPUSVPROC glad_glGetPixelMapusv; -PFNGLSECONDARYCOLORP3UIVPROC glad_glSecondaryColorP3uiv; -PFNGLGETINTEGERVPROC glad_glGetIntegerv; -PFNGLACCUMPROC glad_glAccum; -PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv; -PFNGLGETVERTEXATTRIBIIVPROC glad_glGetVertexAttribIiv; -PFNGLRASTERPOS4DVPROC glad_glRasterPos4dv; -PFNGLTEXCOORD2IVPROC glad_glTexCoord2iv; -PFNGLISQUERYPROC glad_glIsQuery; -PFNGLVERTEXATTRIB4SVPROC glad_glVertexAttrib4sv; -PFNGLWINDOWPOS3DVPROC glad_glWindowPos3dv; -PFNGLTEXIMAGE2DPROC glad_glTexImage2D; -PFNGLSTENCILMASKPROC glad_glStencilMask; -PFNGLDRAWPIXELSPROC glad_glDrawPixels; -PFNGLMULTMATRIXDPROC glad_glMultMatrixd; -PFNGLMULTMATRIXFPROC glad_glMultMatrixf; -PFNGLISTEXTUREPROC glad_glIsTexture; -PFNGLGETMATERIALIVPROC glad_glGetMaterialiv; -PFNGLUNIFORM1FVPROC glad_glUniform1fv; -PFNGLLOADMATRIXFPROC glad_glLoadMatrixf; -PFNGLLOADMATRIXDPROC glad_glLoadMatrixd; -PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv; -PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv; -PFNGLVERTEX4FPROC glad_glVertex4f; -PFNGLRECTSVPROC glad_glRectsv; -PFNGLCOLOR4USVPROC glad_glColor4usv; -PFNGLPOLYGONSTIPPLEPROC glad_glPolygonStipple; -PFNGLINTERLEAVEDARRAYSPROC glad_glInterleavedArrays; -PFNGLNORMAL3IPROC glad_glNormal3i; -PFNGLNORMAL3FPROC glad_glNormal3f; -PFNGLNORMAL3DPROC glad_glNormal3d; -PFNGLNORMAL3BPROC glad_glNormal3b; -PFNGLPIXELMAPUSVPROC glad_glPixelMapusv; -PFNGLGETTEXGENIVPROC glad_glGetTexGeniv; -PFNGLARRAYELEMENTPROC glad_glArrayElement; -PFNGLCOPYBUFFERSUBDATAPROC glad_glCopyBufferSubData; -PFNGLVERTEXATTRIBI1UIVPROC glad_glVertexAttribI1uiv; -PFNGLVERTEXATTRIB2DPROC glad_glVertexAttrib2d; -PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f; -PFNGLVERTEXATTRIB3DVPROC glad_glVertexAttrib3dv; -PFNGLGETQUERYOBJECTUI64VPROC glad_glGetQueryObjectui64v; -PFNGLDEPTHMASKPROC glad_glDepthMask; -PFNGLVERTEXATTRIB2SPROC glad_glVertexAttrib2s; -PFNGLCOLOR3FVPROC glad_glColor3fv; -PFNGLTEXIMAGE3DMULTISAMPLEPROC glad_glTexImage3DMultisample; -PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv; -PFNGLUNIFORM4FVPROC glad_glUniform4fv; -PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform; -PFNGLCOLORPOINTERPROC glad_glColorPointer; -PFNGLFRONTFACEPROC glad_glFrontFace; -PFNGLGETBOOLEANI_VPROC glad_glGetBooleani_v; -PFNGLCLEARBUFFERUIVPROC glad_glClearBufferuiv; -int GLAD_GL_ARB_framebuffer_object; -int GLAD_GL_EXT_framebuffer_object; -int GLAD_GL_ARB_debug_output; -PFNGLDEBUGMESSAGECONTROLARBPROC glad_glDebugMessageControlARB; -PFNGLDEBUGMESSAGEINSERTARBPROC glad_glDebugMessageInsertARB; -PFNGLDEBUGMESSAGECALLBACKARBPROC glad_glDebugMessageCallbackARB; -PFNGLGETDEBUGMESSAGELOGARBPROC glad_glGetDebugMessageLogARB; -PFNGLISRENDERBUFFEREXTPROC glad_glIsRenderbufferEXT; -PFNGLBINDRENDERBUFFEREXTPROC glad_glBindRenderbufferEXT; -PFNGLDELETERENDERBUFFERSEXTPROC glad_glDeleteRenderbuffersEXT; -PFNGLGENRENDERBUFFERSEXTPROC glad_glGenRenderbuffersEXT; -PFNGLRENDERBUFFERSTORAGEEXTPROC glad_glRenderbufferStorageEXT; -PFNGLGETRENDERBUFFERPARAMETERIVEXTPROC glad_glGetRenderbufferParameterivEXT; -PFNGLISFRAMEBUFFEREXTPROC glad_glIsFramebufferEXT; -PFNGLBINDFRAMEBUFFEREXTPROC glad_glBindFramebufferEXT; -PFNGLDELETEFRAMEBUFFERSEXTPROC glad_glDeleteFramebuffersEXT; -PFNGLGENFRAMEBUFFERSEXTPROC glad_glGenFramebuffersEXT; -PFNGLCHECKFRAMEBUFFERSTATUSEXTPROC glad_glCheckFramebufferStatusEXT; -PFNGLFRAMEBUFFERTEXTURE1DEXTPROC glad_glFramebufferTexture1DEXT; -PFNGLFRAMEBUFFERTEXTURE2DEXTPROC glad_glFramebufferTexture2DEXT; -PFNGLFRAMEBUFFERTEXTURE3DEXTPROC glad_glFramebufferTexture3DEXT; -PFNGLFRAMEBUFFERRENDERBUFFEREXTPROC glad_glFramebufferRenderbufferEXT; -PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVEXTPROC glad_glGetFramebufferAttachmentParameterivEXT; -PFNGLGENERATEMIPMAPEXTPROC glad_glGenerateMipmapEXT; +int GLAD_GL_VERSION_1_0 = 0; +int GLAD_GL_VERSION_1_1 = 0; +int GLAD_GL_VERSION_1_2 = 0; +int GLAD_GL_VERSION_1_3 = 0; +int GLAD_GL_VERSION_1_4 = 0; +int GLAD_GL_VERSION_1_5 = 0; +int GLAD_GL_VERSION_2_0 = 0; +int GLAD_GL_VERSION_2_1 = 0; +int GLAD_GL_VERSION_3_0 = 0; +int GLAD_GL_VERSION_3_1 = 0; +int GLAD_GL_VERSION_3_2 = 0; +int GLAD_GL_VERSION_3_3 = 0; +PFNGLACCUMPROC glad_glAccum = NULL; +PFNGLACTIVETEXTUREPROC glad_glActiveTexture = NULL; +PFNGLALPHAFUNCPROC glad_glAlphaFunc = NULL; +PFNGLARETEXTURESRESIDENTPROC glad_glAreTexturesResident = NULL; +PFNGLARRAYELEMENTPROC glad_glArrayElement = NULL; +PFNGLATTACHSHADERPROC glad_glAttachShader = NULL; +PFNGLBEGINPROC glad_glBegin = NULL; +PFNGLBEGINCONDITIONALRENDERPROC glad_glBeginConditionalRender = NULL; +PFNGLBEGINQUERYPROC glad_glBeginQuery = NULL; +PFNGLBEGINTRANSFORMFEEDBACKPROC glad_glBeginTransformFeedback = NULL; +PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation = NULL; +PFNGLBINDBUFFERPROC glad_glBindBuffer = NULL; +PFNGLBINDBUFFERBASEPROC glad_glBindBufferBase = NULL; +PFNGLBINDBUFFERRANGEPROC glad_glBindBufferRange = NULL; +PFNGLBINDFRAGDATALOCATIONPROC glad_glBindFragDataLocation = NULL; +PFNGLBINDFRAGDATALOCATIONINDEXEDPROC glad_glBindFragDataLocationIndexed = NULL; +PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer = NULL; +PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer = NULL; +PFNGLBINDSAMPLERPROC glad_glBindSampler = NULL; +PFNGLBINDTEXTUREPROC glad_glBindTexture = NULL; +PFNGLBINDVERTEXARRAYPROC glad_glBindVertexArray = NULL; +PFNGLBITMAPPROC glad_glBitmap = NULL; +PFNGLBLENDCOLORPROC glad_glBlendColor = NULL; +PFNGLBLENDEQUATIONPROC glad_glBlendEquation = NULL; +PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate = NULL; +PFNGLBLENDFUNCPROC glad_glBlendFunc = NULL; +PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate = NULL; +PFNGLBLITFRAMEBUFFERPROC glad_glBlitFramebuffer = NULL; +PFNGLBUFFERDATAPROC glad_glBufferData = NULL; +PFNGLBUFFERSUBDATAPROC glad_glBufferSubData = NULL; +PFNGLCALLLISTPROC glad_glCallList = NULL; +PFNGLCALLLISTSPROC glad_glCallLists = NULL; +PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus = NULL; +PFNGLCLAMPCOLORPROC glad_glClampColor = NULL; +PFNGLCLEARPROC glad_glClear = NULL; +PFNGLCLEARACCUMPROC glad_glClearAccum = NULL; +PFNGLCLEARBUFFERFIPROC glad_glClearBufferfi = NULL; +PFNGLCLEARBUFFERFVPROC glad_glClearBufferfv = NULL; +PFNGLCLEARBUFFERIVPROC glad_glClearBufferiv = NULL; +PFNGLCLEARBUFFERUIVPROC glad_glClearBufferuiv = NULL; +PFNGLCLEARCOLORPROC glad_glClearColor = NULL; +PFNGLCLEARDEPTHPROC glad_glClearDepth = NULL; +PFNGLCLEARINDEXPROC glad_glClearIndex = NULL; +PFNGLCLEARSTENCILPROC glad_glClearStencil = NULL; +PFNGLCLIENTACTIVETEXTUREPROC glad_glClientActiveTexture = NULL; +PFNGLCLIENTWAITSYNCPROC glad_glClientWaitSync = NULL; +PFNGLCLIPPLANEPROC glad_glClipPlane = NULL; +PFNGLCOLOR3BPROC glad_glColor3b = NULL; +PFNGLCOLOR3BVPROC glad_glColor3bv = NULL; +PFNGLCOLOR3DPROC glad_glColor3d = NULL; +PFNGLCOLOR3DVPROC glad_glColor3dv = NULL; +PFNGLCOLOR3FPROC glad_glColor3f = NULL; +PFNGLCOLOR3FVPROC glad_glColor3fv = NULL; +PFNGLCOLOR3IPROC glad_glColor3i = NULL; +PFNGLCOLOR3IVPROC glad_glColor3iv = NULL; +PFNGLCOLOR3SPROC glad_glColor3s = NULL; +PFNGLCOLOR3SVPROC glad_glColor3sv = NULL; +PFNGLCOLOR3UBPROC glad_glColor3ub = NULL; +PFNGLCOLOR3UBVPROC glad_glColor3ubv = NULL; +PFNGLCOLOR3UIPROC glad_glColor3ui = NULL; +PFNGLCOLOR3UIVPROC glad_glColor3uiv = NULL; +PFNGLCOLOR3USPROC glad_glColor3us = NULL; +PFNGLCOLOR3USVPROC glad_glColor3usv = NULL; +PFNGLCOLOR4BPROC glad_glColor4b = NULL; +PFNGLCOLOR4BVPROC glad_glColor4bv = NULL; +PFNGLCOLOR4DPROC glad_glColor4d = NULL; +PFNGLCOLOR4DVPROC glad_glColor4dv = NULL; +PFNGLCOLOR4FPROC glad_glColor4f = NULL; +PFNGLCOLOR4FVPROC glad_glColor4fv = NULL; +PFNGLCOLOR4IPROC glad_glColor4i = NULL; +PFNGLCOLOR4IVPROC glad_glColor4iv = NULL; +PFNGLCOLOR4SPROC glad_glColor4s = NULL; +PFNGLCOLOR4SVPROC glad_glColor4sv = NULL; +PFNGLCOLOR4UBPROC glad_glColor4ub = NULL; +PFNGLCOLOR4UBVPROC glad_glColor4ubv = NULL; +PFNGLCOLOR4UIPROC glad_glColor4ui = NULL; +PFNGLCOLOR4UIVPROC glad_glColor4uiv = NULL; +PFNGLCOLOR4USPROC glad_glColor4us = NULL; +PFNGLCOLOR4USVPROC glad_glColor4usv = NULL; +PFNGLCOLORMASKPROC glad_glColorMask = NULL; +PFNGLCOLORMASKIPROC glad_glColorMaski = NULL; +PFNGLCOLORMATERIALPROC glad_glColorMaterial = NULL; +PFNGLCOLORP3UIPROC glad_glColorP3ui = NULL; +PFNGLCOLORP3UIVPROC glad_glColorP3uiv = NULL; +PFNGLCOLORP4UIPROC glad_glColorP4ui = NULL; +PFNGLCOLORP4UIVPROC glad_glColorP4uiv = NULL; +PFNGLCOLORPOINTERPROC glad_glColorPointer = NULL; +PFNGLCOMPILESHADERPROC glad_glCompileShader = NULL; +PFNGLCOMPRESSEDTEXIMAGE1DPROC glad_glCompressedTexImage1D = NULL; +PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D = NULL; +PFNGLCOMPRESSEDTEXIMAGE3DPROC glad_glCompressedTexImage3D = NULL; +PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC glad_glCompressedTexSubImage1D = NULL; +PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D = NULL; +PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC glad_glCompressedTexSubImage3D = NULL; +PFNGLCOPYBUFFERSUBDATAPROC glad_glCopyBufferSubData = NULL; +PFNGLCOPYPIXELSPROC glad_glCopyPixels = NULL; +PFNGLCOPYTEXIMAGE1DPROC glad_glCopyTexImage1D = NULL; +PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D = NULL; +PFNGLCOPYTEXSUBIMAGE1DPROC glad_glCopyTexSubImage1D = NULL; +PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D = NULL; +PFNGLCOPYTEXSUBIMAGE3DPROC glad_glCopyTexSubImage3D = NULL; +PFNGLCREATEPROGRAMPROC glad_glCreateProgram = NULL; +PFNGLCREATESHADERPROC glad_glCreateShader = NULL; +PFNGLCULLFACEPROC glad_glCullFace = NULL; +PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers = NULL; +PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers = NULL; +PFNGLDELETELISTSPROC glad_glDeleteLists = NULL; +PFNGLDELETEPROGRAMPROC glad_glDeleteProgram = NULL; +PFNGLDELETEQUERIESPROC glad_glDeleteQueries = NULL; +PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers = NULL; +PFNGLDELETESAMPLERSPROC glad_glDeleteSamplers = NULL; +PFNGLDELETESHADERPROC glad_glDeleteShader = NULL; +PFNGLDELETESYNCPROC glad_glDeleteSync = NULL; +PFNGLDELETETEXTURESPROC glad_glDeleteTextures = NULL; +PFNGLDELETEVERTEXARRAYSPROC glad_glDeleteVertexArrays = NULL; +PFNGLDEPTHFUNCPROC glad_glDepthFunc = NULL; +PFNGLDEPTHMASKPROC glad_glDepthMask = NULL; +PFNGLDEPTHRANGEPROC glad_glDepthRange = NULL; +PFNGLDETACHSHADERPROC glad_glDetachShader = NULL; +PFNGLDISABLEPROC glad_glDisable = NULL; +PFNGLDISABLECLIENTSTATEPROC glad_glDisableClientState = NULL; +PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray = NULL; +PFNGLDISABLEIPROC glad_glDisablei = NULL; +PFNGLDRAWARRAYSPROC glad_glDrawArrays = NULL; +PFNGLDRAWARRAYSINSTANCEDPROC glad_glDrawArraysInstanced = NULL; +PFNGLDRAWBUFFERPROC glad_glDrawBuffer = NULL; +PFNGLDRAWBUFFERSPROC glad_glDrawBuffers = NULL; +PFNGLDRAWELEMENTSPROC glad_glDrawElements = NULL; +PFNGLDRAWELEMENTSBASEVERTEXPROC glad_glDrawElementsBaseVertex = NULL; +PFNGLDRAWELEMENTSINSTANCEDPROC glad_glDrawElementsInstanced = NULL; +PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC glad_glDrawElementsInstancedBaseVertex = NULL; +PFNGLDRAWPIXELSPROC glad_glDrawPixels = NULL; +PFNGLDRAWRANGEELEMENTSPROC glad_glDrawRangeElements = NULL; +PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC glad_glDrawRangeElementsBaseVertex = NULL; +PFNGLEDGEFLAGPROC glad_glEdgeFlag = NULL; +PFNGLEDGEFLAGPOINTERPROC glad_glEdgeFlagPointer = NULL; +PFNGLEDGEFLAGVPROC glad_glEdgeFlagv = NULL; +PFNGLENABLEPROC glad_glEnable = NULL; +PFNGLENABLECLIENTSTATEPROC glad_glEnableClientState = NULL; +PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray = NULL; +PFNGLENABLEIPROC glad_glEnablei = NULL; +PFNGLENDPROC glad_glEnd = NULL; +PFNGLENDCONDITIONALRENDERPROC glad_glEndConditionalRender = NULL; +PFNGLENDLISTPROC glad_glEndList = NULL; +PFNGLENDQUERYPROC glad_glEndQuery = NULL; +PFNGLENDTRANSFORMFEEDBACKPROC glad_glEndTransformFeedback = NULL; +PFNGLEVALCOORD1DPROC glad_glEvalCoord1d = NULL; +PFNGLEVALCOORD1DVPROC glad_glEvalCoord1dv = NULL; +PFNGLEVALCOORD1FPROC glad_glEvalCoord1f = NULL; +PFNGLEVALCOORD1FVPROC glad_glEvalCoord1fv = NULL; +PFNGLEVALCOORD2DPROC glad_glEvalCoord2d = NULL; +PFNGLEVALCOORD2DVPROC glad_glEvalCoord2dv = NULL; +PFNGLEVALCOORD2FPROC glad_glEvalCoord2f = NULL; +PFNGLEVALCOORD2FVPROC glad_glEvalCoord2fv = NULL; +PFNGLEVALMESH1PROC glad_glEvalMesh1 = NULL; +PFNGLEVALMESH2PROC glad_glEvalMesh2 = NULL; +PFNGLEVALPOINT1PROC glad_glEvalPoint1 = NULL; +PFNGLEVALPOINT2PROC glad_glEvalPoint2 = NULL; +PFNGLFEEDBACKBUFFERPROC glad_glFeedbackBuffer = NULL; +PFNGLFENCESYNCPROC glad_glFenceSync = NULL; +PFNGLFINISHPROC glad_glFinish = NULL; +PFNGLFLUSHPROC glad_glFlush = NULL; +PFNGLFLUSHMAPPEDBUFFERRANGEPROC glad_glFlushMappedBufferRange = NULL; +PFNGLFOGCOORDPOINTERPROC glad_glFogCoordPointer = NULL; +PFNGLFOGCOORDDPROC glad_glFogCoordd = NULL; +PFNGLFOGCOORDDVPROC glad_glFogCoorddv = NULL; +PFNGLFOGCOORDFPROC glad_glFogCoordf = NULL; +PFNGLFOGCOORDFVPROC glad_glFogCoordfv = NULL; +PFNGLFOGFPROC glad_glFogf = NULL; +PFNGLFOGFVPROC glad_glFogfv = NULL; +PFNGLFOGIPROC glad_glFogi = NULL; +PFNGLFOGIVPROC glad_glFogiv = NULL; +PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer = NULL; +PFNGLFRAMEBUFFERTEXTUREPROC glad_glFramebufferTexture = NULL; +PFNGLFRAMEBUFFERTEXTURE1DPROC glad_glFramebufferTexture1D = NULL; +PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D = NULL; +PFNGLFRAMEBUFFERTEXTURE3DPROC glad_glFramebufferTexture3D = NULL; +PFNGLFRAMEBUFFERTEXTURELAYERPROC glad_glFramebufferTextureLayer = NULL; +PFNGLFRONTFACEPROC glad_glFrontFace = NULL; +PFNGLFRUSTUMPROC glad_glFrustum = NULL; +PFNGLGENBUFFERSPROC glad_glGenBuffers = NULL; +PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers = NULL; +PFNGLGENLISTSPROC glad_glGenLists = NULL; +PFNGLGENQUERIESPROC glad_glGenQueries = NULL; +PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers = NULL; +PFNGLGENSAMPLERSPROC glad_glGenSamplers = NULL; +PFNGLGENTEXTURESPROC glad_glGenTextures = NULL; +PFNGLGENVERTEXARRAYSPROC glad_glGenVertexArrays = NULL; +PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap = NULL; +PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib = NULL; +PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform = NULL; +PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC glad_glGetActiveUniformBlockName = NULL; +PFNGLGETACTIVEUNIFORMBLOCKIVPROC glad_glGetActiveUniformBlockiv = NULL; +PFNGLGETACTIVEUNIFORMNAMEPROC glad_glGetActiveUniformName = NULL; +PFNGLGETACTIVEUNIFORMSIVPROC glad_glGetActiveUniformsiv = NULL; +PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders = NULL; +PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation = NULL; +PFNGLGETBOOLEANI_VPROC glad_glGetBooleani_v = NULL; +PFNGLGETBOOLEANVPROC glad_glGetBooleanv = NULL; +PFNGLGETBUFFERPARAMETERI64VPROC glad_glGetBufferParameteri64v = NULL; +PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv = NULL; +PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv = NULL; +PFNGLGETBUFFERSUBDATAPROC glad_glGetBufferSubData = NULL; +PFNGLGETCLIPPLANEPROC glad_glGetClipPlane = NULL; +PFNGLGETCOMPRESSEDTEXIMAGEPROC glad_glGetCompressedTexImage = NULL; +PFNGLGETDOUBLEVPROC glad_glGetDoublev = NULL; +PFNGLGETERRORPROC glad_glGetError = NULL; +PFNGLGETFLOATVPROC glad_glGetFloatv = NULL; +PFNGLGETFRAGDATAINDEXPROC glad_glGetFragDataIndex = NULL; +PFNGLGETFRAGDATALOCATIONPROC glad_glGetFragDataLocation = NULL; +PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv = NULL; +PFNGLGETINTEGER64I_VPROC glad_glGetInteger64i_v = NULL; +PFNGLGETINTEGER64VPROC glad_glGetInteger64v = NULL; +PFNGLGETINTEGERI_VPROC glad_glGetIntegeri_v = NULL; +PFNGLGETINTEGERVPROC glad_glGetIntegerv = NULL; +PFNGLGETLIGHTFVPROC glad_glGetLightfv = NULL; +PFNGLGETLIGHTIVPROC glad_glGetLightiv = NULL; +PFNGLGETMAPDVPROC glad_glGetMapdv = NULL; +PFNGLGETMAPFVPROC glad_glGetMapfv = NULL; +PFNGLGETMAPIVPROC glad_glGetMapiv = NULL; +PFNGLGETMATERIALFVPROC glad_glGetMaterialfv = NULL; +PFNGLGETMATERIALIVPROC glad_glGetMaterialiv = NULL; +PFNGLGETMULTISAMPLEFVPROC glad_glGetMultisamplefv = NULL; +PFNGLGETPIXELMAPFVPROC glad_glGetPixelMapfv = NULL; +PFNGLGETPIXELMAPUIVPROC glad_glGetPixelMapuiv = NULL; +PFNGLGETPIXELMAPUSVPROC glad_glGetPixelMapusv = NULL; +PFNGLGETPOINTERVPROC glad_glGetPointerv = NULL; +PFNGLGETPOLYGONSTIPPLEPROC glad_glGetPolygonStipple = NULL; +PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog = NULL; +PFNGLGETPROGRAMIVPROC glad_glGetProgramiv = NULL; +PFNGLGETQUERYOBJECTI64VPROC glad_glGetQueryObjecti64v = NULL; +PFNGLGETQUERYOBJECTIVPROC glad_glGetQueryObjectiv = NULL; +PFNGLGETQUERYOBJECTUI64VPROC glad_glGetQueryObjectui64v = NULL; +PFNGLGETQUERYOBJECTUIVPROC glad_glGetQueryObjectuiv = NULL; +PFNGLGETQUERYIVPROC glad_glGetQueryiv = NULL; +PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv = NULL; +PFNGLGETSAMPLERPARAMETERIIVPROC glad_glGetSamplerParameterIiv = NULL; +PFNGLGETSAMPLERPARAMETERIUIVPROC glad_glGetSamplerParameterIuiv = NULL; +PFNGLGETSAMPLERPARAMETERFVPROC glad_glGetSamplerParameterfv = NULL; +PFNGLGETSAMPLERPARAMETERIVPROC glad_glGetSamplerParameteriv = NULL; +PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog = NULL; +PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource = NULL; +PFNGLGETSHADERIVPROC glad_glGetShaderiv = NULL; +PFNGLGETSTRINGPROC glad_glGetString = NULL; +PFNGLGETSTRINGIPROC glad_glGetStringi = NULL; +PFNGLGETSYNCIVPROC glad_glGetSynciv = NULL; +PFNGLGETTEXENVFVPROC glad_glGetTexEnvfv = NULL; +PFNGLGETTEXENVIVPROC glad_glGetTexEnviv = NULL; +PFNGLGETTEXGENDVPROC glad_glGetTexGendv = NULL; +PFNGLGETTEXGENFVPROC glad_glGetTexGenfv = NULL; +PFNGLGETTEXGENIVPROC glad_glGetTexGeniv = NULL; +PFNGLGETTEXIMAGEPROC glad_glGetTexImage = NULL; +PFNGLGETTEXLEVELPARAMETERFVPROC glad_glGetTexLevelParameterfv = NULL; +PFNGLGETTEXLEVELPARAMETERIVPROC glad_glGetTexLevelParameteriv = NULL; +PFNGLGETTEXPARAMETERIIVPROC glad_glGetTexParameterIiv = NULL; +PFNGLGETTEXPARAMETERIUIVPROC glad_glGetTexParameterIuiv = NULL; +PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv = NULL; +PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv = NULL; +PFNGLGETTRANSFORMFEEDBACKVARYINGPROC glad_glGetTransformFeedbackVarying = NULL; +PFNGLGETUNIFORMBLOCKINDEXPROC glad_glGetUniformBlockIndex = NULL; +PFNGLGETUNIFORMINDICESPROC glad_glGetUniformIndices = NULL; +PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation = NULL; +PFNGLGETUNIFORMFVPROC glad_glGetUniformfv = NULL; +PFNGLGETUNIFORMIVPROC glad_glGetUniformiv = NULL; +PFNGLGETUNIFORMUIVPROC glad_glGetUniformuiv = NULL; +PFNGLGETVERTEXATTRIBIIVPROC glad_glGetVertexAttribIiv = NULL; +PFNGLGETVERTEXATTRIBIUIVPROC glad_glGetVertexAttribIuiv = NULL; +PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv = NULL; +PFNGLGETVERTEXATTRIBDVPROC glad_glGetVertexAttribdv = NULL; +PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv = NULL; +PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv = NULL; +PFNGLHINTPROC glad_glHint = NULL; +PFNGLINDEXMASKPROC glad_glIndexMask = NULL; +PFNGLINDEXPOINTERPROC glad_glIndexPointer = NULL; +PFNGLINDEXDPROC glad_glIndexd = NULL; +PFNGLINDEXDVPROC glad_glIndexdv = NULL; +PFNGLINDEXFPROC glad_glIndexf = NULL; +PFNGLINDEXFVPROC glad_glIndexfv = NULL; +PFNGLINDEXIPROC glad_glIndexi = NULL; +PFNGLINDEXIVPROC glad_glIndexiv = NULL; +PFNGLINDEXSPROC glad_glIndexs = NULL; +PFNGLINDEXSVPROC glad_glIndexsv = NULL; +PFNGLINDEXUBPROC glad_glIndexub = NULL; +PFNGLINDEXUBVPROC glad_glIndexubv = NULL; +PFNGLINITNAMESPROC glad_glInitNames = NULL; +PFNGLINTERLEAVEDARRAYSPROC glad_glInterleavedArrays = NULL; +PFNGLISBUFFERPROC glad_glIsBuffer = NULL; +PFNGLISENABLEDPROC glad_glIsEnabled = NULL; +PFNGLISENABLEDIPROC glad_glIsEnabledi = NULL; +PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer = NULL; +PFNGLISLISTPROC glad_glIsList = NULL; +PFNGLISPROGRAMPROC glad_glIsProgram = NULL; +PFNGLISQUERYPROC glad_glIsQuery = NULL; +PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer = NULL; +PFNGLISSAMPLERPROC glad_glIsSampler = NULL; +PFNGLISSHADERPROC glad_glIsShader = NULL; +PFNGLISSYNCPROC glad_glIsSync = NULL; +PFNGLISTEXTUREPROC glad_glIsTexture = NULL; +PFNGLISVERTEXARRAYPROC glad_glIsVertexArray = NULL; +PFNGLLIGHTMODELFPROC glad_glLightModelf = NULL; +PFNGLLIGHTMODELFVPROC glad_glLightModelfv = NULL; +PFNGLLIGHTMODELIPROC glad_glLightModeli = NULL; +PFNGLLIGHTMODELIVPROC glad_glLightModeliv = NULL; +PFNGLLIGHTFPROC glad_glLightf = NULL; +PFNGLLIGHTFVPROC glad_glLightfv = NULL; +PFNGLLIGHTIPROC glad_glLighti = NULL; +PFNGLLIGHTIVPROC glad_glLightiv = NULL; +PFNGLLINESTIPPLEPROC glad_glLineStipple = NULL; +PFNGLLINEWIDTHPROC glad_glLineWidth = NULL; +PFNGLLINKPROGRAMPROC glad_glLinkProgram = NULL; +PFNGLLISTBASEPROC glad_glListBase = NULL; +PFNGLLOADIDENTITYPROC glad_glLoadIdentity = NULL; +PFNGLLOADMATRIXDPROC glad_glLoadMatrixd = NULL; +PFNGLLOADMATRIXFPROC glad_glLoadMatrixf = NULL; +PFNGLLOADNAMEPROC glad_glLoadName = NULL; +PFNGLLOADTRANSPOSEMATRIXDPROC glad_glLoadTransposeMatrixd = NULL; +PFNGLLOADTRANSPOSEMATRIXFPROC glad_glLoadTransposeMatrixf = NULL; +PFNGLLOGICOPPROC glad_glLogicOp = NULL; +PFNGLMAP1DPROC glad_glMap1d = NULL; +PFNGLMAP1FPROC glad_glMap1f = NULL; +PFNGLMAP2DPROC glad_glMap2d = NULL; +PFNGLMAP2FPROC glad_glMap2f = NULL; +PFNGLMAPBUFFERPROC glad_glMapBuffer = NULL; +PFNGLMAPBUFFERRANGEPROC glad_glMapBufferRange = NULL; +PFNGLMAPGRID1DPROC glad_glMapGrid1d = NULL; +PFNGLMAPGRID1FPROC glad_glMapGrid1f = NULL; +PFNGLMAPGRID2DPROC glad_glMapGrid2d = NULL; +PFNGLMAPGRID2FPROC glad_glMapGrid2f = NULL; +PFNGLMATERIALFPROC glad_glMaterialf = NULL; +PFNGLMATERIALFVPROC glad_glMaterialfv = NULL; +PFNGLMATERIALIPROC glad_glMateriali = NULL; +PFNGLMATERIALIVPROC glad_glMaterialiv = NULL; +PFNGLMATRIXMODEPROC glad_glMatrixMode = NULL; +PFNGLMULTMATRIXDPROC glad_glMultMatrixd = NULL; +PFNGLMULTMATRIXFPROC glad_glMultMatrixf = NULL; +PFNGLMULTTRANSPOSEMATRIXDPROC glad_glMultTransposeMatrixd = NULL; +PFNGLMULTTRANSPOSEMATRIXFPROC glad_glMultTransposeMatrixf = NULL; +PFNGLMULTIDRAWARRAYSPROC glad_glMultiDrawArrays = NULL; +PFNGLMULTIDRAWELEMENTSPROC glad_glMultiDrawElements = NULL; +PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC glad_glMultiDrawElementsBaseVertex = NULL; +PFNGLMULTITEXCOORD1DPROC glad_glMultiTexCoord1d = NULL; +PFNGLMULTITEXCOORD1DVPROC glad_glMultiTexCoord1dv = NULL; +PFNGLMULTITEXCOORD1FPROC glad_glMultiTexCoord1f = NULL; +PFNGLMULTITEXCOORD1FVPROC glad_glMultiTexCoord1fv = NULL; +PFNGLMULTITEXCOORD1IPROC glad_glMultiTexCoord1i = NULL; +PFNGLMULTITEXCOORD1IVPROC glad_glMultiTexCoord1iv = NULL; +PFNGLMULTITEXCOORD1SPROC glad_glMultiTexCoord1s = NULL; +PFNGLMULTITEXCOORD1SVPROC glad_glMultiTexCoord1sv = NULL; +PFNGLMULTITEXCOORD2DPROC glad_glMultiTexCoord2d = NULL; +PFNGLMULTITEXCOORD2DVPROC glad_glMultiTexCoord2dv = NULL; +PFNGLMULTITEXCOORD2FPROC glad_glMultiTexCoord2f = NULL; +PFNGLMULTITEXCOORD2FVPROC glad_glMultiTexCoord2fv = NULL; +PFNGLMULTITEXCOORD2IPROC glad_glMultiTexCoord2i = NULL; +PFNGLMULTITEXCOORD2IVPROC glad_glMultiTexCoord2iv = NULL; +PFNGLMULTITEXCOORD2SPROC glad_glMultiTexCoord2s = NULL; +PFNGLMULTITEXCOORD2SVPROC glad_glMultiTexCoord2sv = NULL; +PFNGLMULTITEXCOORD3DPROC glad_glMultiTexCoord3d = NULL; +PFNGLMULTITEXCOORD3DVPROC glad_glMultiTexCoord3dv = NULL; +PFNGLMULTITEXCOORD3FPROC glad_glMultiTexCoord3f = NULL; +PFNGLMULTITEXCOORD3FVPROC glad_glMultiTexCoord3fv = NULL; +PFNGLMULTITEXCOORD3IPROC glad_glMultiTexCoord3i = NULL; +PFNGLMULTITEXCOORD3IVPROC glad_glMultiTexCoord3iv = NULL; +PFNGLMULTITEXCOORD3SPROC glad_glMultiTexCoord3s = NULL; +PFNGLMULTITEXCOORD3SVPROC glad_glMultiTexCoord3sv = NULL; +PFNGLMULTITEXCOORD4DPROC glad_glMultiTexCoord4d = NULL; +PFNGLMULTITEXCOORD4DVPROC glad_glMultiTexCoord4dv = NULL; +PFNGLMULTITEXCOORD4FPROC glad_glMultiTexCoord4f = NULL; +PFNGLMULTITEXCOORD4FVPROC glad_glMultiTexCoord4fv = NULL; +PFNGLMULTITEXCOORD4IPROC glad_glMultiTexCoord4i = NULL; +PFNGLMULTITEXCOORD4IVPROC glad_glMultiTexCoord4iv = NULL; +PFNGLMULTITEXCOORD4SPROC glad_glMultiTexCoord4s = NULL; +PFNGLMULTITEXCOORD4SVPROC glad_glMultiTexCoord4sv = NULL; +PFNGLMULTITEXCOORDP1UIPROC glad_glMultiTexCoordP1ui = NULL; +PFNGLMULTITEXCOORDP1UIVPROC glad_glMultiTexCoordP1uiv = NULL; +PFNGLMULTITEXCOORDP2UIPROC glad_glMultiTexCoordP2ui = NULL; +PFNGLMULTITEXCOORDP2UIVPROC glad_glMultiTexCoordP2uiv = NULL; +PFNGLMULTITEXCOORDP3UIPROC glad_glMultiTexCoordP3ui = NULL; +PFNGLMULTITEXCOORDP3UIVPROC glad_glMultiTexCoordP3uiv = NULL; +PFNGLMULTITEXCOORDP4UIPROC glad_glMultiTexCoordP4ui = NULL; +PFNGLMULTITEXCOORDP4UIVPROC glad_glMultiTexCoordP4uiv = NULL; +PFNGLNEWLISTPROC glad_glNewList = NULL; +PFNGLNORMAL3BPROC glad_glNormal3b = NULL; +PFNGLNORMAL3BVPROC glad_glNormal3bv = NULL; +PFNGLNORMAL3DPROC glad_glNormal3d = NULL; +PFNGLNORMAL3DVPROC glad_glNormal3dv = NULL; +PFNGLNORMAL3FPROC glad_glNormal3f = NULL; +PFNGLNORMAL3FVPROC glad_glNormal3fv = NULL; +PFNGLNORMAL3IPROC glad_glNormal3i = NULL; +PFNGLNORMAL3IVPROC glad_glNormal3iv = NULL; +PFNGLNORMAL3SPROC glad_glNormal3s = NULL; +PFNGLNORMAL3SVPROC glad_glNormal3sv = NULL; +PFNGLNORMALP3UIPROC glad_glNormalP3ui = NULL; +PFNGLNORMALP3UIVPROC glad_glNormalP3uiv = NULL; +PFNGLNORMALPOINTERPROC glad_glNormalPointer = NULL; +PFNGLORTHOPROC glad_glOrtho = NULL; +PFNGLPASSTHROUGHPROC glad_glPassThrough = NULL; +PFNGLPIXELMAPFVPROC glad_glPixelMapfv = NULL; +PFNGLPIXELMAPUIVPROC glad_glPixelMapuiv = NULL; +PFNGLPIXELMAPUSVPROC glad_glPixelMapusv = NULL; +PFNGLPIXELSTOREFPROC glad_glPixelStoref = NULL; +PFNGLPIXELSTOREIPROC glad_glPixelStorei = NULL; +PFNGLPIXELTRANSFERFPROC glad_glPixelTransferf = NULL; +PFNGLPIXELTRANSFERIPROC glad_glPixelTransferi = NULL; +PFNGLPIXELZOOMPROC glad_glPixelZoom = NULL; +PFNGLPOINTPARAMETERFPROC glad_glPointParameterf = NULL; +PFNGLPOINTPARAMETERFVPROC glad_glPointParameterfv = NULL; +PFNGLPOINTPARAMETERIPROC glad_glPointParameteri = NULL; +PFNGLPOINTPARAMETERIVPROC glad_glPointParameteriv = NULL; +PFNGLPOINTSIZEPROC glad_glPointSize = NULL; +PFNGLPOLYGONMODEPROC glad_glPolygonMode = NULL; +PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset = NULL; +PFNGLPOLYGONSTIPPLEPROC glad_glPolygonStipple = NULL; +PFNGLPOPATTRIBPROC glad_glPopAttrib = NULL; +PFNGLPOPCLIENTATTRIBPROC glad_glPopClientAttrib = NULL; +PFNGLPOPMATRIXPROC glad_glPopMatrix = NULL; +PFNGLPOPNAMEPROC glad_glPopName = NULL; +PFNGLPRIMITIVERESTARTINDEXPROC glad_glPrimitiveRestartIndex = NULL; +PFNGLPRIORITIZETEXTURESPROC glad_glPrioritizeTextures = NULL; +PFNGLPROVOKINGVERTEXPROC glad_glProvokingVertex = NULL; +PFNGLPUSHATTRIBPROC glad_glPushAttrib = NULL; +PFNGLPUSHCLIENTATTRIBPROC glad_glPushClientAttrib = NULL; +PFNGLPUSHMATRIXPROC glad_glPushMatrix = NULL; +PFNGLPUSHNAMEPROC glad_glPushName = NULL; +PFNGLQUERYCOUNTERPROC glad_glQueryCounter = NULL; +PFNGLRASTERPOS2DPROC glad_glRasterPos2d = NULL; +PFNGLRASTERPOS2DVPROC glad_glRasterPos2dv = NULL; +PFNGLRASTERPOS2FPROC glad_glRasterPos2f = NULL; +PFNGLRASTERPOS2FVPROC glad_glRasterPos2fv = NULL; +PFNGLRASTERPOS2IPROC glad_glRasterPos2i = NULL; +PFNGLRASTERPOS2IVPROC glad_glRasterPos2iv = NULL; +PFNGLRASTERPOS2SPROC glad_glRasterPos2s = NULL; +PFNGLRASTERPOS2SVPROC glad_glRasterPos2sv = NULL; +PFNGLRASTERPOS3DPROC glad_glRasterPos3d = NULL; +PFNGLRASTERPOS3DVPROC glad_glRasterPos3dv = NULL; +PFNGLRASTERPOS3FPROC glad_glRasterPos3f = NULL; +PFNGLRASTERPOS3FVPROC glad_glRasterPos3fv = NULL; +PFNGLRASTERPOS3IPROC glad_glRasterPos3i = NULL; +PFNGLRASTERPOS3IVPROC glad_glRasterPos3iv = NULL; +PFNGLRASTERPOS3SPROC glad_glRasterPos3s = NULL; +PFNGLRASTERPOS3SVPROC glad_glRasterPos3sv = NULL; +PFNGLRASTERPOS4DPROC glad_glRasterPos4d = NULL; +PFNGLRASTERPOS4DVPROC glad_glRasterPos4dv = NULL; +PFNGLRASTERPOS4FPROC glad_glRasterPos4f = NULL; +PFNGLRASTERPOS4FVPROC glad_glRasterPos4fv = NULL; +PFNGLRASTERPOS4IPROC glad_glRasterPos4i = NULL; +PFNGLRASTERPOS4IVPROC glad_glRasterPos4iv = NULL; +PFNGLRASTERPOS4SPROC glad_glRasterPos4s = NULL; +PFNGLRASTERPOS4SVPROC glad_glRasterPos4sv = NULL; +PFNGLREADBUFFERPROC glad_glReadBuffer = NULL; +PFNGLREADPIXELSPROC glad_glReadPixels = NULL; +PFNGLRECTDPROC glad_glRectd = NULL; +PFNGLRECTDVPROC glad_glRectdv = NULL; +PFNGLRECTFPROC glad_glRectf = NULL; +PFNGLRECTFVPROC glad_glRectfv = NULL; +PFNGLRECTIPROC glad_glRecti = NULL; +PFNGLRECTIVPROC glad_glRectiv = NULL; +PFNGLRECTSPROC glad_glRects = NULL; +PFNGLRECTSVPROC glad_glRectsv = NULL; +PFNGLRENDERMODEPROC glad_glRenderMode = NULL; +PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage = NULL; +PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glad_glRenderbufferStorageMultisample = NULL; +PFNGLROTATEDPROC glad_glRotated = NULL; +PFNGLROTATEFPROC glad_glRotatef = NULL; +PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage = NULL; +PFNGLSAMPLEMASKIPROC glad_glSampleMaski = NULL; +PFNGLSAMPLERPARAMETERIIVPROC glad_glSamplerParameterIiv = NULL; +PFNGLSAMPLERPARAMETERIUIVPROC glad_glSamplerParameterIuiv = NULL; +PFNGLSAMPLERPARAMETERFPROC glad_glSamplerParameterf = NULL; +PFNGLSAMPLERPARAMETERFVPROC glad_glSamplerParameterfv = NULL; +PFNGLSAMPLERPARAMETERIPROC glad_glSamplerParameteri = NULL; +PFNGLSAMPLERPARAMETERIVPROC glad_glSamplerParameteriv = NULL; +PFNGLSCALEDPROC glad_glScaled = NULL; +PFNGLSCALEFPROC glad_glScalef = NULL; +PFNGLSCISSORPROC glad_glScissor = NULL; +PFNGLSECONDARYCOLOR3BPROC glad_glSecondaryColor3b = NULL; +PFNGLSECONDARYCOLOR3BVPROC glad_glSecondaryColor3bv = NULL; +PFNGLSECONDARYCOLOR3DPROC glad_glSecondaryColor3d = NULL; +PFNGLSECONDARYCOLOR3DVPROC glad_glSecondaryColor3dv = NULL; +PFNGLSECONDARYCOLOR3FPROC glad_glSecondaryColor3f = NULL; +PFNGLSECONDARYCOLOR3FVPROC glad_glSecondaryColor3fv = NULL; +PFNGLSECONDARYCOLOR3IPROC glad_glSecondaryColor3i = NULL; +PFNGLSECONDARYCOLOR3IVPROC glad_glSecondaryColor3iv = NULL; +PFNGLSECONDARYCOLOR3SPROC glad_glSecondaryColor3s = NULL; +PFNGLSECONDARYCOLOR3SVPROC glad_glSecondaryColor3sv = NULL; +PFNGLSECONDARYCOLOR3UBPROC glad_glSecondaryColor3ub = NULL; +PFNGLSECONDARYCOLOR3UBVPROC glad_glSecondaryColor3ubv = NULL; +PFNGLSECONDARYCOLOR3UIPROC glad_glSecondaryColor3ui = NULL; +PFNGLSECONDARYCOLOR3UIVPROC glad_glSecondaryColor3uiv = NULL; +PFNGLSECONDARYCOLOR3USPROC glad_glSecondaryColor3us = NULL; +PFNGLSECONDARYCOLOR3USVPROC glad_glSecondaryColor3usv = NULL; +PFNGLSECONDARYCOLORP3UIPROC glad_glSecondaryColorP3ui = NULL; +PFNGLSECONDARYCOLORP3UIVPROC glad_glSecondaryColorP3uiv = NULL; +PFNGLSECONDARYCOLORPOINTERPROC glad_glSecondaryColorPointer = NULL; +PFNGLSELECTBUFFERPROC glad_glSelectBuffer = NULL; +PFNGLSHADEMODELPROC glad_glShadeModel = NULL; +PFNGLSHADERSOURCEPROC glad_glShaderSource = NULL; +PFNGLSTENCILFUNCPROC glad_glStencilFunc = NULL; +PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate = NULL; +PFNGLSTENCILMASKPROC glad_glStencilMask = NULL; +PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate = NULL; +PFNGLSTENCILOPPROC glad_glStencilOp = NULL; +PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate = NULL; +PFNGLTEXBUFFERPROC glad_glTexBuffer = NULL; +PFNGLTEXCOORD1DPROC glad_glTexCoord1d = NULL; +PFNGLTEXCOORD1DVPROC glad_glTexCoord1dv = NULL; +PFNGLTEXCOORD1FPROC glad_glTexCoord1f = NULL; +PFNGLTEXCOORD1FVPROC glad_glTexCoord1fv = NULL; +PFNGLTEXCOORD1IPROC glad_glTexCoord1i = NULL; +PFNGLTEXCOORD1IVPROC glad_glTexCoord1iv = NULL; +PFNGLTEXCOORD1SPROC glad_glTexCoord1s = NULL; +PFNGLTEXCOORD1SVPROC glad_glTexCoord1sv = NULL; +PFNGLTEXCOORD2DPROC glad_glTexCoord2d = NULL; +PFNGLTEXCOORD2DVPROC glad_glTexCoord2dv = NULL; +PFNGLTEXCOORD2FPROC glad_glTexCoord2f = NULL; +PFNGLTEXCOORD2FVPROC glad_glTexCoord2fv = NULL; +PFNGLTEXCOORD2IPROC glad_glTexCoord2i = NULL; +PFNGLTEXCOORD2IVPROC glad_glTexCoord2iv = NULL; +PFNGLTEXCOORD2SPROC glad_glTexCoord2s = NULL; +PFNGLTEXCOORD2SVPROC glad_glTexCoord2sv = NULL; +PFNGLTEXCOORD3DPROC glad_glTexCoord3d = NULL; +PFNGLTEXCOORD3DVPROC glad_glTexCoord3dv = NULL; +PFNGLTEXCOORD3FPROC glad_glTexCoord3f = NULL; +PFNGLTEXCOORD3FVPROC glad_glTexCoord3fv = NULL; +PFNGLTEXCOORD3IPROC glad_glTexCoord3i = NULL; +PFNGLTEXCOORD3IVPROC glad_glTexCoord3iv = NULL; +PFNGLTEXCOORD3SPROC glad_glTexCoord3s = NULL; +PFNGLTEXCOORD3SVPROC glad_glTexCoord3sv = NULL; +PFNGLTEXCOORD4DPROC glad_glTexCoord4d = NULL; +PFNGLTEXCOORD4DVPROC glad_glTexCoord4dv = NULL; +PFNGLTEXCOORD4FPROC glad_glTexCoord4f = NULL; +PFNGLTEXCOORD4FVPROC glad_glTexCoord4fv = NULL; +PFNGLTEXCOORD4IPROC glad_glTexCoord4i = NULL; +PFNGLTEXCOORD4IVPROC glad_glTexCoord4iv = NULL; +PFNGLTEXCOORD4SPROC glad_glTexCoord4s = NULL; +PFNGLTEXCOORD4SVPROC glad_glTexCoord4sv = NULL; +PFNGLTEXCOORDP1UIPROC glad_glTexCoordP1ui = NULL; +PFNGLTEXCOORDP1UIVPROC glad_glTexCoordP1uiv = NULL; +PFNGLTEXCOORDP2UIPROC glad_glTexCoordP2ui = NULL; +PFNGLTEXCOORDP2UIVPROC glad_glTexCoordP2uiv = NULL; +PFNGLTEXCOORDP3UIPROC glad_glTexCoordP3ui = NULL; +PFNGLTEXCOORDP3UIVPROC glad_glTexCoordP3uiv = NULL; +PFNGLTEXCOORDP4UIPROC glad_glTexCoordP4ui = NULL; +PFNGLTEXCOORDP4UIVPROC glad_glTexCoordP4uiv = NULL; +PFNGLTEXCOORDPOINTERPROC glad_glTexCoordPointer = NULL; +PFNGLTEXENVFPROC glad_glTexEnvf = NULL; +PFNGLTEXENVFVPROC glad_glTexEnvfv = NULL; +PFNGLTEXENVIPROC glad_glTexEnvi = NULL; +PFNGLTEXENVIVPROC glad_glTexEnviv = NULL; +PFNGLTEXGENDPROC glad_glTexGend = NULL; +PFNGLTEXGENDVPROC glad_glTexGendv = NULL; +PFNGLTEXGENFPROC glad_glTexGenf = NULL; +PFNGLTEXGENFVPROC glad_glTexGenfv = NULL; +PFNGLTEXGENIPROC glad_glTexGeni = NULL; +PFNGLTEXGENIVPROC glad_glTexGeniv = NULL; +PFNGLTEXIMAGE1DPROC glad_glTexImage1D = NULL; +PFNGLTEXIMAGE2DPROC glad_glTexImage2D = NULL; +PFNGLTEXIMAGE2DMULTISAMPLEPROC glad_glTexImage2DMultisample = NULL; +PFNGLTEXIMAGE3DPROC glad_glTexImage3D = NULL; +PFNGLTEXIMAGE3DMULTISAMPLEPROC glad_glTexImage3DMultisample = NULL; +PFNGLTEXPARAMETERIIVPROC glad_glTexParameterIiv = NULL; +PFNGLTEXPARAMETERIUIVPROC glad_glTexParameterIuiv = NULL; +PFNGLTEXPARAMETERFPROC glad_glTexParameterf = NULL; +PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv = NULL; +PFNGLTEXPARAMETERIPROC glad_glTexParameteri = NULL; +PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv = NULL; +PFNGLTEXSUBIMAGE1DPROC glad_glTexSubImage1D = NULL; +PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D = NULL; +PFNGLTEXSUBIMAGE3DPROC glad_glTexSubImage3D = NULL; +PFNGLTRANSFORMFEEDBACKVARYINGSPROC glad_glTransformFeedbackVaryings = NULL; +PFNGLTRANSLATEDPROC glad_glTranslated = NULL; +PFNGLTRANSLATEFPROC glad_glTranslatef = NULL; +PFNGLUNIFORM1FPROC glad_glUniform1f = NULL; +PFNGLUNIFORM1FVPROC glad_glUniform1fv = NULL; +PFNGLUNIFORM1IPROC glad_glUniform1i = NULL; +PFNGLUNIFORM1IVPROC glad_glUniform1iv = NULL; +PFNGLUNIFORM1UIPROC glad_glUniform1ui = NULL; +PFNGLUNIFORM1UIVPROC glad_glUniform1uiv = NULL; +PFNGLUNIFORM2FPROC glad_glUniform2f = NULL; +PFNGLUNIFORM2FVPROC glad_glUniform2fv = NULL; +PFNGLUNIFORM2IPROC glad_glUniform2i = NULL; +PFNGLUNIFORM2IVPROC glad_glUniform2iv = NULL; +PFNGLUNIFORM2UIPROC glad_glUniform2ui = NULL; +PFNGLUNIFORM2UIVPROC glad_glUniform2uiv = NULL; +PFNGLUNIFORM3FPROC glad_glUniform3f = NULL; +PFNGLUNIFORM3FVPROC glad_glUniform3fv = NULL; +PFNGLUNIFORM3IPROC glad_glUniform3i = NULL; +PFNGLUNIFORM3IVPROC glad_glUniform3iv = NULL; +PFNGLUNIFORM3UIPROC glad_glUniform3ui = NULL; +PFNGLUNIFORM3UIVPROC glad_glUniform3uiv = NULL; +PFNGLUNIFORM4FPROC glad_glUniform4f = NULL; +PFNGLUNIFORM4FVPROC glad_glUniform4fv = NULL; +PFNGLUNIFORM4IPROC glad_glUniform4i = NULL; +PFNGLUNIFORM4IVPROC glad_glUniform4iv = NULL; +PFNGLUNIFORM4UIPROC glad_glUniform4ui = NULL; +PFNGLUNIFORM4UIVPROC glad_glUniform4uiv = NULL; +PFNGLUNIFORMBLOCKBINDINGPROC glad_glUniformBlockBinding = NULL; +PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv = NULL; +PFNGLUNIFORMMATRIX2X3FVPROC glad_glUniformMatrix2x3fv = NULL; +PFNGLUNIFORMMATRIX2X4FVPROC glad_glUniformMatrix2x4fv = NULL; +PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv = NULL; +PFNGLUNIFORMMATRIX3X2FVPROC glad_glUniformMatrix3x2fv = NULL; +PFNGLUNIFORMMATRIX3X4FVPROC glad_glUniformMatrix3x4fv = NULL; +PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv = NULL; +PFNGLUNIFORMMATRIX4X2FVPROC glad_glUniformMatrix4x2fv = NULL; +PFNGLUNIFORMMATRIX4X3FVPROC glad_glUniformMatrix4x3fv = NULL; +PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer = NULL; +PFNGLUSEPROGRAMPROC glad_glUseProgram = NULL; +PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram = NULL; +PFNGLVERTEX2DPROC glad_glVertex2d = NULL; +PFNGLVERTEX2DVPROC glad_glVertex2dv = NULL; +PFNGLVERTEX2FPROC glad_glVertex2f = NULL; +PFNGLVERTEX2FVPROC glad_glVertex2fv = NULL; +PFNGLVERTEX2IPROC glad_glVertex2i = NULL; +PFNGLVERTEX2IVPROC glad_glVertex2iv = NULL; +PFNGLVERTEX2SPROC glad_glVertex2s = NULL; +PFNGLVERTEX2SVPROC glad_glVertex2sv = NULL; +PFNGLVERTEX3DPROC glad_glVertex3d = NULL; +PFNGLVERTEX3DVPROC glad_glVertex3dv = NULL; +PFNGLVERTEX3FPROC glad_glVertex3f = NULL; +PFNGLVERTEX3FVPROC glad_glVertex3fv = NULL; +PFNGLVERTEX3IPROC glad_glVertex3i = NULL; +PFNGLVERTEX3IVPROC glad_glVertex3iv = NULL; +PFNGLVERTEX3SPROC glad_glVertex3s = NULL; +PFNGLVERTEX3SVPROC glad_glVertex3sv = NULL; +PFNGLVERTEX4DPROC glad_glVertex4d = NULL; +PFNGLVERTEX4DVPROC glad_glVertex4dv = NULL; +PFNGLVERTEX4FPROC glad_glVertex4f = NULL; +PFNGLVERTEX4FVPROC glad_glVertex4fv = NULL; +PFNGLVERTEX4IPROC glad_glVertex4i = NULL; +PFNGLVERTEX4IVPROC glad_glVertex4iv = NULL; +PFNGLVERTEX4SPROC glad_glVertex4s = NULL; +PFNGLVERTEX4SVPROC glad_glVertex4sv = NULL; +PFNGLVERTEXATTRIB1DPROC glad_glVertexAttrib1d = NULL; +PFNGLVERTEXATTRIB1DVPROC glad_glVertexAttrib1dv = NULL; +PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f = NULL; +PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv = NULL; +PFNGLVERTEXATTRIB1SPROC glad_glVertexAttrib1s = NULL; +PFNGLVERTEXATTRIB1SVPROC glad_glVertexAttrib1sv = NULL; +PFNGLVERTEXATTRIB2DPROC glad_glVertexAttrib2d = NULL; +PFNGLVERTEXATTRIB2DVPROC glad_glVertexAttrib2dv = NULL; +PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f = NULL; +PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv = NULL; +PFNGLVERTEXATTRIB2SPROC glad_glVertexAttrib2s = NULL; +PFNGLVERTEXATTRIB2SVPROC glad_glVertexAttrib2sv = NULL; +PFNGLVERTEXATTRIB3DPROC glad_glVertexAttrib3d = NULL; +PFNGLVERTEXATTRIB3DVPROC glad_glVertexAttrib3dv = NULL; +PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f = NULL; +PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv = NULL; +PFNGLVERTEXATTRIB3SPROC glad_glVertexAttrib3s = NULL; +PFNGLVERTEXATTRIB3SVPROC glad_glVertexAttrib3sv = NULL; +PFNGLVERTEXATTRIB4NBVPROC glad_glVertexAttrib4Nbv = NULL; +PFNGLVERTEXATTRIB4NIVPROC glad_glVertexAttrib4Niv = NULL; +PFNGLVERTEXATTRIB4NSVPROC glad_glVertexAttrib4Nsv = NULL; +PFNGLVERTEXATTRIB4NUBPROC glad_glVertexAttrib4Nub = NULL; +PFNGLVERTEXATTRIB4NUBVPROC glad_glVertexAttrib4Nubv = NULL; +PFNGLVERTEXATTRIB4NUIVPROC glad_glVertexAttrib4Nuiv = NULL; +PFNGLVERTEXATTRIB4NUSVPROC glad_glVertexAttrib4Nusv = NULL; +PFNGLVERTEXATTRIB4BVPROC glad_glVertexAttrib4bv = NULL; +PFNGLVERTEXATTRIB4DPROC glad_glVertexAttrib4d = NULL; +PFNGLVERTEXATTRIB4DVPROC glad_glVertexAttrib4dv = NULL; +PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f = NULL; +PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv = NULL; +PFNGLVERTEXATTRIB4IVPROC glad_glVertexAttrib4iv = NULL; +PFNGLVERTEXATTRIB4SPROC glad_glVertexAttrib4s = NULL; +PFNGLVERTEXATTRIB4SVPROC glad_glVertexAttrib4sv = NULL; +PFNGLVERTEXATTRIB4UBVPROC glad_glVertexAttrib4ubv = NULL; +PFNGLVERTEXATTRIB4UIVPROC glad_glVertexAttrib4uiv = NULL; +PFNGLVERTEXATTRIB4USVPROC glad_glVertexAttrib4usv = NULL; +PFNGLVERTEXATTRIBDIVISORPROC glad_glVertexAttribDivisor = NULL; +PFNGLVERTEXATTRIBI1IPROC glad_glVertexAttribI1i = NULL; +PFNGLVERTEXATTRIBI1IVPROC glad_glVertexAttribI1iv = NULL; +PFNGLVERTEXATTRIBI1UIPROC glad_glVertexAttribI1ui = NULL; +PFNGLVERTEXATTRIBI1UIVPROC glad_glVertexAttribI1uiv = NULL; +PFNGLVERTEXATTRIBI2IPROC glad_glVertexAttribI2i = NULL; +PFNGLVERTEXATTRIBI2IVPROC glad_glVertexAttribI2iv = NULL; +PFNGLVERTEXATTRIBI2UIPROC glad_glVertexAttribI2ui = NULL; +PFNGLVERTEXATTRIBI2UIVPROC glad_glVertexAttribI2uiv = NULL; +PFNGLVERTEXATTRIBI3IPROC glad_glVertexAttribI3i = NULL; +PFNGLVERTEXATTRIBI3IVPROC glad_glVertexAttribI3iv = NULL; +PFNGLVERTEXATTRIBI3UIPROC glad_glVertexAttribI3ui = NULL; +PFNGLVERTEXATTRIBI3UIVPROC glad_glVertexAttribI3uiv = NULL; +PFNGLVERTEXATTRIBI4BVPROC glad_glVertexAttribI4bv = NULL; +PFNGLVERTEXATTRIBI4IPROC glad_glVertexAttribI4i = NULL; +PFNGLVERTEXATTRIBI4IVPROC glad_glVertexAttribI4iv = NULL; +PFNGLVERTEXATTRIBI4SVPROC glad_glVertexAttribI4sv = NULL; +PFNGLVERTEXATTRIBI4UBVPROC glad_glVertexAttribI4ubv = NULL; +PFNGLVERTEXATTRIBI4UIPROC glad_glVertexAttribI4ui = NULL; +PFNGLVERTEXATTRIBI4UIVPROC glad_glVertexAttribI4uiv = NULL; +PFNGLVERTEXATTRIBI4USVPROC glad_glVertexAttribI4usv = NULL; +PFNGLVERTEXATTRIBIPOINTERPROC glad_glVertexAttribIPointer = NULL; +PFNGLVERTEXATTRIBP1UIPROC glad_glVertexAttribP1ui = NULL; +PFNGLVERTEXATTRIBP1UIVPROC glad_glVertexAttribP1uiv = NULL; +PFNGLVERTEXATTRIBP2UIPROC glad_glVertexAttribP2ui = NULL; +PFNGLVERTEXATTRIBP2UIVPROC glad_glVertexAttribP2uiv = NULL; +PFNGLVERTEXATTRIBP3UIPROC glad_glVertexAttribP3ui = NULL; +PFNGLVERTEXATTRIBP3UIVPROC glad_glVertexAttribP3uiv = NULL; +PFNGLVERTEXATTRIBP4UIPROC glad_glVertexAttribP4ui = NULL; +PFNGLVERTEXATTRIBP4UIVPROC glad_glVertexAttribP4uiv = NULL; +PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer = NULL; +PFNGLVERTEXP2UIPROC glad_glVertexP2ui = NULL; +PFNGLVERTEXP2UIVPROC glad_glVertexP2uiv = NULL; +PFNGLVERTEXP3UIPROC glad_glVertexP3ui = NULL; +PFNGLVERTEXP3UIVPROC glad_glVertexP3uiv = NULL; +PFNGLVERTEXP4UIPROC glad_glVertexP4ui = NULL; +PFNGLVERTEXP4UIVPROC glad_glVertexP4uiv = NULL; +PFNGLVERTEXPOINTERPROC glad_glVertexPointer = NULL; +PFNGLVIEWPORTPROC glad_glViewport = NULL; +PFNGLWAITSYNCPROC glad_glWaitSync = NULL; +PFNGLWINDOWPOS2DPROC glad_glWindowPos2d = NULL; +PFNGLWINDOWPOS2DVPROC glad_glWindowPos2dv = NULL; +PFNGLWINDOWPOS2FPROC glad_glWindowPos2f = NULL; +PFNGLWINDOWPOS2FVPROC glad_glWindowPos2fv = NULL; +PFNGLWINDOWPOS2IPROC glad_glWindowPos2i = NULL; +PFNGLWINDOWPOS2IVPROC glad_glWindowPos2iv = NULL; +PFNGLWINDOWPOS2SPROC glad_glWindowPos2s = NULL; +PFNGLWINDOWPOS2SVPROC glad_glWindowPos2sv = NULL; +PFNGLWINDOWPOS3DPROC glad_glWindowPos3d = NULL; +PFNGLWINDOWPOS3DVPROC glad_glWindowPos3dv = NULL; +PFNGLWINDOWPOS3FPROC glad_glWindowPos3f = NULL; +PFNGLWINDOWPOS3FVPROC glad_glWindowPos3fv = NULL; +PFNGLWINDOWPOS3IPROC glad_glWindowPos3i = NULL; +PFNGLWINDOWPOS3IVPROC glad_glWindowPos3iv = NULL; +PFNGLWINDOWPOS3SPROC glad_glWindowPos3s = NULL; +PFNGLWINDOWPOS3SVPROC glad_glWindowPos3sv = NULL; +int GLAD_GL_ARB_debug_output = 0; +int GLAD_GL_ARB_framebuffer_object = 0; +int GLAD_GL_EXT_framebuffer_object = 0; +PFNGLDEBUGMESSAGECONTROLARBPROC glad_glDebugMessageControlARB = NULL; +PFNGLDEBUGMESSAGEINSERTARBPROC glad_glDebugMessageInsertARB = NULL; +PFNGLDEBUGMESSAGECALLBACKARBPROC glad_glDebugMessageCallbackARB = NULL; +PFNGLGETDEBUGMESSAGELOGARBPROC glad_glGetDebugMessageLogARB = NULL; +PFNGLISRENDERBUFFEREXTPROC glad_glIsRenderbufferEXT = NULL; +PFNGLBINDRENDERBUFFEREXTPROC glad_glBindRenderbufferEXT = NULL; +PFNGLDELETERENDERBUFFERSEXTPROC glad_glDeleteRenderbuffersEXT = NULL; +PFNGLGENRENDERBUFFERSEXTPROC glad_glGenRenderbuffersEXT = NULL; +PFNGLRENDERBUFFERSTORAGEEXTPROC glad_glRenderbufferStorageEXT = NULL; +PFNGLGETRENDERBUFFERPARAMETERIVEXTPROC glad_glGetRenderbufferParameterivEXT = NULL; +PFNGLISFRAMEBUFFEREXTPROC glad_glIsFramebufferEXT = NULL; +PFNGLBINDFRAMEBUFFEREXTPROC glad_glBindFramebufferEXT = NULL; +PFNGLDELETEFRAMEBUFFERSEXTPROC glad_glDeleteFramebuffersEXT = NULL; +PFNGLGENFRAMEBUFFERSEXTPROC glad_glGenFramebuffersEXT = NULL; +PFNGLCHECKFRAMEBUFFERSTATUSEXTPROC glad_glCheckFramebufferStatusEXT = NULL; +PFNGLFRAMEBUFFERTEXTURE1DEXTPROC glad_glFramebufferTexture1DEXT = NULL; +PFNGLFRAMEBUFFERTEXTURE2DEXTPROC glad_glFramebufferTexture2DEXT = NULL; +PFNGLFRAMEBUFFERTEXTURE3DEXTPROC glad_glFramebufferTexture3DEXT = NULL; +PFNGLFRAMEBUFFERRENDERBUFFEREXTPROC glad_glFramebufferRenderbufferEXT = NULL; +PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVEXTPROC glad_glGetFramebufferAttachmentParameterivEXT = NULL; +PFNGLGENERATEMIPMAPEXTPROC glad_glGenerateMipmapEXT = NULL; static void load_GL_VERSION_1_0(GLADloadproc load) { if(!GLAD_GL_VERSION_1_0) return; glad_glCullFace = (PFNGLCULLFACEPROC)load("glCullFace"); diff --git a/thirdparty/glad/glad/glad.h b/thirdparty/glad/glad/glad.h index 4d92d33b37..52b05e0ae6 100644 --- a/thirdparty/glad/glad/glad.h +++ b/thirdparty/glad/glad/glad.h @@ -1,6 +1,6 @@ /* - OpenGL loader generated by glad 0.1.25 on Sat Jul 28 10:59:43 2018. + OpenGL loader generated by glad 0.1.28 on Thu Nov 22 16:50:04 2018. Language/Generator: C/C++ Specification: gl @@ -13,11 +13,12 @@ Loader: True Local files: False Omit khrplatform: False + Reproducible: False Commandline: --profile="compatibility" --api="gl=3.3" --generator="c" --spec="gl" --extensions="GL_ARB_debug_output,GL_ARB_framebuffer_object,GL_EXT_framebuffer_object" Online: - http://glad.dav1d.de/#profile=compatibility&language=c&specification=gl&loader=on&api=gl%3D3.3&extensions=GL_ARB_debug_output&extensions=GL_ARB_framebuffer_object&extensions=GL_EXT_framebuffer_object + https://glad.dav1d.de/#profile=compatibility&language=c&specification=gl&loader=on&api=gl%3D3.3&extensions=GL_ARB_debug_output&extensions=GL_ARB_framebuffer_object&extensions=GL_EXT_framebuffer_object */ @@ -46,6 +47,10 @@ #define APIENTRYP APIENTRY * #endif +#ifndef GLAPIENTRY +#define GLAPIENTRY APIENTRY +#endif + #ifdef __cplusplus extern "C" { #endif @@ -89,59 +94,21 @@ GLAPI int gladLoadGL(void); GLAPI int gladLoadGLLoader(GLADloadproc); -#include <stddef.h> #include <KHR/khrplatform.h> -#ifndef GLEXT_64_TYPES_DEFINED -/* This code block is duplicated in glxext.h, so must be protected */ -#define GLEXT_64_TYPES_DEFINED -/* Define int32_t, int64_t, and uint64_t types for UST/MSC */ -/* (as used in the GL_EXT_timer_query extension). */ -#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L -#include <inttypes.h> -#elif defined(__sun__) || defined(__digital__) -#include <inttypes.h> -#if defined(__STDC__) -#if defined(__arch64__) || defined(_LP64) -typedef long int int64_t; -typedef unsigned long int uint64_t; -#else -typedef long long int int64_t; -typedef unsigned long long int uint64_t; -#endif /* __arch64__ */ -#endif /* __STDC__ */ -#elif defined( __VMS ) || defined(__sgi) -#include <inttypes.h> -#elif defined(__SCO__) || defined(__USLC__) -#include <stdint.h> -#elif defined(__UNIXOS2__) || defined(__SOL64__) -typedef long int int32_t; -typedef long long int int64_t; -typedef unsigned long long int uint64_t; -#elif defined(_WIN32) && defined(__GNUC__) -#include <stdint.h> -#elif defined(_WIN32) -typedef __int32 int32_t; -typedef __int64 int64_t; -typedef unsigned __int64 uint64_t; -#else -/* Fallback if nothing above works */ -#include <inttypes.h> -#endif -#endif typedef unsigned int GLenum; typedef unsigned char GLboolean; typedef unsigned int GLbitfield; typedef void GLvoid; -typedef signed char GLbyte; -typedef short GLshort; +typedef khronos_int8_t GLbyte; +typedef khronos_uint8_t GLubyte; +typedef khronos_int16_t GLshort; +typedef khronos_uint16_t GLushort; typedef int GLint; -typedef int GLclampx; -typedef unsigned char GLubyte; -typedef unsigned short GLushort; typedef unsigned int GLuint; +typedef khronos_int32_t GLclampx; typedef int GLsizei; -typedef float GLfloat; -typedef float GLclampf; +typedef khronos_float_t GLfloat; +typedef khronos_float_t GLclampf; typedef double GLdouble; typedef double GLclampd; typedef void *GLeglClientBufferEXT; @@ -153,25 +120,17 @@ typedef void *GLhandleARB; #else typedef unsigned int GLhandleARB; #endif -typedef unsigned short GLhalfARB; -typedef unsigned short GLhalf; -typedef GLint GLfixed; +typedef khronos_uint16_t GLhalf; +typedef khronos_uint16_t GLhalfARB; +typedef khronos_int32_t GLfixed; typedef khronos_intptr_t GLintptr; +typedef khronos_intptr_t GLintptrARB; typedef khronos_ssize_t GLsizeiptr; -typedef int64_t GLint64; -typedef uint64_t GLuint64; -#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060) -typedef long GLintptrARB; -#else -typedef ptrdiff_t GLintptrARB; -#endif -#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060) -typedef long GLsizeiptrARB; -#else -typedef ptrdiff_t GLsizeiptrARB; -#endif -typedef int64_t GLint64EXT; -typedef uint64_t GLuint64EXT; +typedef khronos_ssize_t GLsizeiptrARB; +typedef khronos_int64_t GLint64; +typedef khronos_int64_t GLint64EXT; +typedef khronos_uint64_t GLuint64; +typedef khronos_uint64_t GLuint64EXT; typedef struct __GLsync *GLsync; struct _cl_context; struct _cl_event; diff --git a/thirdparty/libwebp/src/dec/alphai_dec.h b/thirdparty/libwebp/src/dec/alphai_dec.h index e0fa281a55..a64104abeb 100644 --- a/thirdparty/libwebp/src/dec/alphai_dec.h +++ b/thirdparty/libwebp/src/dec/alphai_dec.h @@ -51,4 +51,4 @@ void WebPDeallocateAlphaMemory(VP8Decoder* const dec); } // extern "C" #endif -#endif /* WEBP_DEC_ALPHAI_DEC_H_ */ +#endif // WEBP_DEC_ALPHAI_DEC_H_ diff --git a/thirdparty/libwebp/src/dec/buffer_dec.c b/thirdparty/libwebp/src/dec/buffer_dec.c index 75eb3c40b4..3cd94eb4d9 100644 --- a/thirdparty/libwebp/src/dec/buffer_dec.c +++ b/thirdparty/libwebp/src/dec/buffer_dec.c @@ -74,7 +74,8 @@ static VP8StatusCode CheckDecBuffer(const WebPDecBuffer* const buffer) { } else { // RGB checks const WebPRGBABuffer* const buf = &buffer->u.RGBA; const int stride = abs(buf->stride); - const uint64_t size = MIN_BUFFER_SIZE(width, height, stride); + const uint64_t size = + MIN_BUFFER_SIZE(width * kModeBpp[mode], height, stride); ok &= (size <= buf->size); ok &= (stride >= width * kModeBpp[mode]); ok &= (buf->rgba != NULL); diff --git a/thirdparty/libwebp/src/dec/common_dec.h b/thirdparty/libwebp/src/dec/common_dec.h index 9995f1a51a..b158550a80 100644 --- a/thirdparty/libwebp/src/dec/common_dec.h +++ b/thirdparty/libwebp/src/dec/common_dec.h @@ -51,4 +51,4 @@ enum { MB_FEATURE_TREE_PROBS = 3, NUM_PROBAS = 11 }; -#endif // WEBP_DEC_COMMON_DEC_H_ +#endif // WEBP_DEC_COMMON_DEC_H_ diff --git a/thirdparty/libwebp/src/dec/frame_dec.c b/thirdparty/libwebp/src/dec/frame_dec.c index a9d5430d00..bda9e1a6f6 100644 --- a/thirdparty/libwebp/src/dec/frame_dec.c +++ b/thirdparty/libwebp/src/dec/frame_dec.c @@ -338,7 +338,6 @@ void VP8InitDithering(const WebPDecoderOptions* const options, for (s = 0; s < NUM_MB_SEGMENTS; ++s) { VP8QuantMatrix* const dqm = &dec->dqm_[s]; if (dqm->uv_quant_ < DITHER_AMP_TAB_SIZE) { - // TODO(skal): should we specially dither more for uv_quant_ < 0? const int idx = (dqm->uv_quant_ < 0) ? 0 : dqm->uv_quant_; dqm->dither_ = (f * kQuantToDitherAmp[idx]) >> 3; } @@ -669,15 +668,9 @@ int VP8GetThreadMethod(const WebPDecoderOptions* const options, (void)height; assert(headers == NULL || !headers->is_lossless); #if defined(WEBP_USE_THREAD) - if (width < MIN_WIDTH_FOR_THREADS) return 0; - // TODO(skal): tune the heuristic further -#if 0 - if (height < 2 * width) return 2; + if (width >= MIN_WIDTH_FOR_THREADS) return 2; #endif - return 2; -#else // !WEBP_USE_THREAD return 0; -#endif } #undef MT_CACHE_LINES diff --git a/thirdparty/libwebp/src/dec/idec_dec.c b/thirdparty/libwebp/src/dec/idec_dec.c index a371ed7500..9bc9166808 100644 --- a/thirdparty/libwebp/src/dec/idec_dec.c +++ b/thirdparty/libwebp/src/dec/idec_dec.c @@ -140,10 +140,9 @@ static void DoRemap(WebPIDecoder* const idec, ptrdiff_t offset) { if (NeedCompressedAlpha(idec)) { ALPHDecoder* const alph_dec = dec->alph_dec_; dec->alpha_data_ += offset; - if (alph_dec != NULL) { + if (alph_dec != NULL && alph_dec->vp8l_dec_ != NULL) { if (alph_dec->method_ == ALPHA_LOSSLESS_COMPRESSION) { VP8LDecoder* const alph_vp8l_dec = alph_dec->vp8l_dec_; - assert(alph_vp8l_dec != NULL); assert(dec->alpha_data_size_ >= ALPHA_HEADER_LEN); VP8LBitReaderSetBuffer(&alph_vp8l_dec->br_, dec->alpha_data_ + ALPHA_HEADER_LEN, @@ -283,10 +282,8 @@ static void RestoreContext(const MBContext* context, VP8Decoder* const dec, static VP8StatusCode IDecError(WebPIDecoder* const idec, VP8StatusCode error) { if (idec->state_ == STATE_VP8_DATA) { - VP8Io* const io = &idec->io_; - if (io->teardown != NULL) { - io->teardown(io); - } + // Synchronize the thread, clean-up and check for errors. + VP8ExitCritical((VP8Decoder*)idec->dec_, &idec->io_); } idec->state_ = STATE_ERROR; return error; @@ -451,7 +448,10 @@ static VP8StatusCode DecodeRemaining(WebPIDecoder* const idec) { VP8Decoder* const dec = (VP8Decoder*)idec->dec_; VP8Io* const io = &idec->io_; - assert(dec->ready_); + // Make sure partition #0 has been read before, to set dec to ready_. + if (!dec->ready_) { + return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR); + } for (; dec->mb_y_ < dec->mb_h_; ++dec->mb_y_) { if (idec->last_mb_y_ != dec->mb_y_) { if (!VP8ParseIntraModeRow(&dec->br_, dec)) { @@ -473,6 +473,12 @@ static VP8StatusCode DecodeRemaining(WebPIDecoder* const idec) { MemDataSize(&idec->mem_) > MAX_MB_SIZE) { return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR); } + // Synchronize the threads. + if (dec->mt_method_ > 0) { + if (!WebPGetWorkerInterface()->Sync(&dec->worker_)) { + return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR); + } + } RestoreContext(&context, dec, token_br); return VP8_STATUS_SUSPENDED; } @@ -491,6 +497,7 @@ static VP8StatusCode DecodeRemaining(WebPIDecoder* const idec) { } // Synchronize the thread and check for errors. if (!VP8ExitCritical(dec, io)) { + idec->state_ = STATE_ERROR; // prevent re-entry in IDecError return IDecError(idec, VP8_STATUS_USER_ABORT); } dec->ready_ = 0; @@ -571,6 +578,10 @@ static VP8StatusCode IDecode(WebPIDecoder* idec) { status = DecodePartition0(idec); } if (idec->state_ == STATE_VP8_DATA) { + const VP8Decoder* const dec = (VP8Decoder*)idec->dec_; + if (dec == NULL) { + return VP8_STATUS_SUSPENDED; // can't continue if we have no decoder. + } status = DecodeRemaining(idec); } if (idec->state_ == STATE_VP8L_HEADER) { diff --git a/thirdparty/libwebp/src/dec/vp8_dec.h b/thirdparty/libwebp/src/dec/vp8_dec.h index ca85b340cf..a05405df72 100644 --- a/thirdparty/libwebp/src/dec/vp8_dec.h +++ b/thirdparty/libwebp/src/dec/vp8_dec.h @@ -182,4 +182,4 @@ WEBP_EXTERN int VP8LGetInfo( } // extern "C" #endif -#endif /* WEBP_DEC_VP8_DEC_H_ */ +#endif // WEBP_DEC_VP8_DEC_H_ diff --git a/thirdparty/libwebp/src/dec/vp8i_dec.h b/thirdparty/libwebp/src/dec/vp8i_dec.h index c929933e1c..e5e89df57d 100644 --- a/thirdparty/libwebp/src/dec/vp8i_dec.h +++ b/thirdparty/libwebp/src/dec/vp8i_dec.h @@ -32,7 +32,7 @@ extern "C" { // version numbers #define DEC_MAJ_VERSION 1 #define DEC_MIN_VERSION 0 -#define DEC_REV_VERSION 0 +#define DEC_REV_VERSION 1 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline). // Constraints are: We need to store one 16x16 block of luma samples (y), @@ -316,4 +316,4 @@ const uint8_t* VP8DecompressAlphaRows(VP8Decoder* const dec, } // extern "C" #endif -#endif /* WEBP_DEC_VP8I_DEC_H_ */ +#endif // WEBP_DEC_VP8I_DEC_H_ diff --git a/thirdparty/libwebp/src/dec/vp8l_dec.c b/thirdparty/libwebp/src/dec/vp8l_dec.c index 0570f53a77..333bb3e80d 100644 --- a/thirdparty/libwebp/src/dec/vp8l_dec.c +++ b/thirdparty/libwebp/src/dec/vp8l_dec.c @@ -362,12 +362,19 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize, VP8LMetadata* const hdr = &dec->hdr_; uint32_t* huffman_image = NULL; HTreeGroup* htree_groups = NULL; + // When reading htrees, some might be unused, as the format allows it. + // We will still read them but put them in this htree_group_bogus. + HTreeGroup htree_group_bogus; HuffmanCode* huffman_tables = NULL; + HuffmanCode* huffman_tables_bogus = NULL; HuffmanCode* next = NULL; int num_htree_groups = 1; + int num_htree_groups_max = 1; int max_alphabet_size = 0; int* code_lengths = NULL; const int table_size = kTableSize[color_cache_bits]; + int* mapping = NULL; + int ok = 0; if (allow_recursion && VP8LReadBits(br, 1)) { // use meta Huffman codes. @@ -384,9 +391,41 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize, // The huffman data is stored in red and green bytes. const int group = (huffman_image[i] >> 8) & 0xffff; huffman_image[i] = group; - if (group >= num_htree_groups) { - num_htree_groups = group + 1; + if (group >= num_htree_groups_max) { + num_htree_groups_max = group + 1; + } + } + // Check the validity of num_htree_groups_max. If it seems too big, use a + // smaller value for later. This will prevent big memory allocations to end + // up with a bad bitstream anyway. + // The value of 1000 is totally arbitrary. We know that num_htree_groups_max + // is smaller than (1 << 16) and should be smaller than the number of pixels + // (though the format allows it to be bigger). + if (num_htree_groups_max > 1000 || num_htree_groups_max > xsize * ysize) { + // Create a mapping from the used indices to the minimal set of used + // values [0, num_htree_groups) + mapping = (int*)WebPSafeMalloc(num_htree_groups_max, sizeof(*mapping)); + if (mapping == NULL) { + dec->status_ = VP8_STATUS_OUT_OF_MEMORY; + goto Error; + } + // -1 means a value is unmapped, and therefore unused in the Huffman + // image. + memset(mapping, 0xff, num_htree_groups_max * sizeof(*mapping)); + for (num_htree_groups = 0, i = 0; i < huffman_pixs; ++i) { + // Get the current mapping for the group and remap the Huffman image. + int* const mapped_group = &mapping[huffman_image[i]]; + if (*mapped_group == -1) *mapped_group = num_htree_groups++; + huffman_image[i] = *mapped_group; + } + huffman_tables_bogus = (HuffmanCode*)WebPSafeMalloc( + table_size, sizeof(*huffman_tables_bogus)); + if (huffman_tables_bogus == NULL) { + dec->status_ = VP8_STATUS_OUT_OF_MEMORY; + goto Error; } + } else { + num_htree_groups = num_htree_groups_max; } } @@ -403,11 +442,11 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize, } } + code_lengths = (int*)WebPSafeCalloc((uint64_t)max_alphabet_size, + sizeof(*code_lengths)); huffman_tables = (HuffmanCode*)WebPSafeMalloc(num_htree_groups * table_size, sizeof(*huffman_tables)); htree_groups = VP8LHtreeGroupsNew(num_htree_groups); - code_lengths = (int*)WebPSafeCalloc((uint64_t)max_alphabet_size, - sizeof(*code_lengths)); if (htree_groups == NULL || code_lengths == NULL || huffman_tables == NULL) { dec->status_ = VP8_STATUS_OUT_OF_MEMORY; @@ -415,28 +454,35 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize, } next = huffman_tables; - for (i = 0; i < num_htree_groups; ++i) { - HTreeGroup* const htree_group = &htree_groups[i]; + for (i = 0; i < num_htree_groups_max; ++i) { + // If the index "i" is unused in the Huffman image, read the coefficients + // but store them to a bogus htree_group. + const int is_bogus = (mapping != NULL && mapping[i] == -1); + HTreeGroup* const htree_group = + is_bogus ? &htree_group_bogus : + &htree_groups[(mapping == NULL) ? i : mapping[i]]; HuffmanCode** const htrees = htree_group->htrees; + HuffmanCode* huffman_tables_i = is_bogus ? huffman_tables_bogus : next; int size; int total_size = 0; int is_trivial_literal = 1; int max_bits = 0; for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) { int alphabet_size = kAlphabetSize[j]; - htrees[j] = next; + htrees[j] = huffman_tables_i; if (j == 0 && color_cache_bits > 0) { alphabet_size += 1 << color_cache_bits; } - size = ReadHuffmanCode(alphabet_size, dec, code_lengths, next); + size = + ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_tables_i); if (size == 0) { goto Error; } if (is_trivial_literal && kLiteralMap[j] == 1) { - is_trivial_literal = (next->bits == 0); + is_trivial_literal = (huffman_tables_i->bits == 0); } - total_size += next->bits; - next += size; + total_size += huffman_tables_i->bits; + huffman_tables_i += size; if (j <= ALPHA) { int local_max_bits = code_lengths[0]; int k; @@ -448,38 +494,41 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize, max_bits += local_max_bits; } } + if (!is_bogus) next = huffman_tables_i; htree_group->is_trivial_literal = is_trivial_literal; htree_group->is_trivial_code = 0; if (is_trivial_literal) { const int red = htrees[RED][0].value; const int blue = htrees[BLUE][0].value; const int alpha = htrees[ALPHA][0].value; - htree_group->literal_arb = - ((uint32_t)alpha << 24) | (red << 16) | blue; + htree_group->literal_arb = ((uint32_t)alpha << 24) | (red << 16) | blue; if (total_size == 0 && htrees[GREEN][0].value < NUM_LITERAL_CODES) { htree_group->is_trivial_code = 1; htree_group->literal_arb |= htrees[GREEN][0].value << 8; } } - htree_group->use_packed_table = !htree_group->is_trivial_code && - (max_bits < HUFFMAN_PACKED_BITS); + htree_group->use_packed_table = + !htree_group->is_trivial_code && (max_bits < HUFFMAN_PACKED_BITS); if (htree_group->use_packed_table) BuildPackedTable(htree_group); } - WebPSafeFree(code_lengths); + ok = 1; - // All OK. Finalize pointers and return. + // All OK. Finalize pointers. hdr->huffman_image_ = huffman_image; hdr->num_htree_groups_ = num_htree_groups; hdr->htree_groups_ = htree_groups; hdr->huffman_tables_ = huffman_tables; - return 1; Error: WebPSafeFree(code_lengths); - WebPSafeFree(huffman_image); - WebPSafeFree(huffman_tables); - VP8LHtreeGroupsFree(htree_groups); - return 0; + WebPSafeFree(huffman_tables_bogus); + WebPSafeFree(mapping); + if (!ok) { + WebPSafeFree(huffman_image); + WebPSafeFree(huffman_tables); + VP8LHtreeGroupsFree(htree_groups); + } + return ok; } //------------------------------------------------------------------------------ @@ -884,7 +933,11 @@ static WEBP_INLINE void CopyBlock8b(uint8_t* const dst, int dist, int length) { #endif break; case 2: +#if !defined(WORDS_BIGENDIAN) memcpy(&pattern, src, sizeof(uint16_t)); +#else + pattern = ((uint32_t)src[0] << 8) | src[1]; +#endif #if defined(__arm__) || defined(_M_ARM) pattern |= pattern << 16; #elif defined(WEBP_USE_MIPS_DSP_R2) @@ -1523,7 +1576,6 @@ int VP8LDecodeAlphaHeader(ALPHDecoder* const alph_dec, if (dec == NULL) return 0; assert(alph_dec != NULL); - alph_dec->vp8l_dec_ = dec; dec->width_ = alph_dec->width_; dec->height_ = alph_dec->height_; @@ -1555,11 +1607,12 @@ int VP8LDecodeAlphaHeader(ALPHDecoder* const alph_dec, if (!ok) goto Err; + // Only set here, once we are sure it is valid (to avoid thread races). + alph_dec->vp8l_dec_ = dec; return 1; Err: - VP8LDelete(alph_dec->vp8l_dec_); - alph_dec->vp8l_dec_ = NULL; + VP8LDelete(dec); return 0; } diff --git a/thirdparty/libwebp/src/dec/vp8li_dec.h b/thirdparty/libwebp/src/dec/vp8li_dec.h index 8e500cf9ff..0a4d613f99 100644 --- a/thirdparty/libwebp/src/dec/vp8li_dec.h +++ b/thirdparty/libwebp/src/dec/vp8li_dec.h @@ -132,4 +132,4 @@ void VP8LDelete(VP8LDecoder* const dec); } // extern "C" #endif -#endif /* WEBP_DEC_VP8LI_DEC_H_ */ +#endif // WEBP_DEC_VP8LI_DEC_H_ diff --git a/thirdparty/libwebp/src/dec/webpi_dec.h b/thirdparty/libwebp/src/dec/webpi_dec.h index c378ba6fc3..24baff5d27 100644 --- a/thirdparty/libwebp/src/dec/webpi_dec.h +++ b/thirdparty/libwebp/src/dec/webpi_dec.h @@ -130,4 +130,4 @@ int WebPAvoidSlowMemory(const WebPDecBuffer* const output, } // extern "C" #endif -#endif /* WEBP_DEC_WEBPI_DEC_H_ */ +#endif // WEBP_DEC_WEBPI_DEC_H_ diff --git a/thirdparty/libwebp/src/demux/demux.c b/thirdparty/libwebp/src/demux/demux.c index 684215e3de..a69c65b7cf 100644 --- a/thirdparty/libwebp/src/demux/demux.c +++ b/thirdparty/libwebp/src/demux/demux.c @@ -25,7 +25,7 @@ #define DMUX_MAJ_VERSION 1 #define DMUX_MIN_VERSION 0 -#define DMUX_REV_VERSION 0 +#define DMUX_REV_VERSION 1 typedef struct { size_t start_; // start location of the data diff --git a/thirdparty/libwebp/src/dsp/dsp.h b/thirdparty/libwebp/src/dsp/dsp.h index 4ab77a5130..fafc2d05d3 100644 --- a/thirdparty/libwebp/src/dsp/dsp.h +++ b/thirdparty/libwebp/src/dsp/dsp.h @@ -76,10 +76,6 @@ extern "C" { #define WEBP_USE_SSE41 #endif -#if defined(__AVX2__) || defined(WEBP_HAVE_AVX2) -#define WEBP_USE_AVX2 -#endif - // The intrinsics currently cause compiler errors with arm-nacl-gcc and the // inline assembly would need to be modified for use with Native Client. #if (defined(__ARM_NEON__) || \ @@ -679,4 +675,4 @@ void VP8FiltersInit(void); } // extern "C" #endif -#endif /* WEBP_DSP_DSP_H_ */ +#endif // WEBP_DSP_DSP_H_ diff --git a/thirdparty/libwebp/src/dsp/enc.c b/thirdparty/libwebp/src/dsp/enc.c index fa23b40a30..2fddbc4c52 100644 --- a/thirdparty/libwebp/src/dsp/enc.c +++ b/thirdparty/libwebp/src/dsp/enc.c @@ -734,7 +734,6 @@ VP8BlockCopy VP8Copy16x8; extern void VP8EncDspInitSSE2(void); extern void VP8EncDspInitSSE41(void); -extern void VP8EncDspInitAVX2(void); extern void VP8EncDspInitNEON(void); extern void VP8EncDspInitMIPS32(void); extern void VP8EncDspInitMIPSdspR2(void); @@ -784,11 +783,6 @@ WEBP_DSP_INIT_FUNC(VP8EncDspInit) { #endif } #endif -#if defined(WEBP_USE_AVX2) - if (VP8GetCPUInfo(kAVX2)) { - VP8EncDspInitAVX2(); - } -#endif #if defined(WEBP_USE_MIPS32) if (VP8GetCPUInfo(kMIPS32)) { VP8EncDspInitMIPS32(); diff --git a/thirdparty/libwebp/src/dsp/enc_avx2.c b/thirdparty/libwebp/src/dsp/enc_avx2.c deleted file mode 100644 index 8bc5798fee..0000000000 --- a/thirdparty/libwebp/src/dsp/enc_avx2.c +++ /dev/null @@ -1,21 +0,0 @@ -// Copyright 2014 Google Inc. All Rights Reserved. -// -// Use of this source code is governed by a BSD-style license -// that can be found in the COPYING file in the root of the source -// tree. An additional intellectual property rights grant can be found -// in the file PATENTS. All contributing project authors may -// be found in the AUTHORS file in the root of the source tree. -// ----------------------------------------------------------------------------- -// -// AVX2 version of speed-critical encoding functions. - -#include "src/dsp/dsp.h" - -#if defined(WEBP_USE_AVX2) - -#endif // WEBP_USE_AVX2 - -//------------------------------------------------------------------------------ -// Entry point - -WEBP_DSP_INIT_STUB(VP8EncDspInitAVX2) diff --git a/thirdparty/libwebp/src/dsp/lossless.c b/thirdparty/libwebp/src/dsp/lossless.c index f9b3c182d3..d21aa6a0a0 100644 --- a/thirdparty/libwebp/src/dsp/lossless.c +++ b/thirdparty/libwebp/src/dsp/lossless.c @@ -23,8 +23,6 @@ #include "src/dsp/lossless.h" #include "src/dsp/lossless_common.h" -#define MAX_DIFF_COST (1e30f) - //------------------------------------------------------------------------------ // Image transforms. diff --git a/thirdparty/libwebp/src/dsp/lossless.h b/thirdparty/libwebp/src/dsp/lossless.h index b2bbdfc93c..f709cc86b2 100644 --- a/thirdparty/libwebp/src/dsp/lossless.h +++ b/thirdparty/libwebp/src/dsp/lossless.h @@ -163,7 +163,7 @@ extern VP8LCostCombinedFunc VP8LExtraCostCombined; extern VP8LCombinedShannonEntropyFunc VP8LCombinedShannonEntropy; typedef struct { // small struct to hold counters - int counts[2]; // index: 0=zero steak, 1=non-zero streak + int counts[2]; // index: 0=zero streak, 1=non-zero streak int streaks[2][2]; // [zero/non-zero][streak<3 / streak>=3] } VP8LStreaks; @@ -194,10 +194,14 @@ extern VP8LGetEntropyUnrefinedFunc VP8LGetEntropyUnrefined; void VP8LBitsEntropyUnrefined(const uint32_t* const array, int n, VP8LBitEntropy* const entropy); -typedef void (*VP8LHistogramAddFunc)(const VP8LHistogram* const a, - const VP8LHistogram* const b, - VP8LHistogram* const out); -extern VP8LHistogramAddFunc VP8LHistogramAdd; +typedef void (*VP8LAddVectorFunc)(const uint32_t* a, const uint32_t* b, + uint32_t* out, int size); +extern VP8LAddVectorFunc VP8LAddVector; +typedef void (*VP8LAddVectorEqFunc)(const uint32_t* a, uint32_t* out, int size); +extern VP8LAddVectorEqFunc VP8LAddVectorEq; +void VP8LHistogramAdd(const VP8LHistogram* const a, + const VP8LHistogram* const b, + VP8LHistogram* const out); // ----------------------------------------------------------------------------- // PrefixEncode() diff --git a/thirdparty/libwebp/src/dsp/lossless_enc.c b/thirdparty/libwebp/src/dsp/lossless_enc.c index d608326fef..1408fbf580 100644 --- a/thirdparty/libwebp/src/dsp/lossless_enc.c +++ b/thirdparty/libwebp/src/dsp/lossless_enc.c @@ -632,38 +632,67 @@ static double ExtraCostCombined_C(const uint32_t* X, const uint32_t* Y, //------------------------------------------------------------------------------ -static void HistogramAdd_C(const VP8LHistogram* const a, - const VP8LHistogram* const b, - VP8LHistogram* const out) { +static void AddVector_C(const uint32_t* a, const uint32_t* b, uint32_t* out, + int size) { + int i; + for (i = 0; i < size; ++i) out[i] = a[i] + b[i]; +} + +static void AddVectorEq_C(const uint32_t* a, uint32_t* out, int size) { + int i; + for (i = 0; i < size; ++i) out[i] += a[i]; +} + +#define ADD(X, ARG, LEN) do { \ + if (a->is_used_[X]) { \ + if (b->is_used_[X]) { \ + VP8LAddVector(a->ARG, b->ARG, out->ARG, (LEN)); \ + } else { \ + memcpy(&out->ARG[0], &a->ARG[0], (LEN) * sizeof(out->ARG[0])); \ + } \ + } else if (b->is_used_[X]) { \ + memcpy(&out->ARG[0], &b->ARG[0], (LEN) * sizeof(out->ARG[0])); \ + } else { \ + memset(&out->ARG[0], 0, (LEN) * sizeof(out->ARG[0])); \ + } \ +} while (0) + +#define ADD_EQ(X, ARG, LEN) do { \ + if (a->is_used_[X]) { \ + if (out->is_used_[X]) { \ + VP8LAddVectorEq(a->ARG, out->ARG, (LEN)); \ + } else { \ + memcpy(&out->ARG[0], &a->ARG[0], (LEN) * sizeof(out->ARG[0])); \ + } \ + } \ +} while (0) + +void VP8LHistogramAdd(const VP8LHistogram* const a, + const VP8LHistogram* const b, VP8LHistogram* const out) { int i; const int literal_size = VP8LHistogramNumCodes(a->palette_code_bits_); assert(a->palette_code_bits_ == b->palette_code_bits_); + if (b != out) { - for (i = 0; i < literal_size; ++i) { - out->literal_[i] = a->literal_[i] + b->literal_[i]; - } - for (i = 0; i < NUM_DISTANCE_CODES; ++i) { - out->distance_[i] = a->distance_[i] + b->distance_[i]; - } - for (i = 0; i < NUM_LITERAL_CODES; ++i) { - out->red_[i] = a->red_[i] + b->red_[i]; - out->blue_[i] = a->blue_[i] + b->blue_[i]; - out->alpha_[i] = a->alpha_[i] + b->alpha_[i]; + ADD(0, literal_, literal_size); + ADD(1, red_, NUM_LITERAL_CODES); + ADD(2, blue_, NUM_LITERAL_CODES); + ADD(3, alpha_, NUM_LITERAL_CODES); + ADD(4, distance_, NUM_DISTANCE_CODES); + for (i = 0; i < 5; ++i) { + out->is_used_[i] = (a->is_used_[i] | b->is_used_[i]); } } else { - for (i = 0; i < literal_size; ++i) { - out->literal_[i] += a->literal_[i]; - } - for (i = 0; i < NUM_DISTANCE_CODES; ++i) { - out->distance_[i] += a->distance_[i]; - } - for (i = 0; i < NUM_LITERAL_CODES; ++i) { - out->red_[i] += a->red_[i]; - out->blue_[i] += a->blue_[i]; - out->alpha_[i] += a->alpha_[i]; - } + ADD_EQ(0, literal_, literal_size); + ADD_EQ(1, red_, NUM_LITERAL_CODES); + ADD_EQ(2, blue_, NUM_LITERAL_CODES); + ADD_EQ(3, alpha_, NUM_LITERAL_CODES); + ADD_EQ(4, distance_, NUM_DISTANCE_CODES); + for (i = 0; i < 5; ++i) out->is_used_[i] |= a->is_used_[i]; } } +#undef ADD +#undef ADD_EQ //------------------------------------------------------------------------------ // Image transforms. @@ -848,7 +877,8 @@ VP8LCombinedShannonEntropyFunc VP8LCombinedShannonEntropy; VP8LGetEntropyUnrefinedFunc VP8LGetEntropyUnrefined; VP8LGetCombinedEntropyUnrefinedFunc VP8LGetCombinedEntropyUnrefined; -VP8LHistogramAddFunc VP8LHistogramAdd; +VP8LAddVectorFunc VP8LAddVector; +VP8LAddVectorEqFunc VP8LAddVectorEq; VP8LVectorMismatchFunc VP8LVectorMismatch; VP8LBundleColorMapFunc VP8LBundleColorMap; @@ -885,7 +915,8 @@ WEBP_DSP_INIT_FUNC(VP8LEncDspInit) { VP8LGetEntropyUnrefined = GetEntropyUnrefined_C; VP8LGetCombinedEntropyUnrefined = GetCombinedEntropyUnrefined_C; - VP8LHistogramAdd = HistogramAdd_C; + VP8LAddVector = AddVector_C; + VP8LAddVectorEq = AddVectorEq_C; VP8LVectorMismatch = VectorMismatch_C; VP8LBundleColorMap = VP8LBundleColorMap_C; @@ -971,7 +1002,8 @@ WEBP_DSP_INIT_FUNC(VP8LEncDspInit) { assert(VP8LCombinedShannonEntropy != NULL); assert(VP8LGetEntropyUnrefined != NULL); assert(VP8LGetCombinedEntropyUnrefined != NULL); - assert(VP8LHistogramAdd != NULL); + assert(VP8LAddVector != NULL); + assert(VP8LAddVectorEq != NULL); assert(VP8LVectorMismatch != NULL); assert(VP8LBundleColorMap != NULL); assert(VP8LPredictorsSub[0] != NULL); diff --git a/thirdparty/libwebp/src/dsp/lossless_enc_mips32.c b/thirdparty/libwebp/src/dsp/lossless_enc_mips32.c index e7b58f4e8c..0412a093cf 100644 --- a/thirdparty/libwebp/src/dsp/lossless_enc_mips32.c +++ b/thirdparty/libwebp/src/dsp/lossless_enc_mips32.c @@ -344,65 +344,29 @@ static void GetCombinedEntropyUnrefined_MIPS32(const uint32_t X[], ASM_END_COMMON_0 \ ASM_END_COMMON_1 -#define ADD_VECTOR(A, B, OUT, SIZE, EXTRA_SIZE) do { \ - const uint32_t* pa = (const uint32_t*)(A); \ - const uint32_t* pb = (const uint32_t*)(B); \ - uint32_t* pout = (uint32_t*)(OUT); \ - const uint32_t* const LoopEnd = pa + (SIZE); \ - assert((SIZE) % 4 == 0); \ - ASM_START \ - ADD_TO_OUT(0, 4, 8, 12, 1, pa, pb, pout) \ - ASM_END_0 \ - if ((EXTRA_SIZE) > 0) { \ - const int last = (EXTRA_SIZE); \ - int i; \ - for (i = 0; i < last; ++i) pout[i] = pa[i] + pb[i]; \ - } \ -} while (0) - -#define ADD_VECTOR_EQ(A, OUT, SIZE, EXTRA_SIZE) do { \ - const uint32_t* pa = (const uint32_t*)(A); \ - uint32_t* pout = (uint32_t*)(OUT); \ - const uint32_t* const LoopEnd = pa + (SIZE); \ - assert((SIZE) % 4 == 0); \ - ASM_START \ - ADD_TO_OUT(0, 4, 8, 12, 0, pa, pout, pout) \ - ASM_END_1 \ - if ((EXTRA_SIZE) > 0) { \ - const int last = (EXTRA_SIZE); \ - int i; \ - for (i = 0; i < last; ++i) pout[i] += pa[i]; \ - } \ -} while (0) - -static void HistogramAdd_MIPS32(const VP8LHistogram* const a, - const VP8LHistogram* const b, - VP8LHistogram* const out) { +static void AddVector_MIPS32(const uint32_t* pa, const uint32_t* pb, + uint32_t* pout, int size) { uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; - const int extra_cache_size = VP8LHistogramNumCodes(a->palette_code_bits_) - - (NUM_LITERAL_CODES + NUM_LENGTH_CODES); - assert(a->palette_code_bits_ == b->palette_code_bits_); - - if (b != out) { - ADD_VECTOR(a->literal_, b->literal_, out->literal_, - NUM_LITERAL_CODES + NUM_LENGTH_CODES, extra_cache_size); - ADD_VECTOR(a->distance_, b->distance_, out->distance_, - NUM_DISTANCE_CODES, 0); - ADD_VECTOR(a->red_, b->red_, out->red_, NUM_LITERAL_CODES, 0); - ADD_VECTOR(a->blue_, b->blue_, out->blue_, NUM_LITERAL_CODES, 0); - ADD_VECTOR(a->alpha_, b->alpha_, out->alpha_, NUM_LITERAL_CODES, 0); - } else { - ADD_VECTOR_EQ(a->literal_, out->literal_, - NUM_LITERAL_CODES + NUM_LENGTH_CODES, extra_cache_size); - ADD_VECTOR_EQ(a->distance_, out->distance_, NUM_DISTANCE_CODES, 0); - ADD_VECTOR_EQ(a->red_, out->red_, NUM_LITERAL_CODES, 0); - ADD_VECTOR_EQ(a->blue_, out->blue_, NUM_LITERAL_CODES, 0); - ADD_VECTOR_EQ(a->alpha_, out->alpha_, NUM_LITERAL_CODES, 0); - } + const uint32_t end = ((size) / 4) * 4; + const uint32_t* const LoopEnd = pa + end; + int i; + ASM_START + ADD_TO_OUT(0, 4, 8, 12, 1, pa, pb, pout) + ASM_END_0 + for (i = end; i < size; ++i) pout[i] = pa[i] + pb[i]; +} + +static void AddVectorEq_MIPS32(const uint32_t* pa, uint32_t* pout, int size) { + uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; + const uint32_t end = ((size) / 4) * 4; + const uint32_t* const LoopEnd = pa + end; + int i; + ASM_START + ADD_TO_OUT(0, 4, 8, 12, 0, pa, pout, pout) + ASM_END_1 + for (i = end; i < size; ++i) pout[i] += pa[i]; } -#undef ADD_VECTOR_EQ -#undef ADD_VECTOR #undef ASM_END_1 #undef ASM_END_0 #undef ASM_END_COMMON_1 @@ -422,7 +386,8 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitMIPS32(void) { VP8LExtraCostCombined = ExtraCostCombined_MIPS32; VP8LGetEntropyUnrefined = GetEntropyUnrefined_MIPS32; VP8LGetCombinedEntropyUnrefined = GetCombinedEntropyUnrefined_MIPS32; - VP8LHistogramAdd = HistogramAdd_MIPS32; + VP8LAddVector = AddVector_MIPS32; + VP8LAddVectorEq = AddVectorEq_MIPS32; } #else // !WEBP_USE_MIPS32 diff --git a/thirdparty/libwebp/src/dsp/lossless_enc_sse2.c b/thirdparty/libwebp/src/dsp/lossless_enc_sse2.c index f84a9909e1..36478c4912 100644 --- a/thirdparty/libwebp/src/dsp/lossless_enc_sse2.c +++ b/thirdparty/libwebp/src/dsp/lossless_enc_sse2.c @@ -170,12 +170,13 @@ static void CollectColorRedTransforms_SSE2(const uint32_t* argb, int stride, //------------------------------------------------------------------------------ +// Note we are adding uint32_t's as *signed* int32's (using _mm_add_epi32). But +// that's ok since the histogram values are less than 1<<28 (max picture size). #define LINE_SIZE 16 // 8 or 16 static void AddVector_SSE2(const uint32_t* a, const uint32_t* b, uint32_t* out, int size) { int i; - assert(size % LINE_SIZE == 0); - for (i = 0; i < size; i += LINE_SIZE) { + for (i = 0; i + LINE_SIZE <= size; i += LINE_SIZE) { const __m128i a0 = _mm_loadu_si128((const __m128i*)&a[i + 0]); const __m128i a1 = _mm_loadu_si128((const __m128i*)&a[i + 4]); #if (LINE_SIZE == 16) @@ -195,12 +196,14 @@ static void AddVector_SSE2(const uint32_t* a, const uint32_t* b, uint32_t* out, _mm_storeu_si128((__m128i*)&out[i + 12], _mm_add_epi32(a3, b3)); #endif } + for (; i < size; ++i) { + out[i] = a[i] + b[i]; + } } static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) { int i; - assert(size % LINE_SIZE == 0); - for (i = 0; i < size; i += LINE_SIZE) { + for (i = 0; i + LINE_SIZE <= size; i += LINE_SIZE) { const __m128i a0 = _mm_loadu_si128((const __m128i*)&a[i + 0]); const __m128i a1 = _mm_loadu_si128((const __m128i*)&a[i + 4]); #if (LINE_SIZE == 16) @@ -220,35 +223,11 @@ static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) { _mm_storeu_si128((__m128i*)&out[i + 12], _mm_add_epi32(a3, b3)); #endif } -} -#undef LINE_SIZE - -// Note we are adding uint32_t's as *signed* int32's (using _mm_add_epi32). But -// that's ok since the histogram values are less than 1<<28 (max picture size). -static void HistogramAdd_SSE2(const VP8LHistogram* const a, - const VP8LHistogram* const b, - VP8LHistogram* const out) { - int i; - const int literal_size = VP8LHistogramNumCodes(a->palette_code_bits_); - assert(a->palette_code_bits_ == b->palette_code_bits_); - if (b != out) { - AddVector_SSE2(a->literal_, b->literal_, out->literal_, NUM_LITERAL_CODES); - AddVector_SSE2(a->red_, b->red_, out->red_, NUM_LITERAL_CODES); - AddVector_SSE2(a->blue_, b->blue_, out->blue_, NUM_LITERAL_CODES); - AddVector_SSE2(a->alpha_, b->alpha_, out->alpha_, NUM_LITERAL_CODES); - } else { - AddVectorEq_SSE2(a->literal_, out->literal_, NUM_LITERAL_CODES); - AddVectorEq_SSE2(a->red_, out->red_, NUM_LITERAL_CODES); - AddVectorEq_SSE2(a->blue_, out->blue_, NUM_LITERAL_CODES); - AddVectorEq_SSE2(a->alpha_, out->alpha_, NUM_LITERAL_CODES); - } - for (i = NUM_LITERAL_CODES; i < literal_size; ++i) { - out->literal_[i] = a->literal_[i] + b->literal_[i]; - } - for (i = 0; i < NUM_DISTANCE_CODES; ++i) { - out->distance_[i] = a->distance_[i] + b->distance_[i]; + for (; i < size; ++i) { + out[i] += a[i]; } } +#undef LINE_SIZE //------------------------------------------------------------------------------ // Entropy @@ -675,7 +654,8 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitSSE2(void) { VP8LTransformColor = TransformColor_SSE2; VP8LCollectColorBlueTransforms = CollectColorBlueTransforms_SSE2; VP8LCollectColorRedTransforms = CollectColorRedTransforms_SSE2; - VP8LHistogramAdd = HistogramAdd_SSE2; + VP8LAddVector = AddVector_SSE2; + VP8LAddVectorEq = AddVectorEq_SSE2; VP8LCombinedShannonEntropy = CombinedShannonEntropy_SSE2; VP8LVectorMismatch = VectorMismatch_SSE2; VP8LBundleColorMap = BundleColorMap_SSE2; diff --git a/thirdparty/libwebp/src/dsp/msa_macro.h b/thirdparty/libwebp/src/dsp/msa_macro.h index dfacda6ccd..de026a1d9e 100644 --- a/thirdparty/libwebp/src/dsp/msa_macro.h +++ b/thirdparty/libwebp/src/dsp/msa_macro.h @@ -1389,4 +1389,4 @@ static WEBP_INLINE uint32_t func_hadd_uh_u32(v8u16 in) { } while (0) #define AVER_UB2_UB(...) AVER_UB2(v16u8, __VA_ARGS__) -#endif /* WEBP_DSP_MSA_MACRO_H_ */ +#endif // WEBP_DSP_MSA_MACRO_H_ diff --git a/thirdparty/libwebp/src/dsp/rescaler.c b/thirdparty/libwebp/src/dsp/rescaler.c index f307d35056..753f84fcf4 100644 --- a/thirdparty/libwebp/src/dsp/rescaler.c +++ b/thirdparty/libwebp/src/dsp/rescaler.c @@ -21,6 +21,7 @@ #define ROUNDER (WEBP_RESCALER_ONE >> 1) #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX) +#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX) //------------------------------------------------------------------------------ // Row import @@ -138,7 +139,7 @@ void WebPRescalerExportRowShrink_C(WebPRescaler* const wrk) { if (yscale) { for (x_out = 0; x_out < x_out_max; ++x_out) { const uint32_t frac = (uint32_t)MULT_FIX(frow[x_out], yscale); - const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale); + const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale); assert(v >= 0 && v <= 255); dst[x_out] = v; irow[x_out] = frac; // new fractional start @@ -153,6 +154,7 @@ void WebPRescalerExportRowShrink_C(WebPRescaler* const wrk) { } } +#undef MULT_FIX_FLOOR #undef MULT_FIX #undef ROUNDER diff --git a/thirdparty/libwebp/src/dsp/rescaler_mips32.c b/thirdparty/libwebp/src/dsp/rescaler_mips32.c index 542f7e5970..61f63c616c 100644 --- a/thirdparty/libwebp/src/dsp/rescaler_mips32.c +++ b/thirdparty/libwebp/src/dsp/rescaler_mips32.c @@ -209,6 +209,7 @@ static void ExportRowExpand_MIPS32(WebPRescaler* const wrk) { } } +#if 0 // disabled for now. TODO(skal): make match the C-code static void ExportRowShrink_MIPS32(WebPRescaler* const wrk) { const int x_out_max = wrk->dst_width * wrk->num_channels; uint8_t* dst = wrk->dst; @@ -273,6 +274,7 @@ static void ExportRowShrink_MIPS32(WebPRescaler* const wrk) { ); } } +#endif // 0 //------------------------------------------------------------------------------ // Entry point @@ -283,7 +285,7 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInitMIPS32(void) { WebPRescalerImportRowExpand = ImportRowExpand_MIPS32; WebPRescalerImportRowShrink = ImportRowShrink_MIPS32; WebPRescalerExportRowExpand = ExportRowExpand_MIPS32; - WebPRescalerExportRowShrink = ExportRowShrink_MIPS32; +// WebPRescalerExportRowShrink = ExportRowShrink_MIPS32; } #else // !WEBP_USE_MIPS32 diff --git a/thirdparty/libwebp/src/dsp/rescaler_mips_dsp_r2.c b/thirdparty/libwebp/src/dsp/rescaler_mips_dsp_r2.c index b78aac15e6..ce9e64862e 100644 --- a/thirdparty/libwebp/src/dsp/rescaler_mips_dsp_r2.c +++ b/thirdparty/libwebp/src/dsp/rescaler_mips_dsp_r2.c @@ -20,10 +20,12 @@ #define ROUNDER (WEBP_RESCALER_ONE >> 1) #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX) +#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX) //------------------------------------------------------------------------------ // Row export +#if 0 // disabled for now. TODO(skal): make match the C-code static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) { int i; const int x_out_max = wrk->dst_width * wrk->num_channels; @@ -106,7 +108,7 @@ static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) { } for (i = 0; i < (x_out_max & 0x3); ++i) { const uint32_t frac = (uint32_t)MULT_FIX(*frow++, yscale); - const int v = (int)MULT_FIX(*irow - frac, wrk->fxy_scale); + const int v = (int)MULT_FIX_FLOOR(*irow - frac, wrk->fxy_scale); assert(v >= 0 && v <= 255); *dst++ = v; *irow++ = frac; // new fractional start @@ -154,13 +156,14 @@ static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) { ); } for (i = 0; i < (x_out_max & 0x3); ++i) { - const int v = (int)MULT_FIX(*irow, wrk->fxy_scale); + const int v = (int)MULT_FIX_FLOOR(*irow, wrk->fxy_scale); assert(v >= 0 && v <= 255); *dst++ = v; *irow++ = 0; } } } +#endif // 0 static void ExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) { int i; @@ -294,6 +297,7 @@ static void ExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) { } } +#undef MULT_FIX_FLOOR #undef MULT_FIX #undef ROUNDER @@ -304,7 +308,7 @@ extern void WebPRescalerDspInitMIPSdspR2(void); WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInitMIPSdspR2(void) { WebPRescalerExportRowExpand = ExportRowExpand_MIPSdspR2; - WebPRescalerExportRowShrink = ExportRowShrink_MIPSdspR2; +// WebPRescalerExportRowShrink = ExportRowShrink_MIPSdspR2; } #else // !WEBP_USE_MIPS_DSP_R2 diff --git a/thirdparty/libwebp/src/dsp/rescaler_msa.c b/thirdparty/libwebp/src/dsp/rescaler_msa.c index f3bc99f1cd..c559254836 100644 --- a/thirdparty/libwebp/src/dsp/rescaler_msa.c +++ b/thirdparty/libwebp/src/dsp/rescaler_msa.c @@ -22,6 +22,7 @@ #define ROUNDER (WEBP_RESCALER_ONE >> 1) #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX) +#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX) #define CALC_MULT_FIX_16(in0, in1, in2, in3, scale, shift, dst) do { \ v4u32 tmp0, tmp1, tmp2, tmp3; \ @@ -262,6 +263,7 @@ static void RescalerExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) { } } +#if 0 // disabled for now. TODO(skal): make match the C-code static WEBP_INLINE void ExportRowShrink_0(const uint32_t* frow, uint32_t* irow, uint8_t* dst, int length, const uint32_t yscale, @@ -341,7 +343,7 @@ static WEBP_INLINE void ExportRowShrink_0(const uint32_t* frow, uint32_t* irow, } for (x_out = 0; x_out < length; ++x_out) { const uint32_t frac = (uint32_t)MULT_FIX(frow[x_out], yscale); - const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale); + const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale); assert(v >= 0 && v <= 255); dst[x_out] = v; irow[x_out] = frac; @@ -426,6 +428,7 @@ static void RescalerExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) { ExportRowShrink_1(irow, dst, x_out_max, wrk); } } +#endif // 0 //------------------------------------------------------------------------------ // Entry point @@ -434,7 +437,7 @@ extern void WebPRescalerDspInitMSA(void); WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInitMSA(void) { WebPRescalerExportRowExpand = RescalerExportRowExpand_MIPSdspR2; - WebPRescalerExportRowShrink = RescalerExportRowShrink_MIPSdspR2; +// WebPRescalerExportRowShrink = RescalerExportRowShrink_MIPSdspR2; } #else // !WEBP_USE_MSA diff --git a/thirdparty/libwebp/src/dsp/rescaler_neon.c b/thirdparty/libwebp/src/dsp/rescaler_neon.c index 3eff9fbaf4..a553f06f79 100644 --- a/thirdparty/libwebp/src/dsp/rescaler_neon.c +++ b/thirdparty/libwebp/src/dsp/rescaler_neon.c @@ -22,6 +22,7 @@ #define ROUNDER (WEBP_RESCALER_ONE >> 1) #define MULT_FIX_C(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX) +#define MULT_FIX_FLOOR_C(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX) #define LOAD_32x4(SRC, DST) const uint32x4_t DST = vld1q_u32((SRC)) #define LOAD_32x8(SRC, DST0, DST1) \ @@ -35,8 +36,11 @@ #if (WEBP_RESCALER_RFIX == 32) #define MAKE_HALF_CST(C) vdupq_n_s32((int32_t)((C) >> 1)) -#define MULT_FIX(A, B) /* note: B is actualy scale>>1. See MAKE_HALF_CST */ \ +// note: B is actualy scale>>1. See MAKE_HALF_CST +#define MULT_FIX(A, B) \ vreinterpretq_u32_s32(vqrdmulhq_s32(vreinterpretq_s32_u32((A)), (B))) +#define MULT_FIX_FLOOR(A, B) \ + vreinterpretq_u32_s32(vqdmulhq_s32(vreinterpretq_s32_u32((A)), (B))) #else #error "MULT_FIX/WEBP_RESCALER_RFIX need some more work" #endif @@ -135,8 +139,8 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) { const uint32x4_t A1 = MULT_FIX(in1, yscale_half); const uint32x4_t B0 = vqsubq_u32(in2, A0); const uint32x4_t B1 = vqsubq_u32(in3, A1); - const uint32x4_t C0 = MULT_FIX(B0, fxy_scale_half); - const uint32x4_t C1 = MULT_FIX(B1, fxy_scale_half); + const uint32x4_t C0 = MULT_FIX_FLOOR(B0, fxy_scale_half); + const uint32x4_t C1 = MULT_FIX_FLOOR(B1, fxy_scale_half); const uint16x4_t D0 = vmovn_u32(C0); const uint16x4_t D1 = vmovn_u32(C1); const uint8x8_t E = vmovn_u16(vcombine_u16(D0, D1)); @@ -145,7 +149,7 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) { } for (; x_out < x_out_max; ++x_out) { const uint32_t frac = (uint32_t)MULT_FIX_C(frow[x_out], yscale); - const int v = (int)MULT_FIX_C(irow[x_out] - frac, wrk->fxy_scale); + const int v = (int)MULT_FIX_FLOOR_C(irow[x_out] - frac, fxy_scale); assert(v >= 0 && v <= 255); dst[x_out] = v; irow[x_out] = frac; // new fractional start @@ -170,6 +174,12 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) { } } +#undef MULT_FIX_FLOOR_C +#undef MULT_FIX_C +#undef MULT_FIX_FLOOR +#undef MULT_FIX +#undef ROUNDER + //------------------------------------------------------------------------------ extern void WebPRescalerDspInitNEON(void); diff --git a/thirdparty/libwebp/src/dsp/rescaler_sse2.c b/thirdparty/libwebp/src/dsp/rescaler_sse2.c index 64c50deab5..f7461a452c 100644 --- a/thirdparty/libwebp/src/dsp/rescaler_sse2.c +++ b/thirdparty/libwebp/src/dsp/rescaler_sse2.c @@ -25,6 +25,7 @@ #define ROUNDER (WEBP_RESCALER_ONE >> 1) #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX) +#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX) // input: 8 bytes ABCDEFGH -> output: A0E0B0F0C0G0D0H0 static void LoadTwoPixels_SSE2(const uint8_t* const src, __m128i* out) { @@ -224,6 +225,35 @@ static WEBP_INLINE void ProcessRow_SSE2(const __m128i* const A0, _mm_storel_epi64((__m128i*)dst, G); } +static WEBP_INLINE void ProcessRow_Floor_SSE2(const __m128i* const A0, + const __m128i* const A1, + const __m128i* const A2, + const __m128i* const A3, + const __m128i* const mult, + uint8_t* const dst) { + const __m128i mask = _mm_set_epi32(0xffffffffu, 0, 0xffffffffu, 0); + const __m128i B0 = _mm_mul_epu32(*A0, *mult); + const __m128i B1 = _mm_mul_epu32(*A1, *mult); + const __m128i B2 = _mm_mul_epu32(*A2, *mult); + const __m128i B3 = _mm_mul_epu32(*A3, *mult); + const __m128i D0 = _mm_srli_epi64(B0, WEBP_RESCALER_RFIX); + const __m128i D1 = _mm_srli_epi64(B1, WEBP_RESCALER_RFIX); +#if (WEBP_RESCALER_RFIX < 32) + const __m128i D2 = + _mm_and_si128(_mm_slli_epi64(B2, 32 - WEBP_RESCALER_RFIX), mask); + const __m128i D3 = + _mm_and_si128(_mm_slli_epi64(B3, 32 - WEBP_RESCALER_RFIX), mask); +#else + const __m128i D2 = _mm_and_si128(B2, mask); + const __m128i D3 = _mm_and_si128(B3, mask); +#endif + const __m128i E0 = _mm_or_si128(D0, D2); + const __m128i E1 = _mm_or_si128(D1, D3); + const __m128i F = _mm_packs_epi32(E0, E1); + const __m128i G = _mm_packus_epi16(F, F); + _mm_storel_epi64((__m128i*)dst, G); +} + static void RescalerExportRowExpand_SSE2(WebPRescaler* const wrk) { int x_out; uint8_t* const dst = wrk->dst; @@ -322,12 +352,12 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) { const __m128i G1 = _mm_or_si128(D1, F3); _mm_storeu_si128((__m128i*)(irow + x_out + 0), G0); _mm_storeu_si128((__m128i*)(irow + x_out + 4), G1); - ProcessRow_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out); + ProcessRow_Floor_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out); } } for (; x_out < x_out_max; ++x_out) { const uint32_t frac = (int)MULT_FIX(frow[x_out], yscale); - const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale); + const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale); assert(v >= 0 && v <= 255); dst[x_out] = v; irow[x_out] = frac; // new fractional start @@ -352,6 +382,7 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) { } } +#undef MULT_FIX_FLOOR #undef MULT_FIX #undef ROUNDER diff --git a/thirdparty/libwebp/src/dsp/yuv.h b/thirdparty/libwebp/src/dsp/yuv.h index eb787270d2..c12be1d094 100644 --- a/thirdparty/libwebp/src/dsp/yuv.h +++ b/thirdparty/libwebp/src/dsp/yuv.h @@ -207,4 +207,4 @@ static WEBP_INLINE int VP8RGBToV(int r, int g, int b, int rounding) { } // extern "C" #endif -#endif /* WEBP_DSP_YUV_H_ */ +#endif // WEBP_DSP_YUV_H_ diff --git a/thirdparty/libwebp/src/enc/analysis_enc.c b/thirdparty/libwebp/src/enc/analysis_enc.c index a47ff7d4e8..687757ae03 100644 --- a/thirdparty/libwebp/src/enc/analysis_enc.c +++ b/thirdparty/libwebp/src/enc/analysis_enc.c @@ -458,7 +458,7 @@ static void MergeJobs(const SegmentJob* const src, SegmentJob* const dst) { dst->uv_alpha += src->uv_alpha; } -// initialize the job struct with some TODOs +// initialize the job struct with some tasks to perform static void InitSegmentJob(VP8Encoder* const enc, SegmentJob* const job, int start_row, int end_row) { WebPGetWorkerInterface()->Init(&job->worker); diff --git a/thirdparty/libwebp/src/enc/backward_references_cost_enc.c b/thirdparty/libwebp/src/enc/backward_references_cost_enc.c index 7175496c7f..516abd73eb 100644 --- a/thirdparty/libwebp/src/enc/backward_references_cost_enc.c +++ b/thirdparty/libwebp/src/enc/backward_references_cost_enc.c @@ -67,7 +67,7 @@ static int CostModelBuild(CostModel* const m, int xsize, int cache_bits, // The following code is similar to VP8LHistogramCreate but converts the // distance to plane code. - VP8LHistogramInit(histo, cache_bits); + VP8LHistogramInit(histo, cache_bits, /*init_arrays=*/ 1); while (VP8LRefsCursorOk(&c)) { VP8LHistogramAddSinglePixOrCopy(histo, c.cur_pos, VP8LDistanceToPlaneCode, xsize); diff --git a/thirdparty/libwebp/src/enc/backward_references_enc.c b/thirdparty/libwebp/src/enc/backward_references_enc.c index 39230188b9..3ab7b0ac7d 100644 --- a/thirdparty/libwebp/src/enc/backward_references_enc.c +++ b/thirdparty/libwebp/src/enc/backward_references_enc.c @@ -715,6 +715,7 @@ static int CalculateBestCacheSize(const uint32_t* argb, int quality, for (i = 0; i <= cache_bits_max; ++i) { histos[i] = VP8LAllocateHistogram(i); if (histos[i] == NULL) goto Error; + VP8LHistogramInit(histos[i], i, /*init_arrays=*/ 1); if (i == 0) continue; cc_init[i] = VP8LColorCacheInit(&hashers[i], i); if (!cc_init[i]) goto Error; diff --git a/thirdparty/libwebp/src/enc/cost_enc.h b/thirdparty/libwebp/src/enc/cost_enc.h index bdce1e6a3b..a4b177b342 100644 --- a/thirdparty/libwebp/src/enc/cost_enc.h +++ b/thirdparty/libwebp/src/enc/cost_enc.h @@ -79,4 +79,4 @@ extern const uint16_t VP8FixedCostsI4[NUM_BMODES][NUM_BMODES][NUM_BMODES]; } // extern "C" #endif -#endif /* WEBP_ENC_COST_ENC_H_ */ +#endif // WEBP_ENC_COST_ENC_H_ diff --git a/thirdparty/libwebp/src/enc/delta_palettization_enc.c b/thirdparty/libwebp/src/enc/delta_palettization_enc.c deleted file mode 100644 index a61c8e6c93..0000000000 --- a/thirdparty/libwebp/src/enc/delta_palettization_enc.c +++ /dev/null @@ -1,455 +0,0 @@ -// Copyright 2015 Google Inc. All Rights Reserved. -// -// Use of this source code is governed by a BSD-style license -// that can be found in the COPYING file in the root of the source -// tree. An additional intellectual property rights grant can be found -// in the file PATENTS. All contributing project authors may -// be found in the AUTHORS file in the root of the source tree. -// ----------------------------------------------------------------------------- -// -// Author: Mislav Bradac (mislavm@google.com) -// - -#include "src/enc/delta_palettization_enc.h" - -#ifdef WEBP_EXPERIMENTAL_FEATURES -#include "src/webp/types.h" -#include "src/dsp/lossless.h" - -#define MK_COL(r, g, b) (((r) << 16) + ((g) << 8) + (b)) - -// Format allows palette up to 256 entries, but more palette entries produce -// bigger entropy. In the future it will probably be useful to add more entries -// that are far from the origin of the palette or choose remaining entries -// dynamically. -#define DELTA_PALETTE_SIZE 226 - -// Palette used for delta_palettization. Entries are roughly sorted by distance -// of their signed equivalents from the origin. -static const uint32_t kDeltaPalette[DELTA_PALETTE_SIZE] = { - MK_COL(0u, 0u, 0u), - MK_COL(255u, 255u, 255u), - MK_COL(1u, 1u, 1u), - MK_COL(254u, 254u, 254u), - MK_COL(2u, 2u, 2u), - MK_COL(4u, 4u, 4u), - MK_COL(252u, 252u, 252u), - MK_COL(250u, 0u, 0u), - MK_COL(0u, 250u, 0u), - MK_COL(0u, 0u, 250u), - MK_COL(6u, 0u, 0u), - MK_COL(0u, 6u, 0u), - MK_COL(0u, 0u, 6u), - MK_COL(0u, 0u, 248u), - MK_COL(0u, 0u, 8u), - MK_COL(0u, 248u, 0u), - MK_COL(0u, 248u, 248u), - MK_COL(0u, 248u, 8u), - MK_COL(0u, 8u, 0u), - MK_COL(0u, 8u, 248u), - MK_COL(0u, 8u, 8u), - MK_COL(8u, 8u, 8u), - MK_COL(248u, 0u, 0u), - MK_COL(248u, 0u, 248u), - MK_COL(248u, 0u, 8u), - MK_COL(248u, 248u, 0u), - MK_COL(248u, 8u, 0u), - MK_COL(8u, 0u, 0u), - MK_COL(8u, 0u, 248u), - MK_COL(8u, 0u, 8u), - MK_COL(8u, 248u, 0u), - MK_COL(8u, 8u, 0u), - MK_COL(23u, 23u, 23u), - MK_COL(13u, 13u, 13u), - MK_COL(232u, 232u, 232u), - MK_COL(244u, 244u, 244u), - MK_COL(245u, 245u, 250u), - MK_COL(50u, 50u, 50u), - MK_COL(204u, 204u, 204u), - MK_COL(236u, 236u, 236u), - MK_COL(16u, 16u, 16u), - MK_COL(240u, 16u, 16u), - MK_COL(16u, 240u, 16u), - MK_COL(240u, 240u, 16u), - MK_COL(16u, 16u, 240u), - MK_COL(240u, 16u, 240u), - MK_COL(16u, 240u, 240u), - MK_COL(240u, 240u, 240u), - MK_COL(0u, 0u, 232u), - MK_COL(0u, 232u, 0u), - MK_COL(232u, 0u, 0u), - MK_COL(0u, 0u, 24u), - MK_COL(0u, 24u, 0u), - MK_COL(24u, 0u, 0u), - MK_COL(32u, 32u, 32u), - MK_COL(224u, 32u, 32u), - MK_COL(32u, 224u, 32u), - MK_COL(224u, 224u, 32u), - MK_COL(32u, 32u, 224u), - MK_COL(224u, 32u, 224u), - MK_COL(32u, 224u, 224u), - MK_COL(224u, 224u, 224u), - MK_COL(0u, 0u, 176u), - MK_COL(0u, 0u, 80u), - MK_COL(0u, 176u, 0u), - MK_COL(0u, 176u, 176u), - MK_COL(0u, 176u, 80u), - MK_COL(0u, 80u, 0u), - MK_COL(0u, 80u, 176u), - MK_COL(0u, 80u, 80u), - MK_COL(176u, 0u, 0u), - MK_COL(176u, 0u, 176u), - MK_COL(176u, 0u, 80u), - MK_COL(176u, 176u, 0u), - MK_COL(176u, 80u, 0u), - MK_COL(80u, 0u, 0u), - MK_COL(80u, 0u, 176u), - MK_COL(80u, 0u, 80u), - MK_COL(80u, 176u, 0u), - MK_COL(80u, 80u, 0u), - MK_COL(0u, 0u, 152u), - MK_COL(0u, 0u, 104u), - MK_COL(0u, 152u, 0u), - MK_COL(0u, 152u, 152u), - MK_COL(0u, 152u, 104u), - MK_COL(0u, 104u, 0u), - MK_COL(0u, 104u, 152u), - MK_COL(0u, 104u, 104u), - MK_COL(152u, 0u, 0u), - MK_COL(152u, 0u, 152u), - MK_COL(152u, 0u, 104u), - MK_COL(152u, 152u, 0u), - MK_COL(152u, 104u, 0u), - MK_COL(104u, 0u, 0u), - MK_COL(104u, 0u, 152u), - MK_COL(104u, 0u, 104u), - MK_COL(104u, 152u, 0u), - MK_COL(104u, 104u, 0u), - MK_COL(216u, 216u, 216u), - MK_COL(216u, 216u, 40u), - MK_COL(216u, 216u, 176u), - MK_COL(216u, 216u, 80u), - MK_COL(216u, 40u, 216u), - MK_COL(216u, 40u, 40u), - MK_COL(216u, 40u, 176u), - MK_COL(216u, 40u, 80u), - MK_COL(216u, 176u, 216u), - MK_COL(216u, 176u, 40u), - MK_COL(216u, 176u, 176u), - MK_COL(216u, 176u, 80u), - MK_COL(216u, 80u, 216u), - MK_COL(216u, 80u, 40u), - MK_COL(216u, 80u, 176u), - MK_COL(216u, 80u, 80u), - MK_COL(40u, 216u, 216u), - MK_COL(40u, 216u, 40u), - MK_COL(40u, 216u, 176u), - MK_COL(40u, 216u, 80u), - MK_COL(40u, 40u, 216u), - MK_COL(40u, 40u, 40u), - MK_COL(40u, 40u, 176u), - MK_COL(40u, 40u, 80u), - MK_COL(40u, 176u, 216u), - MK_COL(40u, 176u, 40u), - MK_COL(40u, 176u, 176u), - MK_COL(40u, 176u, 80u), - MK_COL(40u, 80u, 216u), - MK_COL(40u, 80u, 40u), - MK_COL(40u, 80u, 176u), - MK_COL(40u, 80u, 80u), - MK_COL(80u, 216u, 216u), - MK_COL(80u, 216u, 40u), - MK_COL(80u, 216u, 176u), - MK_COL(80u, 216u, 80u), - MK_COL(80u, 40u, 216u), - MK_COL(80u, 40u, 40u), - MK_COL(80u, 40u, 176u), - MK_COL(80u, 40u, 80u), - MK_COL(80u, 176u, 216u), - MK_COL(80u, 176u, 40u), - MK_COL(80u, 176u, 176u), - MK_COL(80u, 176u, 80u), - MK_COL(80u, 80u, 216u), - MK_COL(80u, 80u, 40u), - MK_COL(80u, 80u, 176u), - MK_COL(80u, 80u, 80u), - MK_COL(0u, 0u, 192u), - MK_COL(0u, 0u, 64u), - MK_COL(0u, 0u, 128u), - MK_COL(0u, 192u, 0u), - MK_COL(0u, 192u, 192u), - MK_COL(0u, 192u, 64u), - MK_COL(0u, 192u, 128u), - MK_COL(0u, 64u, 0u), - MK_COL(0u, 64u, 192u), - MK_COL(0u, 64u, 64u), - MK_COL(0u, 64u, 128u), - MK_COL(0u, 128u, 0u), - MK_COL(0u, 128u, 192u), - MK_COL(0u, 128u, 64u), - MK_COL(0u, 128u, 128u), - MK_COL(176u, 216u, 216u), - MK_COL(176u, 216u, 40u), - MK_COL(176u, 216u, 176u), - MK_COL(176u, 216u, 80u), - MK_COL(176u, 40u, 216u), - MK_COL(176u, 40u, 40u), - MK_COL(176u, 40u, 176u), - MK_COL(176u, 40u, 80u), - MK_COL(176u, 176u, 216u), - MK_COL(176u, 176u, 40u), - MK_COL(176u, 176u, 176u), - MK_COL(176u, 176u, 80u), - MK_COL(176u, 80u, 216u), - MK_COL(176u, 80u, 40u), - MK_COL(176u, 80u, 176u), - MK_COL(176u, 80u, 80u), - MK_COL(192u, 0u, 0u), - MK_COL(192u, 0u, 192u), - MK_COL(192u, 0u, 64u), - MK_COL(192u, 0u, 128u), - MK_COL(192u, 192u, 0u), - MK_COL(192u, 192u, 192u), - MK_COL(192u, 192u, 64u), - MK_COL(192u, 192u, 128u), - MK_COL(192u, 64u, 0u), - MK_COL(192u, 64u, 192u), - MK_COL(192u, 64u, 64u), - MK_COL(192u, 64u, 128u), - MK_COL(192u, 128u, 0u), - MK_COL(192u, 128u, 192u), - MK_COL(192u, 128u, 64u), - MK_COL(192u, 128u, 128u), - MK_COL(64u, 0u, 0u), - MK_COL(64u, 0u, 192u), - MK_COL(64u, 0u, 64u), - MK_COL(64u, 0u, 128u), - MK_COL(64u, 192u, 0u), - MK_COL(64u, 192u, 192u), - MK_COL(64u, 192u, 64u), - MK_COL(64u, 192u, 128u), - MK_COL(64u, 64u, 0u), - MK_COL(64u, 64u, 192u), - MK_COL(64u, 64u, 64u), - MK_COL(64u, 64u, 128u), - MK_COL(64u, 128u, 0u), - MK_COL(64u, 128u, 192u), - MK_COL(64u, 128u, 64u), - MK_COL(64u, 128u, 128u), - MK_COL(128u, 0u, 0u), - MK_COL(128u, 0u, 192u), - MK_COL(128u, 0u, 64u), - MK_COL(128u, 0u, 128u), - MK_COL(128u, 192u, 0u), - MK_COL(128u, 192u, 192u), - MK_COL(128u, 192u, 64u), - MK_COL(128u, 192u, 128u), - MK_COL(128u, 64u, 0u), - MK_COL(128u, 64u, 192u), - MK_COL(128u, 64u, 64u), - MK_COL(128u, 64u, 128u), - MK_COL(128u, 128u, 0u), - MK_COL(128u, 128u, 192u), - MK_COL(128u, 128u, 64u), - MK_COL(128u, 128u, 128u), -}; - -#undef MK_COL - -//------------------------------------------------------------------------------ -// TODO(skal): move the functions to dsp/lossless.c when the correct -// granularity is found. For now, we'll just copy-paste some useful bits -// here instead. - -// In-place sum of each component with mod 256. -static WEBP_INLINE void AddPixelsEq(uint32_t* a, uint32_t b) { - const uint32_t alpha_and_green = (*a & 0xff00ff00u) + (b & 0xff00ff00u); - const uint32_t red_and_blue = (*a & 0x00ff00ffu) + (b & 0x00ff00ffu); - *a = (alpha_and_green & 0xff00ff00u) | (red_and_blue & 0x00ff00ffu); -} - -static WEBP_INLINE uint32_t Clip255(uint32_t a) { - if (a < 256) { - return a; - } - // return 0, when a is a negative integer. - // return 255, when a is positive. - return ~a >> 24; -} - -// Delta palettization functions. -static WEBP_INLINE int Square(int x) { - return x * x; -} - -static WEBP_INLINE uint32_t Intensity(uint32_t a) { - return - 30 * ((a >> 16) & 0xff) + - 59 * ((a >> 8) & 0xff) + - 11 * ((a >> 0) & 0xff); -} - -static uint32_t CalcDist(uint32_t predicted_value, uint32_t actual_value, - uint32_t palette_entry) { - int i; - uint32_t distance = 0; - AddPixelsEq(&predicted_value, palette_entry); - for (i = 0; i < 32; i += 8) { - const int32_t av = (actual_value >> i) & 0xff; - const int32_t pv = (predicted_value >> i) & 0xff; - distance += Square(pv - av); - } - // We sum square of intensity difference with factor 10, but because Intensity - // returns 100 times real intensity we need to multiply differences of colors - // by 1000. - distance *= 1000u; - distance += Square(Intensity(predicted_value) - - Intensity(actual_value)); - return distance; -} - -static uint32_t Predict(int x, int y, uint32_t* image) { - const uint32_t t = (y == 0) ? ARGB_BLACK : image[x]; - const uint32_t l = (x == 0) ? ARGB_BLACK : image[x - 1]; - const uint32_t p = - (((((t >> 24) & 0xff) + ((l >> 24) & 0xff)) / 2) << 24) + - (((((t >> 16) & 0xff) + ((l >> 16) & 0xff)) / 2) << 16) + - (((((t >> 8) & 0xff) + ((l >> 8) & 0xff)) / 2) << 8) + - (((((t >> 0) & 0xff) + ((l >> 0) & 0xff)) / 2) << 0); - if (x == 0 && y == 0) return ARGB_BLACK; - if (x == 0) return t; - if (y == 0) return l; - return p; -} - -static WEBP_INLINE int AddSubtractComponentFullWithCoefficient( - int a, int b, int c) { - return Clip255(a + ((b - c) >> 2)); -} - -static WEBP_INLINE uint32_t ClampedAddSubtractFullWithCoefficient( - uint32_t c0, uint32_t c1, uint32_t c2) { - const int a = AddSubtractComponentFullWithCoefficient( - c0 >> 24, c1 >> 24, c2 >> 24); - const int r = AddSubtractComponentFullWithCoefficient((c0 >> 16) & 0xff, - (c1 >> 16) & 0xff, - (c2 >> 16) & 0xff); - const int g = AddSubtractComponentFullWithCoefficient((c0 >> 8) & 0xff, - (c1 >> 8) & 0xff, - (c2 >> 8) & 0xff); - const int b = AddSubtractComponentFullWithCoefficient( - c0 & 0xff, c1 & 0xff, c2 & 0xff); - return ((uint32_t)a << 24) | (r << 16) | (g << 8) | b; -} - -//------------------------------------------------------------------------------ - -// Find palette entry with minimum error from difference of actual pixel value -// and predicted pixel value. Propagate error of pixel to its top and left pixel -// in src array. Write predicted_value + palette_entry to new_image. Return -// index of best palette entry. -static int FindBestPaletteEntry(uint32_t src, uint32_t predicted_value, - const uint32_t palette[], int palette_size) { - int i; - int idx = 0; - uint32_t best_distance = CalcDist(predicted_value, src, palette[0]); - for (i = 1; i < palette_size; ++i) { - const uint32_t distance = CalcDist(predicted_value, src, palette[i]); - if (distance < best_distance) { - best_distance = distance; - idx = i; - } - } - return idx; -} - -static void ApplyBestPaletteEntry(int x, int y, - uint32_t new_value, uint32_t palette_value, - uint32_t* src, int src_stride, - uint32_t* new_image) { - AddPixelsEq(&new_value, palette_value); - if (x > 0) { - src[x - 1] = ClampedAddSubtractFullWithCoefficient(src[x - 1], - new_value, src[x]); - } - if (y > 0) { - src[x - src_stride] = - ClampedAddSubtractFullWithCoefficient(src[x - src_stride], - new_value, src[x]); - } - new_image[x] = new_value; -} - -//------------------------------------------------------------------------------ -// Main entry point - -static WebPEncodingError ApplyDeltaPalette(uint32_t* src, uint32_t* dst, - uint32_t src_stride, - uint32_t dst_stride, - const uint32_t* palette, - int palette_size, - int width, int height, - int num_passes) { - int x, y; - WebPEncodingError err = VP8_ENC_OK; - uint32_t* new_image = (uint32_t*)WebPSafeMalloc(width, sizeof(*new_image)); - uint8_t* const tmp_row = (uint8_t*)WebPSafeMalloc(width, sizeof(*tmp_row)); - if (new_image == NULL || tmp_row == NULL) { - err = VP8_ENC_ERROR_OUT_OF_MEMORY; - goto Error; - } - - while (num_passes--) { - uint32_t* cur_src = src; - uint32_t* cur_dst = dst; - for (y = 0; y < height; ++y) { - for (x = 0; x < width; ++x) { - const uint32_t predicted_value = Predict(x, y, new_image); - tmp_row[x] = FindBestPaletteEntry(cur_src[x], predicted_value, - palette, palette_size); - ApplyBestPaletteEntry(x, y, predicted_value, palette[tmp_row[x]], - cur_src, src_stride, new_image); - } - for (x = 0; x < width; ++x) { - cur_dst[x] = palette[tmp_row[x]]; - } - cur_src += src_stride; - cur_dst += dst_stride; - } - } - Error: - WebPSafeFree(new_image); - WebPSafeFree(tmp_row); - return err; -} - -// replaces enc->argb_ by a palettizable approximation of it, -// and generates optimal enc->palette_[] -WebPEncodingError WebPSearchOptimalDeltaPalette(VP8LEncoder* const enc) { - const WebPPicture* const pic = enc->pic_; - uint32_t* src = pic->argb; - uint32_t* dst = enc->argb_; - const int width = pic->width; - const int height = pic->height; - - WebPEncodingError err = VP8_ENC_OK; - memcpy(enc->palette_, kDeltaPalette, sizeof(kDeltaPalette)); - enc->palette_[DELTA_PALETTE_SIZE - 1] = src[0] - 0xff000000u; - enc->palette_size_ = DELTA_PALETTE_SIZE; - err = ApplyDeltaPalette(src, dst, pic->argb_stride, enc->current_width_, - enc->palette_, enc->palette_size_, - width, height, 2); - if (err != VP8_ENC_OK) goto Error; - - Error: - return err; -} - -#else // !WEBP_EXPERIMENTAL_FEATURES - -WebPEncodingError WebPSearchOptimalDeltaPalette(VP8LEncoder* const enc) { - (void)enc; - return VP8_ENC_ERROR_INVALID_CONFIGURATION; -} - -#endif // WEBP_EXPERIMENTAL_FEATURES diff --git a/thirdparty/libwebp/src/enc/delta_palettization_enc.h b/thirdparty/libwebp/src/enc/delta_palettization_enc.h deleted file mode 100644 index b15e2cd487..0000000000 --- a/thirdparty/libwebp/src/enc/delta_palettization_enc.h +++ /dev/null @@ -1,25 +0,0 @@ -// Copyright 2015 Google Inc. All Rights Reserved. -// -// Use of this source code is governed by a BSD-style license -// that can be found in the COPYING file in the root of the source -// tree. An additional intellectual property rights grant can be found -// in the file PATENTS. All contributing project authors may -// be found in the AUTHORS file in the root of the source tree. -// ----------------------------------------------------------------------------- -// -// Author: Mislav Bradac (mislavm@google.com) -// - -#ifndef WEBP_ENC_DELTA_PALETTIZATION_ENC_H_ -#define WEBP_ENC_DELTA_PALETTIZATION_ENC_H_ - -#include "src/webp/encode.h" -#include "src/enc/vp8li_enc.h" - -// Replaces enc->argb_[] input by a palettizable approximation of it, -// and generates optimal enc->palette_[]. -// This function can revert enc->use_palette_ / enc->use_predict_ flag -// if delta-palettization is not producing expected saving. -WebPEncodingError WebPSearchOptimalDeltaPalette(VP8LEncoder* const enc); - -#endif // WEBP_ENC_DELTA_PALETTIZATION_ENC_H_ diff --git a/thirdparty/libwebp/src/enc/histogram_enc.c b/thirdparty/libwebp/src/enc/histogram_enc.c index 9fdbc627a1..4e49e0a201 100644 --- a/thirdparty/libwebp/src/enc/histogram_enc.c +++ b/thirdparty/libwebp/src/enc/histogram_enc.c @@ -51,10 +51,12 @@ static void HistogramCopy(const VP8LHistogram* const src, VP8LHistogram* const dst) { uint32_t* const dst_literal = dst->literal_; const int dst_cache_bits = dst->palette_code_bits_; + const int literal_size = VP8LHistogramNumCodes(dst_cache_bits); const int histo_size = VP8LGetHistogramSize(dst_cache_bits); assert(src->palette_code_bits_ == dst_cache_bits); memcpy(dst, src, histo_size); dst->literal_ = dst_literal; + memcpy(dst->literal_, src->literal_, literal_size * sizeof(*dst->literal_)); } int VP8LGetHistogramSize(int cache_bits) { @@ -91,9 +93,19 @@ void VP8LHistogramCreate(VP8LHistogram* const p, VP8LHistogramStoreRefs(refs, p); } -void VP8LHistogramInit(VP8LHistogram* const p, int palette_code_bits) { +void VP8LHistogramInit(VP8LHistogram* const p, int palette_code_bits, + int init_arrays) { p->palette_code_bits_ = palette_code_bits; - HistogramClear(p); + if (init_arrays) { + HistogramClear(p); + } else { + p->trivial_symbol_ = 0; + p->bit_cost_ = 0.; + p->literal_cost_ = 0.; + p->red_cost_ = 0.; + p->blue_cost_ = 0.; + memset(p->is_used_, 0, sizeof(p->is_used_)); + } } VP8LHistogram* VP8LAllocateHistogram(int cache_bits) { @@ -104,37 +116,70 @@ VP8LHistogram* VP8LAllocateHistogram(int cache_bits) { histo = (VP8LHistogram*)memory; // literal_ won't necessary be aligned. histo->literal_ = (uint32_t*)(memory + sizeof(VP8LHistogram)); - VP8LHistogramInit(histo, cache_bits); + VP8LHistogramInit(histo, cache_bits, /*init_arrays=*/ 0); return histo; } +// Resets the pointers of the histograms to point to the bit buffer in the set. +static void HistogramSetResetPointers(VP8LHistogramSet* const set, + int cache_bits) { + int i; + const int histo_size = VP8LGetHistogramSize(cache_bits); + uint8_t* memory = (uint8_t*) (set->histograms); + memory += set->max_size * sizeof(*set->histograms); + for (i = 0; i < set->max_size; ++i) { + memory = (uint8_t*) WEBP_ALIGN(memory); + set->histograms[i] = (VP8LHistogram*) memory; + // literal_ won't necessary be aligned. + set->histograms[i]->literal_ = (uint32_t*)(memory + sizeof(VP8LHistogram)); + memory += histo_size; + } +} + +// Returns the total size of the VP8LHistogramSet. +static size_t HistogramSetTotalSize(int size, int cache_bits) { + const int histo_size = VP8LGetHistogramSize(cache_bits); + return (sizeof(VP8LHistogramSet) + size * (sizeof(VP8LHistogram*) + + histo_size + WEBP_ALIGN_CST)); +} + VP8LHistogramSet* VP8LAllocateHistogramSet(int size, int cache_bits) { int i; VP8LHistogramSet* set; - const int histo_size = VP8LGetHistogramSize(cache_bits); - const size_t total_size = - sizeof(*set) + size * (sizeof(*set->histograms) + - histo_size + WEBP_ALIGN_CST); + const size_t total_size = HistogramSetTotalSize(size, cache_bits); uint8_t* memory = (uint8_t*)WebPSafeMalloc(total_size, sizeof(*memory)); if (memory == NULL) return NULL; set = (VP8LHistogramSet*)memory; memory += sizeof(*set); set->histograms = (VP8LHistogram**)memory; - memory += size * sizeof(*set->histograms); set->max_size = size; set->size = size; + HistogramSetResetPointers(set, cache_bits); for (i = 0; i < size; ++i) { - memory = (uint8_t*)WEBP_ALIGN(memory); - set->histograms[i] = (VP8LHistogram*)memory; - // literal_ won't necessary be aligned. - set->histograms[i]->literal_ = (uint32_t*)(memory + sizeof(VP8LHistogram)); - VP8LHistogramInit(set->histograms[i], cache_bits); - memory += histo_size; + VP8LHistogramInit(set->histograms[i], cache_bits, /*init_arrays=*/ 0); } return set; } +void VP8LHistogramSetClear(VP8LHistogramSet* const set) { + int i; + const int cache_bits = set->histograms[0]->palette_code_bits_; + const int size = set->size; + const size_t total_size = HistogramSetTotalSize(size, cache_bits); + uint8_t* memory = (uint8_t*)set; + + memset(memory, 0, total_size); + memory += sizeof(*set); + set->histograms = (VP8LHistogram**)memory; + set->max_size = size; + set->size = size; + HistogramSetResetPointers(set, cache_bits); + for (i = 0; i < size; ++i) { + set->histograms[i]->palette_code_bits_ = cache_bits; + } +} + // ----------------------------------------------------------------------------- void VP8LHistogramAddSinglePixOrCopy(VP8LHistogram* const histo, @@ -237,7 +282,8 @@ static double FinalHuffmanCost(const VP8LStreaks* const stats) { // Get the symbol entropy for the distribution 'population'. // Set 'trivial_sym', if there's only one symbol present in the distribution. static double PopulationCost(const uint32_t* const population, int length, - uint32_t* const trivial_sym) { + uint32_t* const trivial_sym, + uint8_t* const is_used) { VP8LBitEntropy bit_entropy; VP8LStreaks stats; VP8LGetEntropyUnrefined(population, length, &bit_entropy, &stats); @@ -245,6 +291,8 @@ static double PopulationCost(const uint32_t* const population, int length, *trivial_sym = (bit_entropy.nonzeros == 1) ? bit_entropy.nonzero_code : VP8L_NON_TRIVIAL_SYM; } + // The histogram is used if there is at least one non-zero streak. + *is_used = (stats.streaks[1][0] != 0 || stats.streaks[1][1] != 0); return BitsEntropyRefine(&bit_entropy) + FinalHuffmanCost(&stats); } @@ -253,7 +301,9 @@ static double PopulationCost(const uint32_t* const population, int length, // non-zero: both the zero-th one, or both the last one. static WEBP_INLINE double GetCombinedEntropy(const uint32_t* const X, const uint32_t* const Y, - int length, int trivial_at_end) { + int length, int is_X_used, + int is_Y_used, + int trivial_at_end) { VP8LStreaks stats; if (trivial_at_end) { // This configuration is due to palettization that transforms an indexed @@ -262,28 +312,43 @@ static WEBP_INLINE double GetCombinedEntropy(const uint32_t* const X, // Only FinalHuffmanCost needs to be evaluated. memset(&stats, 0, sizeof(stats)); // Deal with the non-zero value at index 0 or length-1. - stats.streaks[1][0] += 1; + stats.streaks[1][0] = 1; // Deal with the following/previous zero streak. - stats.counts[0] += 1; - stats.streaks[0][1] += length - 1; + stats.counts[0] = 1; + stats.streaks[0][1] = length - 1; return FinalHuffmanCost(&stats); } else { VP8LBitEntropy bit_entropy; - VP8LGetCombinedEntropyUnrefined(X, Y, length, &bit_entropy, &stats); + if (is_X_used) { + if (is_Y_used) { + VP8LGetCombinedEntropyUnrefined(X, Y, length, &bit_entropy, &stats); + } else { + VP8LGetEntropyUnrefined(X, length, &bit_entropy, &stats); + } + } else { + if (is_Y_used) { + VP8LGetEntropyUnrefined(Y, length, &bit_entropy, &stats); + } else { + memset(&stats, 0, sizeof(stats)); + stats.counts[0] = 1; + stats.streaks[0][length > 3] = length; + VP8LBitEntropyInit(&bit_entropy); + } + } return BitsEntropyRefine(&bit_entropy) + FinalHuffmanCost(&stats); } } // Estimates the Entropy + Huffman + other block overhead size cost. -double VP8LHistogramEstimateBits(const VP8LHistogram* const p) { +double VP8LHistogramEstimateBits(VP8LHistogram* const p) { return - PopulationCost( - p->literal_, VP8LHistogramNumCodes(p->palette_code_bits_), NULL) - + PopulationCost(p->red_, NUM_LITERAL_CODES, NULL) - + PopulationCost(p->blue_, NUM_LITERAL_CODES, NULL) - + PopulationCost(p->alpha_, NUM_LITERAL_CODES, NULL) - + PopulationCost(p->distance_, NUM_DISTANCE_CODES, NULL) + PopulationCost(p->literal_, VP8LHistogramNumCodes(p->palette_code_bits_), + NULL, &p->is_used_[0]) + + PopulationCost(p->red_, NUM_LITERAL_CODES, NULL, &p->is_used_[1]) + + PopulationCost(p->blue_, NUM_LITERAL_CODES, NULL, &p->is_used_[2]) + + PopulationCost(p->alpha_, NUM_LITERAL_CODES, NULL, &p->is_used_[3]) + + PopulationCost(p->distance_, NUM_DISTANCE_CODES, NULL, &p->is_used_[4]) + VP8LExtraCost(p->literal_ + NUM_LITERAL_CODES, NUM_LENGTH_CODES) + VP8LExtraCost(p->distance_, NUM_DISTANCE_CODES); } @@ -299,7 +364,8 @@ static int GetCombinedHistogramEntropy(const VP8LHistogram* const a, int trivial_at_end = 0; assert(a->palette_code_bits_ == b->palette_code_bits_); *cost += GetCombinedEntropy(a->literal_, b->literal_, - VP8LHistogramNumCodes(palette_code_bits), 0); + VP8LHistogramNumCodes(palette_code_bits), + a->is_used_[0], b->is_used_[0], 0); *cost += VP8LExtraCostCombined(a->literal_ + NUM_LITERAL_CODES, b->literal_ + NUM_LITERAL_CODES, NUM_LENGTH_CODES); @@ -319,19 +385,23 @@ static int GetCombinedHistogramEntropy(const VP8LHistogram* const a, } *cost += - GetCombinedEntropy(a->red_, b->red_, NUM_LITERAL_CODES, trivial_at_end); + GetCombinedEntropy(a->red_, b->red_, NUM_LITERAL_CODES, a->is_used_[1], + b->is_used_[1], trivial_at_end); if (*cost > cost_threshold) return 0; *cost += - GetCombinedEntropy(a->blue_, b->blue_, NUM_LITERAL_CODES, trivial_at_end); + GetCombinedEntropy(a->blue_, b->blue_, NUM_LITERAL_CODES, a->is_used_[2], + b->is_used_[2], trivial_at_end); if (*cost > cost_threshold) return 0; - *cost += GetCombinedEntropy(a->alpha_, b->alpha_, NUM_LITERAL_CODES, - trivial_at_end); + *cost += + GetCombinedEntropy(a->alpha_, b->alpha_, NUM_LITERAL_CODES, + a->is_used_[3], b->is_used_[3], trivial_at_end); if (*cost > cost_threshold) return 0; *cost += - GetCombinedEntropy(a->distance_, b->distance_, NUM_DISTANCE_CODES, 0); + GetCombinedEntropy(a->distance_, b->distance_, NUM_DISTANCE_CODES, + a->is_used_[4], b->is_used_[4], 0); *cost += VP8LExtraCostCombined(a->distance_, b->distance_, NUM_DISTANCE_CODES); if (*cost > cost_threshold) return 0; @@ -419,16 +489,19 @@ static void UpdateDominantCostRange( static void UpdateHistogramCost(VP8LHistogram* const h) { uint32_t alpha_sym, red_sym, blue_sym; const double alpha_cost = - PopulationCost(h->alpha_, NUM_LITERAL_CODES, &alpha_sym); + PopulationCost(h->alpha_, NUM_LITERAL_CODES, &alpha_sym, + &h->is_used_[3]); const double distance_cost = - PopulationCost(h->distance_, NUM_DISTANCE_CODES, NULL) + + PopulationCost(h->distance_, NUM_DISTANCE_CODES, NULL, &h->is_used_[4]) + VP8LExtraCost(h->distance_, NUM_DISTANCE_CODES); const int num_codes = VP8LHistogramNumCodes(h->palette_code_bits_); - h->literal_cost_ = PopulationCost(h->literal_, num_codes, NULL) + - VP8LExtraCost(h->literal_ + NUM_LITERAL_CODES, - NUM_LENGTH_CODES); - h->red_cost_ = PopulationCost(h->red_, NUM_LITERAL_CODES, &red_sym); - h->blue_cost_ = PopulationCost(h->blue_, NUM_LITERAL_CODES, &blue_sym); + h->literal_cost_ = + PopulationCost(h->literal_, num_codes, NULL, &h->is_used_[0]) + + VP8LExtraCost(h->literal_ + NUM_LITERAL_CODES, NUM_LENGTH_CODES); + h->red_cost_ = + PopulationCost(h->red_, NUM_LITERAL_CODES, &red_sym, &h->is_used_[1]); + h->blue_cost_ = + PopulationCost(h->blue_, NUM_LITERAL_CODES, &blue_sym, &h->is_used_[2]); h->bit_cost_ = h->literal_cost_ + h->red_cost_ + h->blue_cost_ + alpha_cost + distance_cost; if ((alpha_sym | red_sym | blue_sym) == VP8L_NON_TRIVIAL_SYM) { @@ -473,6 +546,7 @@ static void HistogramBuild( VP8LHistogram** const histograms = image_histo->histograms; VP8LRefsCursor c = VP8LRefsCursorInit(backward_refs); assert(histo_bits > 0); + VP8LHistogramSetClear(image_histo); while (VP8LRefsCursorOk(&c)) { const PixOrCopy* const v = c.cur_pos; const int ix = (y >> histo_bits) * histo_xsize + (x >> histo_bits); @@ -493,11 +567,19 @@ static void HistogramCopyAndAnalyze( const int histo_size = orig_histo->size; VP8LHistogram** const orig_histograms = orig_histo->histograms; VP8LHistogram** const histograms = image_histo->histograms; + image_histo->size = 0; for (i = 0; i < histo_size; ++i) { VP8LHistogram* const histo = orig_histograms[i]; UpdateHistogramCost(histo); + + // Skip the histogram if it is completely empty, which can happen for tiles + // with no information (when they are skipped because of LZ77). + if (!histo->is_used_[0] && !histo->is_used_[1] && !histo->is_used_[2] + && !histo->is_used_[3] && !histo->is_used_[4]) { + continue; + } // Copy histograms from orig_histo[] to image_histo[]. - HistogramCopy(histo, histograms[i]); + HistogramCopy(histo, histograms[image_histo->size++]); } } @@ -674,6 +756,18 @@ static void HistoQueueUpdateHead(HistoQueue* const histo_queue, } } +// Update the cost diff and combo of a pair of histograms. This needs to be +// called when the the histograms have been merged with a third one. +static void HistoQueueUpdatePair(const VP8LHistogram* const h1, + const VP8LHistogram* const h2, + double threshold, + HistogramPair* const pair) { + const double sum_cost = h1->bit_cost_ + h2->bit_cost_; + pair->cost_combo = 0.; + GetCombinedHistogramEntropy(h1, h2, sum_cost + threshold, &pair->cost_combo); + pair->cost_diff = pair->cost_combo - sum_cost; +} + // Create a pair from indices "idx1" and "idx2" provided its cost // is inferior to "threshold", a negative entropy. // It returns the cost of the pair, or 0. if it superior to threshold. @@ -683,7 +777,6 @@ static double HistoQueuePush(HistoQueue* const histo_queue, const VP8LHistogram* h1; const VP8LHistogram* h2; HistogramPair pair; - double sum_cost; assert(threshold <= 0.); if (idx1 > idx2) { @@ -695,10 +788,8 @@ static double HistoQueuePush(HistoQueue* const histo_queue, pair.idx2 = idx2; h1 = histograms[idx1]; h2 = histograms[idx2]; - sum_cost = h1->bit_cost_ + h2->bit_cost_; - pair.cost_combo = 0.; - GetCombinedHistogramEntropy(h1, h2, sum_cost + threshold, &pair.cost_combo); - pair.cost_diff = pair.cost_combo - sum_cost; + + HistoQueueUpdatePair(h1, h2, threshold, &pair); // Do not even consider the pair if it does not improve the entropy. if (pair.cost_diff >= threshold) return 0.; @@ -891,8 +982,7 @@ static int HistogramCombineStochastic(VP8LHistogramSet* const image_histo, } if (do_eval) { // Re-evaluate the cost of an updated pair. - GetCombinedHistogramEntropy(histograms[p->idx1], histograms[p->idx2], 0, - &p->cost_diff); + HistoQueueUpdatePair(histograms[p->idx1], histograms[p->idx2], 0., p); if (p->cost_diff >= 0.) { HistoQueuePopPair(&histo_queue, p); continue; @@ -987,8 +1077,7 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize, // histograms of small sizes (as bin_map will be very sparse) and // maximum quality q==100 (to preserve the compression gains at that level). const int entropy_combine_num_bins = low_effort ? NUM_PARTITIONS : BIN_SIZE; - const int entropy_combine = - (orig_histo->size > entropy_combine_num_bins * 2) && (quality < 100); + int entropy_combine; if (orig_histo == NULL) goto Error; @@ -996,15 +1085,16 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize, HistogramBuild(xsize, histo_bits, refs, orig_histo); // Copies the histograms and computes its bit_cost. HistogramCopyAndAnalyze(orig_histo, image_histo); - + entropy_combine = + (image_histo->size > entropy_combine_num_bins * 2) && (quality < 100); if (entropy_combine) { - const int bin_map_size = orig_histo->size; + const int bin_map_size = image_histo->size; // Reuse histogram_symbols storage. By definition, it's guaranteed to be ok. uint16_t* const bin_map = histogram_symbols; const double combine_cost_factor = GetCombineCostFactor(image_histo_raw_size, quality); - HistogramAnalyzeEntropyBin(orig_histo, bin_map, low_effort); + HistogramAnalyzeEntropyBin(image_histo, bin_map, low_effort); // Collapse histograms with similar entropy. HistogramCombineEntropyBin(image_histo, tmp_histo, bin_map, bin_map_size, entropy_combine_num_bins, combine_cost_factor, diff --git a/thirdparty/libwebp/src/enc/histogram_enc.h b/thirdparty/libwebp/src/enc/histogram_enc.h index e8c4c83f6f..54c2d21783 100644 --- a/thirdparty/libwebp/src/enc/histogram_enc.h +++ b/thirdparty/libwebp/src/enc/histogram_enc.h @@ -44,6 +44,7 @@ typedef struct { double literal_cost_; // Cached values of dominant entropy costs: double red_cost_; // literal, red & blue. double blue_cost_; + uint8_t is_used_[5]; // 5 for literal, red, blue, alpha, distance } VP8LHistogram; // Collection of histograms with fixed capacity, allocated as one @@ -67,7 +68,9 @@ void VP8LHistogramCreate(VP8LHistogram* const p, int VP8LGetHistogramSize(int palette_code_bits); // Set the palette_code_bits and reset the stats. -void VP8LHistogramInit(VP8LHistogram* const p, int palette_code_bits); +// If init_arrays is true, the arrays are also filled with 0's. +void VP8LHistogramInit(VP8LHistogram* const p, int palette_code_bits, + int init_arrays); // Collect all the references into a histogram (without reset) void VP8LHistogramStoreRefs(const VP8LBackwardRefs* const refs, @@ -83,6 +86,9 @@ void VP8LFreeHistogramSet(VP8LHistogramSet* const histo); // using 'cache_bits'. Return NULL in case of memory error. VP8LHistogramSet* VP8LAllocateHistogramSet(int size, int cache_bits); +// Set the histograms in set to 0. +void VP8LHistogramSetClear(VP8LHistogramSet* const set); + // Allocate and initialize histogram object with specified 'cache_bits'. // Returns NULL in case of memory error. // Special case of VP8LAllocateHistogramSet, with size equals 1. @@ -113,7 +119,7 @@ double VP8LBitsEntropy(const uint32_t* const array, int n); // Estimate how many bits the combined entropy of literals and distance // approximately maps to. -double VP8LHistogramEstimateBits(const VP8LHistogram* const p); +double VP8LHistogramEstimateBits(VP8LHistogram* const p); #ifdef __cplusplus } diff --git a/thirdparty/libwebp/src/enc/iterator_enc.c b/thirdparty/libwebp/src/enc/iterator_enc.c index 7c47d51272..29f91d8315 100644 --- a/thirdparty/libwebp/src/enc/iterator_enc.c +++ b/thirdparty/libwebp/src/enc/iterator_enc.c @@ -128,7 +128,7 @@ static void ImportLine(const uint8_t* src, int src_stride, for (; i < total_len; ++i) dst[i] = dst[len - 1]; } -void VP8IteratorImport(VP8EncIterator* const it, uint8_t* tmp_32) { +void VP8IteratorImport(VP8EncIterator* const it, uint8_t* const tmp_32) { const VP8Encoder* const enc = it->enc_; const int x = it->x_, y = it->y_; const WebPPicture* const pic = enc->pic_; diff --git a/thirdparty/libwebp/src/enc/picture_tools_enc.c b/thirdparty/libwebp/src/enc/picture_tools_enc.c index be292d4391..d0e8a495da 100644 --- a/thirdparty/libwebp/src/enc/picture_tools_enc.c +++ b/thirdparty/libwebp/src/enc/picture_tools_enc.c @@ -16,10 +16,6 @@ #include "src/enc/vp8i_enc.h" #include "src/dsp/yuv.h" -static WEBP_INLINE uint32_t MakeARGB32(int r, int g, int b) { - return (0xff000000u | (r << 16) | (g << 8) | b); -} - //------------------------------------------------------------------------------ // Helper: clean up fully transparent area to help compressibility. @@ -195,6 +191,10 @@ void WebPCleanupTransparentAreaLossless(WebPPicture* const pic) { #define BLEND_10BIT(V0, V1, ALPHA) \ ((((V0) * (1020 - (ALPHA)) + (V1) * (ALPHA)) * 0x101 + 1024) >> 18) +static WEBP_INLINE uint32_t MakeARGB32(int r, int g, int b) { + return (0xff000000u | (r << 16) | (g << 8) | b); +} + void WebPBlendAlpha(WebPPicture* pic, uint32_t background_rgb) { const int red = (background_rgb >> 16) & 0xff; const int green = (background_rgb >> 8) & 0xff; @@ -208,39 +208,44 @@ void WebPBlendAlpha(WebPPicture* pic, uint32_t background_rgb) { const int U0 = VP8RGBToU(4 * red, 4 * green, 4 * blue, 4 * YUV_HALF); const int V0 = VP8RGBToV(4 * red, 4 * green, 4 * blue, 4 * YUV_HALF); const int has_alpha = pic->colorspace & WEBP_CSP_ALPHA_BIT; - if (!has_alpha || pic->a == NULL) return; // nothing to do + uint8_t* y_ptr = pic->y; + uint8_t* u_ptr = pic->u; + uint8_t* v_ptr = pic->v; + uint8_t* a_ptr = pic->a; + if (!has_alpha || a_ptr == NULL) return; // nothing to do for (y = 0; y < pic->height; ++y) { // Luma blending - uint8_t* const y_ptr = pic->y + y * pic->y_stride; - uint8_t* const a_ptr = pic->a + y * pic->a_stride; for (x = 0; x < pic->width; ++x) { - const int alpha = a_ptr[x]; + const uint8_t alpha = a_ptr[x]; if (alpha < 0xff) { - y_ptr[x] = BLEND(Y0, y_ptr[x], a_ptr[x]); + y_ptr[x] = BLEND(Y0, y_ptr[x], alpha); } } // Chroma blending every even line if ((y & 1) == 0) { - uint8_t* const u = pic->u + (y >> 1) * pic->uv_stride; - uint8_t* const v = pic->v + (y >> 1) * pic->uv_stride; uint8_t* const a_ptr2 = (y + 1 == pic->height) ? a_ptr : a_ptr + pic->a_stride; for (x = 0; x < uv_width; ++x) { // Average four alpha values into a single blending weight. // TODO(skal): might lead to visible contouring. Can we do better? - const int alpha = + const uint32_t alpha = a_ptr[2 * x + 0] + a_ptr[2 * x + 1] + a_ptr2[2 * x + 0] + a_ptr2[2 * x + 1]; - u[x] = BLEND_10BIT(U0, u[x], alpha); - v[x] = BLEND_10BIT(V0, v[x], alpha); + u_ptr[x] = BLEND_10BIT(U0, u_ptr[x], alpha); + v_ptr[x] = BLEND_10BIT(V0, v_ptr[x], alpha); } if (pic->width & 1) { // rightmost pixel - const int alpha = 2 * (a_ptr[2 * x + 0] + a_ptr2[2 * x + 0]); - u[x] = BLEND_10BIT(U0, u[x], alpha); - v[x] = BLEND_10BIT(V0, v[x], alpha); + const uint32_t alpha = 2 * (a_ptr[2 * x + 0] + a_ptr2[2 * x + 0]); + u_ptr[x] = BLEND_10BIT(U0, u_ptr[x], alpha); + v_ptr[x] = BLEND_10BIT(V0, v_ptr[x], alpha); } + } else { + u_ptr += pic->uv_stride; + v_ptr += pic->uv_stride; } - memset(a_ptr, 0xff, pic->width); + memset(a_ptr, 0xff, pic->width); // reset alpha value to opaque + a_ptr += pic->a_stride; + y_ptr += pic->y_stride; } } else { uint32_t* argb = pic->argb; diff --git a/thirdparty/libwebp/src/enc/vp8i_enc.h b/thirdparty/libwebp/src/enc/vp8i_enc.h index 624e8f8e66..92439febb8 100644 --- a/thirdparty/libwebp/src/enc/vp8i_enc.h +++ b/thirdparty/libwebp/src/enc/vp8i_enc.h @@ -32,7 +32,7 @@ extern "C" { // version numbers #define ENC_MAJ_VERSION 1 #define ENC_MIN_VERSION 0 -#define ENC_REV_VERSION 0 +#define ENC_REV_VERSION 1 enum { MAX_LF_LEVELS = 64, // Maximum loop filter level MAX_VARIABLE_LEVEL = 67, // last (inclusive) level with variable cost @@ -278,7 +278,7 @@ int VP8IteratorIsDone(const VP8EncIterator* const it); // Import uncompressed samples from source. // If tmp_32 is not NULL, import boundary samples too. // tmp_32 is a 32-bytes scratch buffer that must be aligned in memory. -void VP8IteratorImport(VP8EncIterator* const it, uint8_t* tmp_32); +void VP8IteratorImport(VP8EncIterator* const it, uint8_t* const tmp_32); // export decimated samples void VP8IteratorExport(const VP8EncIterator* const it); // go to next macroblock. Returns false if not finished. @@ -515,4 +515,4 @@ void WebPCleanupTransparentAreaLossless(WebPPicture* const pic); } // extern "C" #endif -#endif /* WEBP_ENC_VP8I_ENC_H_ */ +#endif // WEBP_ENC_VP8I_ENC_H_ diff --git a/thirdparty/libwebp/src/enc/vp8l_enc.c b/thirdparty/libwebp/src/enc/vp8l_enc.c index a89184eb08..2713edcd95 100644 --- a/thirdparty/libwebp/src/enc/vp8l_enc.c +++ b/thirdparty/libwebp/src/enc/vp8l_enc.c @@ -809,6 +809,7 @@ static WebPEncodingError EncodeImageNoHuffman(VP8LBitWriter* const bw, err = VP8_ENC_ERROR_OUT_OF_MEMORY; goto Error; } + VP8LHistogramSetClear(histogram_image); // Build histogram image and symbols from backward references. VP8LHistogramStoreRefs(refs, histogram_image->histograms[0]); @@ -1248,14 +1249,20 @@ static WebPEncodingError MakeInputImageCopy(VP8LEncoder* const enc) { const WebPPicture* const picture = enc->pic_; const int width = picture->width; const int height = picture->height; - int y; + err = AllocateTransformBuffer(enc, width, height); if (err != VP8_ENC_OK) return err; if (enc->argb_content_ == kEncoderARGB) return VP8_ENC_OK; - for (y = 0; y < height; ++y) { - memcpy(enc->argb_ + y * width, - picture->argb + y * picture->argb_stride, - width * sizeof(*enc->argb_)); + + { + uint32_t* dst = enc->argb_; + const uint32_t* src = picture->argb; + int y; + for (y = 0; y < height; ++y) { + memcpy(dst, src, width * sizeof(*dst)); + dst += width; + src += picture->argb_stride; + } } enc->argb_content_ = kEncoderARGB; assert(enc->current_width_ == width); diff --git a/thirdparty/libwebp/src/enc/vp8li_enc.h b/thirdparty/libwebp/src/enc/vp8li_enc.h index 298a4a0014..d2d0fc509c 100644 --- a/thirdparty/libwebp/src/enc/vp8li_enc.h +++ b/thirdparty/libwebp/src/enc/vp8li_enc.h @@ -115,4 +115,4 @@ void VP8LColorSpaceTransform(int width, int height, int bits, int quality, } // extern "C" #endif -#endif /* WEBP_ENC_VP8LI_ENC_H_ */ +#endif // WEBP_ENC_VP8LI_ENC_H_ diff --git a/thirdparty/libwebp/src/mux/animi.h b/thirdparty/libwebp/src/mux/animi.h index 88899532aa..34c45ba4da 100644 --- a/thirdparty/libwebp/src/mux/animi.h +++ b/thirdparty/libwebp/src/mux/animi.h @@ -40,4 +40,4 @@ int WebPAnimEncoderRefineRect( } // extern "C" #endif -#endif /* WEBP_MUX_ANIMI_H_ */ +#endif // WEBP_MUX_ANIMI_H_ diff --git a/thirdparty/libwebp/src/mux/muxedit.c b/thirdparty/libwebp/src/mux/muxedit.c index 7a027b3cb4..ccf14b2a0c 100644 --- a/thirdparty/libwebp/src/mux/muxedit.c +++ b/thirdparty/libwebp/src/mux/muxedit.c @@ -69,12 +69,12 @@ void WebPMuxDelete(WebPMux* mux) { if (idx == (INDEX)) { \ err = ChunkAssignData(&chunk, data, copy_data, tag); \ if (err == WEBP_MUX_OK) { \ - err = ChunkSetNth(&chunk, (LIST), nth); \ + err = ChunkSetHead(&chunk, (LIST)); \ } \ return err; \ } -static WebPMuxError MuxSet(WebPMux* const mux, uint32_t tag, uint32_t nth, +static WebPMuxError MuxSet(WebPMux* const mux, uint32_t tag, const WebPData* const data, int copy_data) { WebPChunk chunk; WebPMuxError err = WEBP_MUX_NOT_FOUND; @@ -190,7 +190,7 @@ WebPMuxError WebPMuxSetChunk(WebPMux* mux, const char fourcc[4], if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err; // Add the given chunk. - return MuxSet(mux, tag, 1, chunk_data, copy_data); + return MuxSet(mux, tag, chunk_data, copy_data); } // Creates a chunk from given 'data' and sets it as 1st chunk in 'chunk_list'. @@ -202,7 +202,7 @@ static WebPMuxError AddDataToChunkList( ChunkInit(&chunk); err = ChunkAssignData(&chunk, data, copy_data, tag); if (err != WEBP_MUX_OK) goto Err; - err = ChunkSetNth(&chunk, chunk_list, 1); + err = ChunkSetHead(&chunk, chunk_list); if (err != WEBP_MUX_OK) goto Err; return WEBP_MUX_OK; Err: @@ -266,14 +266,14 @@ WebPMuxError WebPMuxPushFrame(WebPMux* mux, const WebPMuxFrameInfo* info, int copy_data) { WebPMuxImage wpi; WebPMuxError err; - const WebPData* const bitstream = &info->bitstream; // Sanity checks. if (mux == NULL || info == NULL) return WEBP_MUX_INVALID_ARGUMENT; if (info->id != WEBP_CHUNK_ANMF) return WEBP_MUX_INVALID_ARGUMENT; - if (bitstream->bytes == NULL || bitstream->size > MAX_CHUNK_PAYLOAD) { + if (info->bitstream.bytes == NULL || + info->bitstream.size > MAX_CHUNK_PAYLOAD) { return WEBP_MUX_INVALID_ARGUMENT; } @@ -287,7 +287,7 @@ WebPMuxError WebPMuxPushFrame(WebPMux* mux, const WebPMuxFrameInfo* info, } MuxImageInit(&wpi); - err = SetAlphaAndImageChunks(bitstream, copy_data, &wpi); + err = SetAlphaAndImageChunks(&info->bitstream, copy_data, &wpi); if (err != WEBP_MUX_OK) goto Err; assert(wpi.img_ != NULL); // As SetAlphaAndImageChunks() was successful. @@ -342,7 +342,7 @@ WebPMuxError WebPMuxSetAnimationParams(WebPMux* mux, // Set the animation parameters. PutLE32(data, params->bgcolor); PutLE16(data + 4, params->loop_count); - return MuxSet(mux, kChunks[IDX_ANIM].tag, 1, &anim, 1); + return MuxSet(mux, kChunks[IDX_ANIM].tag, &anim, 1); } WebPMuxError WebPMuxSetCanvasSize(WebPMux* mux, @@ -540,7 +540,7 @@ static WebPMuxError CreateVP8XChunk(WebPMux* const mux) { PutLE24(data + 4, width - 1); // canvas width. PutLE24(data + 7, height - 1); // canvas height. - return MuxSet(mux, kChunks[IDX_VP8X].tag, 1, &vp8x, 1); + return MuxSet(mux, kChunks[IDX_VP8X].tag, &vp8x, 1); } // Cleans up 'mux' by removing any unnecessary chunks. diff --git a/thirdparty/libwebp/src/mux/muxi.h b/thirdparty/libwebp/src/mux/muxi.h index 6b57eea30f..df9f74c63c 100644 --- a/thirdparty/libwebp/src/mux/muxi.h +++ b/thirdparty/libwebp/src/mux/muxi.h @@ -14,6 +14,7 @@ #ifndef WEBP_MUX_MUXI_H_ #define WEBP_MUX_MUXI_H_ +#include <assert.h> #include <stdlib.h> #include "src/dec/vp8i_dec.h" #include "src/dec/vp8li_dec.h" @@ -28,7 +29,7 @@ extern "C" { #define MUX_MAJ_VERSION 1 #define MUX_MIN_VERSION 0 -#define MUX_REV_VERSION 0 +#define MUX_REV_VERSION 1 // Chunk object. typedef struct WebPChunk WebPChunk; @@ -126,11 +127,14 @@ WebPChunk* ChunkSearchList(WebPChunk* first, uint32_t nth, uint32_t tag); WebPMuxError ChunkAssignData(WebPChunk* chunk, const WebPData* const data, int copy_data, uint32_t tag); -// Sets 'chunk' at nth position in the 'chunk_list'. -// nth = 0 has the special meaning "last of the list". +// Sets 'chunk' as the only element in 'chunk_list' if it is empty. // On success ownership is transferred from 'chunk' to the 'chunk_list'. -WebPMuxError ChunkSetNth(WebPChunk* chunk, WebPChunk** chunk_list, - uint32_t nth); +WebPMuxError ChunkSetHead(WebPChunk* const chunk, WebPChunk** const chunk_list); +// Sets 'chunk' at last position in the 'chunk_list'. +// On success ownership is transferred from 'chunk' to the 'chunk_list'. +// *chunk_list also points towards the last valid element of the initial +// *chunk_list. +WebPMuxError ChunkAppend(WebPChunk* const chunk, WebPChunk*** const chunk_list); // Releases chunk and returns chunk->next_. WebPChunk* ChunkRelease(WebPChunk* const chunk); @@ -143,13 +147,13 @@ void ChunkListDelete(WebPChunk** const chunk_list); // Returns size of the chunk including chunk header and padding byte (if any). static WEBP_INLINE size_t SizeWithPadding(size_t chunk_size) { + assert(chunk_size <= MAX_CHUNK_PAYLOAD); return CHUNK_HEADER_SIZE + ((chunk_size + 1) & ~1U); } // Size of a chunk including header and padding. static WEBP_INLINE size_t ChunkDiskSize(const WebPChunk* chunk) { const size_t data_size = chunk->data_.size; - assert(data_size < MAX_CHUNK_PAYLOAD); return SizeWithPadding(data_size); } @@ -227,4 +231,4 @@ WebPMuxError MuxValidate(const WebPMux* const mux); } // extern "C" #endif -#endif /* WEBP_MUX_MUXI_H_ */ +#endif // WEBP_MUX_MUXI_H_ diff --git a/thirdparty/libwebp/src/mux/muxinternal.c b/thirdparty/libwebp/src/mux/muxinternal.c index 1473f100e5..b9ee6717d3 100644 --- a/thirdparty/libwebp/src/mux/muxinternal.c +++ b/thirdparty/libwebp/src/mux/muxinternal.c @@ -111,27 +111,6 @@ WebPChunk* ChunkSearchList(WebPChunk* first, uint32_t nth, uint32_t tag) { return ((nth > 0) && (iter > 0)) ? NULL : first; } -// Outputs a pointer to 'prev_chunk->next_', -// where 'prev_chunk' is the pointer to the chunk at position (nth - 1). -// Returns true if nth chunk was found. -static int ChunkSearchListToSet(WebPChunk** chunk_list, uint32_t nth, - WebPChunk*** const location) { - uint32_t count = 0; - assert(chunk_list != NULL); - *location = chunk_list; - - while (*chunk_list != NULL) { - WebPChunk* const cur_chunk = *chunk_list; - ++count; - if (count == nth) return 1; // Found. - chunk_list = &cur_chunk->next_; - *location = chunk_list; - } - - // *chunk_list is ok to be NULL if adding at last location. - return (nth == 0 || (count == nth - 1)) ? 1 : 0; -} - //------------------------------------------------------------------------------ // Chunk writer methods. @@ -156,11 +135,12 @@ WebPMuxError ChunkAssignData(WebPChunk* chunk, const WebPData* const data, return WEBP_MUX_OK; } -WebPMuxError ChunkSetNth(WebPChunk* chunk, WebPChunk** chunk_list, - uint32_t nth) { +WebPMuxError ChunkSetHead(WebPChunk* const chunk, + WebPChunk** const chunk_list) { WebPChunk* new_chunk; - if (!ChunkSearchListToSet(chunk_list, nth, &chunk_list)) { + assert(chunk_list != NULL); + if (*chunk_list != NULL) { return WEBP_MUX_NOT_FOUND; } @@ -168,11 +148,26 @@ WebPMuxError ChunkSetNth(WebPChunk* chunk, WebPChunk** chunk_list, if (new_chunk == NULL) return WEBP_MUX_MEMORY_ERROR; *new_chunk = *chunk; chunk->owner_ = 0; - new_chunk->next_ = *chunk_list; + new_chunk->next_ = NULL; *chunk_list = new_chunk; return WEBP_MUX_OK; } +WebPMuxError ChunkAppend(WebPChunk* const chunk, + WebPChunk*** const chunk_list) { + assert(chunk_list != NULL && *chunk_list != NULL); + + if (**chunk_list == NULL) { + ChunkSetHead(chunk, *chunk_list); + } else { + WebPChunk* last_chunk = **chunk_list; + while (last_chunk->next_ != NULL) last_chunk = last_chunk->next_; + ChunkSetHead(chunk, &last_chunk->next_); + *chunk_list = &last_chunk->next_; + } + return WEBP_MUX_OK; +} + //------------------------------------------------------------------------------ // Chunk deletion method(s). @@ -232,9 +227,11 @@ void MuxImageInit(WebPMuxImage* const wpi) { WebPMuxImage* MuxImageRelease(WebPMuxImage* const wpi) { WebPMuxImage* next; if (wpi == NULL) return NULL; - ChunkDelete(wpi->header_); - ChunkDelete(wpi->alpha_); - ChunkDelete(wpi->img_); + // There should be at most one chunk of header_, alpha_, img_ but we call + // ChunkListDelete to be safe + ChunkListDelete(&wpi->header_); + ChunkListDelete(&wpi->alpha_); + ChunkListDelete(&wpi->img_); ChunkListDelete(&wpi->unknown_); next = wpi->next_; diff --git a/thirdparty/libwebp/src/mux/muxread.c b/thirdparty/libwebp/src/mux/muxread.c index 0b55286862..268f6acb53 100644 --- a/thirdparty/libwebp/src/mux/muxread.c +++ b/thirdparty/libwebp/src/mux/muxread.c @@ -59,6 +59,7 @@ static WebPMuxError ChunkVerifyAndAssign(WebPChunk* chunk, // Sanity checks. if (data_size < CHUNK_HEADER_SIZE) return WEBP_MUX_NOT_ENOUGH_DATA; chunk_size = GetLE32(data + TAG_SIZE); + if (chunk_size > MAX_CHUNK_PAYLOAD) return WEBP_MUX_BAD_DATA; { const size_t chunk_disk_size = SizeWithPadding(chunk_size); @@ -102,6 +103,7 @@ static int MuxImageParse(const WebPChunk* const chunk, int copy_data, const uint8_t* const last = bytes + size; WebPChunk subchunk; size_t subchunk_size; + WebPChunk** unknown_chunk_list = &wpi->unknown_; ChunkInit(&subchunk); assert(chunk->tag_ == kChunks[IDX_ANMF].tag); @@ -116,7 +118,7 @@ static int MuxImageParse(const WebPChunk* const chunk, int copy_data, if (size < hdr_size) goto Fail; ChunkAssignData(&subchunk, &temp, copy_data, chunk->tag_); } - ChunkSetNth(&subchunk, &wpi->header_, 1); + ChunkSetHead(&subchunk, &wpi->header_); wpi->is_partial_ = 1; // Waiting for ALPH and/or VP8/VP8L chunks. // Rest of the chunks. @@ -133,18 +135,23 @@ static int MuxImageParse(const WebPChunk* const chunk, int copy_data, switch (ChunkGetIdFromTag(subchunk.tag_)) { case WEBP_CHUNK_ALPHA: if (wpi->alpha_ != NULL) goto Fail; // Consecutive ALPH chunks. - if (ChunkSetNth(&subchunk, &wpi->alpha_, 1) != WEBP_MUX_OK) goto Fail; + if (ChunkSetHead(&subchunk, &wpi->alpha_) != WEBP_MUX_OK) goto Fail; wpi->is_partial_ = 1; // Waiting for a VP8 chunk. break; case WEBP_CHUNK_IMAGE: - if (ChunkSetNth(&subchunk, &wpi->img_, 1) != WEBP_MUX_OK) goto Fail; + if (wpi->img_ != NULL) goto Fail; // Only 1 image chunk allowed. + if (ChunkSetHead(&subchunk, &wpi->img_) != WEBP_MUX_OK) goto Fail; if (!MuxImageFinalize(wpi)) goto Fail; wpi->is_partial_ = 0; // wpi is completely filled. break; case WEBP_CHUNK_UNKNOWN: - if (wpi->is_partial_) goto Fail; // Encountered an unknown chunk - // before some image chunks. - if (ChunkSetNth(&subchunk, &wpi->unknown_, 0) != WEBP_MUX_OK) goto Fail; + if (wpi->is_partial_) { + goto Fail; // Encountered an unknown chunk + // before some image chunks. + } + if (ChunkAppend(&subchunk, &unknown_chunk_list) != WEBP_MUX_OK) { + goto Fail; + } break; default: goto Fail; @@ -175,6 +182,9 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, const uint8_t* data; size_t size; WebPChunk chunk; + // Stores the end of the chunk lists so that it is faster to append data to + // their ends. + WebPChunk** chunk_list_ends[WEBP_CHUNK_NIL + 1] = { NULL }; ChunkInit(&chunk); // Sanity checks. @@ -187,7 +197,7 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, size = bitstream->size; if (data == NULL) return NULL; - if (size < RIFF_HEADER_SIZE) return NULL; + if (size < RIFF_HEADER_SIZE + CHUNK_HEADER_SIZE) return NULL; if (GetLE32(data + 0) != MKFOURCC('R', 'I', 'F', 'F') || GetLE32(data + CHUNK_HEADER_SIZE) != MKFOURCC('W', 'E', 'B', 'P')) { return NULL; @@ -196,8 +206,6 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, mux = WebPMuxNew(); if (mux == NULL) return NULL; - if (size < RIFF_HEADER_SIZE + TAG_SIZE) goto Err; - tag = GetLE32(data + RIFF_HEADER_SIZE); if (tag != kChunks[IDX_VP8].tag && tag != kChunks[IDX_VP8L].tag && @@ -205,13 +213,17 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, goto Err; // First chunk should be VP8, VP8L or VP8X. } - riff_size = SizeWithPadding(GetLE32(data + TAG_SIZE)); - if (riff_size > MAX_CHUNK_PAYLOAD || riff_size > size) { - goto Err; - } else { - if (riff_size < size) { // Redundant data after last chunk. - size = riff_size; // To make sure we don't read any data beyond mux_size. - } + riff_size = GetLE32(data + TAG_SIZE); + if (riff_size > MAX_CHUNK_PAYLOAD) goto Err; + + // Note this padding is historical and differs from demux.c which does not + // pad the file size. + riff_size = SizeWithPadding(riff_size); + if (riff_size < CHUNK_HEADER_SIZE) goto Err; + if (riff_size > size) goto Err; + // There's no point in reading past the end of the RIFF chunk. + if (size > riff_size + CHUNK_HEADER_SIZE) { + size = riff_size + CHUNK_HEADER_SIZE; } end = data + size; @@ -226,7 +238,6 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, while (data != end) { size_t data_size; WebPChunkId id; - WebPChunk** chunk_list; if (ChunkVerifyAndAssign(&chunk, data, size, riff_size, copy_data) != WEBP_MUX_OK) { goto Err; @@ -236,11 +247,11 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, switch (id) { case WEBP_CHUNK_ALPHA: if (wpi->alpha_ != NULL) goto Err; // Consecutive ALPH chunks. - if (ChunkSetNth(&chunk, &wpi->alpha_, 1) != WEBP_MUX_OK) goto Err; + if (ChunkSetHead(&chunk, &wpi->alpha_) != WEBP_MUX_OK) goto Err; wpi->is_partial_ = 1; // Waiting for a VP8 chunk. break; case WEBP_CHUNK_IMAGE: - if (ChunkSetNth(&chunk, &wpi->img_, 1) != WEBP_MUX_OK) goto Err; + if (ChunkSetHead(&chunk, &wpi->img_) != WEBP_MUX_OK) goto Err; if (!MuxImageFinalize(wpi)) goto Err; wpi->is_partial_ = 0; // wpi is completely filled. PushImage: @@ -257,9 +268,13 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data, default: // A non-image chunk. if (wpi->is_partial_) goto Err; // Encountered a non-image chunk before // getting all chunks of an image. - chunk_list = MuxGetChunkListFromId(mux, id); // List to add this chunk. - if (ChunkSetNth(&chunk, chunk_list, 0) != WEBP_MUX_OK) goto Err; + if (chunk_list_ends[id] == NULL) { + chunk_list_ends[id] = + MuxGetChunkListFromId(mux, id); // List to add this chunk. + } + if (ChunkAppend(&chunk, &chunk_list_ends[id]) != WEBP_MUX_OK) goto Err; if (id == WEBP_CHUNK_VP8X) { // grab global specs + if (data_size < CHUNK_HEADER_SIZE + VP8X_CHUNK_SIZE) goto Err; mux->canvas_width_ = GetLE24(data + 12) + 1; mux->canvas_height_ = GetLE24(data + 15) + 1; } @@ -385,6 +400,10 @@ static WebPMuxError SynthesizeBitstream(const WebPMuxImage* const wpi, uint8_t* const data = (uint8_t*)WebPSafeMalloc(1ULL, size); if (data == NULL) return WEBP_MUX_MEMORY_ERROR; + // There should be at most one alpha_ chunk and exactly one img_ chunk. + assert(wpi->alpha_ == NULL || wpi->alpha_->next_ == NULL); + assert(wpi->img_ != NULL && wpi->img_->next_ == NULL); + // Main RIFF header. dst = MuxEmitRiffHeader(data, size); diff --git a/thirdparty/libwebp/src/utils/bit_reader_inl_utils.h b/thirdparty/libwebp/src/utils/bit_reader_inl_utils.h index 2ccc6ed326..7e607f370a 100644 --- a/thirdparty/libwebp/src/utils/bit_reader_inl_utils.h +++ b/thirdparty/libwebp/src/utils/bit_reader_inl_utils.h @@ -187,4 +187,4 @@ static WEBP_INLINE int VP8GetBitAlt(VP8BitReader* const br, int prob) { } // extern "C" #endif -#endif // WEBP_UTILS_BIT_READER_INL_UTILS_H_ +#endif // WEBP_UTILS_BIT_READER_INL_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/bit_reader_utils.h b/thirdparty/libwebp/src/utils/bit_reader_utils.h index 04f9804409..de810d402a 100644 --- a/thirdparty/libwebp/src/utils/bit_reader_utils.h +++ b/thirdparty/libwebp/src/utils/bit_reader_utils.h @@ -182,4 +182,4 @@ static WEBP_INLINE void VP8LFillBitWindow(VP8LBitReader* const br) { } // extern "C" #endif -#endif /* WEBP_UTILS_BIT_READER_UTILS_H_ */ +#endif // WEBP_UTILS_BIT_READER_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/bit_writer_utils.h b/thirdparty/libwebp/src/utils/bit_writer_utils.h index 2cf5976fe3..b9d5102a5a 100644 --- a/thirdparty/libwebp/src/utils/bit_writer_utils.h +++ b/thirdparty/libwebp/src/utils/bit_writer_utils.h @@ -151,4 +151,4 @@ static WEBP_INLINE void VP8LPutBits(VP8LBitWriter* const bw, } // extern "C" #endif -#endif /* WEBP_UTILS_BIT_WRITER_UTILS_H_ */ +#endif // WEBP_UTILS_BIT_WRITER_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/filters_utils.h b/thirdparty/libwebp/src/utils/filters_utils.h index 410f2fcdf2..61da66e212 100644 --- a/thirdparty/libwebp/src/utils/filters_utils.h +++ b/thirdparty/libwebp/src/utils/filters_utils.h @@ -29,4 +29,4 @@ WEBP_FILTER_TYPE WebPEstimateBestFilter(const uint8_t* data, } // extern "C" #endif -#endif /* WEBP_UTILS_FILTERS_UTILS_H_ */ +#endif // WEBP_UTILS_FILTERS_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/quant_levels_dec_utils.c b/thirdparty/libwebp/src/utils/quant_levels_dec_utils.c index 3818a78b93..f65b6cdbb6 100644 --- a/thirdparty/libwebp/src/utils/quant_levels_dec_utils.c +++ b/thirdparty/libwebp/src/utils/quant_levels_dec_utils.c @@ -261,9 +261,15 @@ static void CleanupParams(SmoothParams* const p) { int WebPDequantizeLevels(uint8_t* const data, int width, int height, int stride, int strength) { - const int radius = 4 * strength / 100; + int radius = 4 * strength / 100; + if (strength < 0 || strength > 100) return 0; if (data == NULL || width <= 0 || height <= 0) return 0; // bad params + + // limit the filter size to not exceed the image dimensions + if (2 * radius + 1 > width) radius = (width - 1) >> 1; + if (2 * radius + 1 > height) radius = (height - 1) >> 1; + if (radius > 0) { SmoothParams p; memset(&p, 0, sizeof(p)); diff --git a/thirdparty/libwebp/src/utils/quant_levels_dec_utils.h b/thirdparty/libwebp/src/utils/quant_levels_dec_utils.h index f822107a72..327f19f336 100644 --- a/thirdparty/libwebp/src/utils/quant_levels_dec_utils.h +++ b/thirdparty/libwebp/src/utils/quant_levels_dec_utils.h @@ -32,4 +32,4 @@ int WebPDequantizeLevels(uint8_t* const data, int width, int height, int stride, } // extern "C" #endif -#endif /* WEBP_UTILS_QUANT_LEVELS_DEC_UTILS_H_ */ +#endif // WEBP_UTILS_QUANT_LEVELS_DEC_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/quant_levels_utils.h b/thirdparty/libwebp/src/utils/quant_levels_utils.h index 75df2ba6a4..9ee3ea0075 100644 --- a/thirdparty/libwebp/src/utils/quant_levels_utils.h +++ b/thirdparty/libwebp/src/utils/quant_levels_utils.h @@ -33,4 +33,4 @@ int QuantizeLevels(uint8_t* const data, int width, int height, int num_levels, } // extern "C" #endif -#endif /* WEBP_UTILS_QUANT_LEVELS_UTILS_H_ */ +#endif // WEBP_UTILS_QUANT_LEVELS_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/random_utils.h b/thirdparty/libwebp/src/utils/random_utils.h index 6d36c667e7..a5006f84f7 100644 --- a/thirdparty/libwebp/src/utils/random_utils.h +++ b/thirdparty/libwebp/src/utils/random_utils.h @@ -60,4 +60,4 @@ static WEBP_INLINE int VP8RandomBits(VP8Random* const rg, int num_bits) { } // extern "C" #endif -#endif /* WEBP_UTILS_RANDOM_UTILS_H_ */ +#endif // WEBP_UTILS_RANDOM_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/rescaler_utils.h b/thirdparty/libwebp/src/utils/rescaler_utils.h index 8890e6fa13..ca41e42c4a 100644 --- a/thirdparty/libwebp/src/utils/rescaler_utils.h +++ b/thirdparty/libwebp/src/utils/rescaler_utils.h @@ -98,4 +98,4 @@ int WebPRescalerHasPendingOutput(const WebPRescaler* const rescaler) { } // extern "C" #endif -#endif /* WEBP_UTILS_RESCALER_UTILS_H_ */ +#endif // WEBP_UTILS_RESCALER_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/thread_utils.h b/thirdparty/libwebp/src/utils/thread_utils.h index c8ae6c9033..29ad49f74b 100644 --- a/thirdparty/libwebp/src/utils/thread_utils.h +++ b/thirdparty/libwebp/src/utils/thread_utils.h @@ -87,4 +87,4 @@ WEBP_EXTERN const WebPWorkerInterface* WebPGetWorkerInterface(void); } // extern "C" #endif -#endif /* WEBP_UTILS_THREAD_UTILS_H_ */ +#endif // WEBP_UTILS_THREAD_UTILS_H_ diff --git a/thirdparty/libwebp/src/utils/utils.h b/thirdparty/libwebp/src/utils/utils.h index 52921bf24e..da97b5d38f 100644 --- a/thirdparty/libwebp/src/utils/utils.h +++ b/thirdparty/libwebp/src/utils/utils.h @@ -175,4 +175,4 @@ WEBP_EXTERN int WebPGetColorPalette(const struct WebPPicture* const pic, } // extern "C" #endif -#endif /* WEBP_UTILS_UTILS_H_ */ +#endif // WEBP_UTILS_UTILS_H_ diff --git a/thirdparty/libwebp/src/webp/decode.h b/thirdparty/libwebp/src/webp/decode.h index 2165e96c95..95d31e7619 100644 --- a/thirdparty/libwebp/src/webp/decode.h +++ b/thirdparty/libwebp/src/webp/decode.h @@ -491,4 +491,4 @@ WEBP_EXTERN VP8StatusCode WebPDecode(const uint8_t* data, size_t data_size, } // extern "C" #endif -#endif /* WEBP_WEBP_DECODE_H_ */ +#endif // WEBP_WEBP_DECODE_H_ diff --git a/thirdparty/libwebp/src/webp/demux.h b/thirdparty/libwebp/src/webp/demux.h index 555d641338..846eeb15a9 100644 --- a/thirdparty/libwebp/src/webp/demux.h +++ b/thirdparty/libwebp/src/webp/demux.h @@ -360,4 +360,4 @@ WEBP_EXTERN void WebPAnimDecoderDelete(WebPAnimDecoder* dec); } // extern "C" #endif -#endif /* WEBP_WEBP_DEMUX_H_ */ +#endif // WEBP_WEBP_DEMUX_H_ diff --git a/thirdparty/libwebp/src/webp/encode.h b/thirdparty/libwebp/src/webp/encode.h index 7ec3543dc2..549cf07730 100644 --- a/thirdparty/libwebp/src/webp/encode.h +++ b/thirdparty/libwebp/src/webp/encode.h @@ -542,4 +542,4 @@ WEBP_EXTERN int WebPEncode(const WebPConfig* config, WebPPicture* picture); } // extern "C" #endif -#endif /* WEBP_WEBP_ENCODE_H_ */ +#endif // WEBP_WEBP_ENCODE_H_ diff --git a/thirdparty/libwebp/src/webp/format_constants.h b/thirdparty/libwebp/src/webp/format_constants.h index 329fc8a3b0..eca6981a47 100644 --- a/thirdparty/libwebp/src/webp/format_constants.h +++ b/thirdparty/libwebp/src/webp/format_constants.h @@ -84,4 +84,4 @@ typedef enum { // overflow a uint32_t. #define MAX_CHUNK_PAYLOAD (~0U - CHUNK_HEADER_SIZE - 1) -#endif /* WEBP_WEBP_FORMAT_CONSTANTS_H_ */ +#endif // WEBP_WEBP_FORMAT_CONSTANTS_H_ diff --git a/thirdparty/libwebp/src/webp/mux.h b/thirdparty/libwebp/src/webp/mux.h index 28bb4a41c9..66096a92e0 100644 --- a/thirdparty/libwebp/src/webp/mux.h +++ b/thirdparty/libwebp/src/webp/mux.h @@ -527,4 +527,4 @@ WEBP_EXTERN void WebPAnimEncoderDelete(WebPAnimEncoder* enc); } // extern "C" #endif -#endif /* WEBP_WEBP_MUX_H_ */ +#endif // WEBP_WEBP_MUX_H_ diff --git a/thirdparty/libwebp/src/webp/mux_types.h b/thirdparty/libwebp/src/webp/mux_types.h index b37e2c67aa..ceea77dfc6 100644 --- a/thirdparty/libwebp/src/webp/mux_types.h +++ b/thirdparty/libwebp/src/webp/mux_types.h @@ -95,4 +95,4 @@ static WEBP_INLINE int WebPDataCopy(const WebPData* src, WebPData* dst) { } // extern "C" #endif -#endif /* WEBP_WEBP_MUX_TYPES_H_ */ +#endif // WEBP_WEBP_MUX_TYPES_H_ diff --git a/thirdparty/libwebp/src/webp/types.h b/thirdparty/libwebp/src/webp/types.h index 989a763f0d..0ce2622e41 100644 --- a/thirdparty/libwebp/src/webp/types.h +++ b/thirdparty/libwebp/src/webp/types.h @@ -49,4 +49,4 @@ typedef long long int int64_t; // Macro to check ABI compatibility (same major revision number) #define WEBP_ABI_IS_INCOMPATIBLE(a, b) (((a) >> 8) != ((b) >> 8)) -#endif /* WEBP_WEBP_TYPES_H_ */ +#endif // WEBP_WEBP_TYPES_H_ diff --git a/thirdparty/libwebsockets/core/context.c b/thirdparty/libwebsockets/core/context.c index db9151b95f..7be004df33 100644 --- a/thirdparty/libwebsockets/core/context.c +++ b/thirdparty/libwebsockets/core/context.c @@ -134,7 +134,7 @@ lws_protocol_vh_priv_get(struct lws_vhost *vhost, { int n = 0; - if (!vhost || !vhost->protocol_vh_privs) + if (!vhost || !vhost->protocol_vh_privs || !prot) return NULL; while (n < vhost->count_protocols && &vhost->protocols[n] != prot) @@ -808,7 +808,7 @@ lws_create_vhost(struct lws_context *context, #ifdef LWS_WITH_ACCESS_LOG if (info->log_filepath) { - vh->log_fd = open(info->log_filepath, + vh->log_fd = lws_open(info->log_filepath, O_CREAT | O_APPEND | O_RDWR, 0600); if (vh->log_fd == (int)LWS_INVALID_FILE) { lwsl_err("unable to open log filepath %s\n", @@ -936,24 +936,29 @@ lws_create_event_pipes(struct lws_context *context) wsi->tsi = n; wsi->vhost = NULL; wsi->event_pipe = 1; + wsi->desc.sockfd = LWS_SOCK_INVALID; + context->pt[n].pipe_wsi = wsi; + context->count_wsi_allocated++; - if (lws_plat_pipe_create(wsi)) { - lws_free(wsi); + if (lws_plat_pipe_create(wsi)) + /* + * platform code returns 0 if it actually created pipes + * and initialized pt->dummy_pipe_fds[]. If it used + * some other mechanism outside of signaling in the + * normal event loop, we skip treating the pipe as + * related to dummy_pipe_fds[], adding it to the fds, + * etc. + */ continue; - } + wsi->desc.sockfd = context->pt[n].dummy_pipe_fds[0]; lwsl_debug("event pipe fd %d\n", wsi->desc.sockfd); - context->pt[n].pipe_wsi = wsi; - if (context->event_loop_ops->accept) context->event_loop_ops->accept(wsi); if (__insert_wsi_socket_into_fds(context, wsi)) return 1; - - //lws_change_pollfd(context->pt[n].pipe_wsi, 0, LWS_POLLIN); - context->count_wsi_allocated++; } return 0; diff --git a/thirdparty/libwebsockets/core/libwebsockets.c b/thirdparty/libwebsockets/core/libwebsockets.c index 0da02b17e4..58f00226f6 100644 --- a/thirdparty/libwebsockets/core/libwebsockets.c +++ b/thirdparty/libwebsockets/core/libwebsockets.c @@ -28,7 +28,7 @@ #ifdef LWS_WITH_IPV6 #if defined(WIN32) || defined(_WIN32) #include <wincrypt.h> -#include <Iphlpapi.h> +#include <iphlpapi.h> #else #include <net/if.h> #endif @@ -58,6 +58,28 @@ static const char * const log_level_names[] = { }; #endif +int lws_open(const char *__file, int __oflag, ...) +{ + va_list ap; + int n; + + va_start(ap, __oflag); + if (((__oflag & O_CREAT) == O_CREAT) +#if defined(O_TMPFILE) + || ((__oflag & O_TMPFILE) == O_TMPFILE) +#endif + ) + /* last arg is really a mode_t. But windows... */ + n = open(__file, __oflag, va_arg(ap, uint32_t)); + else + n = open(__file, __oflag); + va_end(ap); + + lws_plat_apply_FD_CLOEXEC(n); + + return n; +} + #if defined (_DEBUG) void lwsi_set_role(struct lws *wsi, lws_wsi_state_t role) { @@ -826,7 +848,15 @@ just_kill_connection: if (!wsi->protocol) pro = &wsi->vhost->protocols[0]; - pro->callback(wsi, + if (!wsi->upgraded_to_http2 || !lwsi_role_client(wsi)) + /* + * The network wsi for a client h2 connection shouldn't + * call back for its role: the child stream connections + * own the role. Otherwise h2 will call back closed + * one too many times as the children do it and then + * the closing network stream. + */ + pro->callback(wsi, wsi->role_ops->close_cb[lwsi_role_server(wsi)], wsi->user_space, NULL, 0); wsi->told_user_closed = 1; @@ -1453,7 +1483,7 @@ lws_vfs_select_fops(const struct lws_plat_file_ops *fops, const char *vfs_path, pf = fops->next; while (pf) { n = 0; - while (n < (int)ARRAY_SIZE(pf->fi) && pf->fi[n].sig) { + while (n < (int)LWS_ARRAY_SIZE(pf->fi) && pf->fi[n].sig) { if (p >= vfs_path + pf->fi[n].len) if (!strncmp(p - (pf->fi[n].len - 1), pf->fi[n].sig, @@ -1935,9 +1965,9 @@ static const char * const colours[] = { "[32;1m", /* LLL_INFO */ "[34;1m", /* LLL_DEBUG */ "[33;1m", /* LLL_PARSER */ - "[33;1m", /* LLL_HEADER */ - "[33;1m", /* LLL_EXT */ - "[33;1m", /* LLL_CLIENT */ + "[33m", /* LLL_HEADER */ + "[33m", /* LLL_EXT */ + "[33m", /* LLL_CLIENT */ "[33;1m", /* LLL_LATENCY */ "[30;1m", /* LLL_USER */ }; @@ -1946,14 +1976,14 @@ LWS_VISIBLE void lwsl_emit_stderr(int level, const char *line) { char buf[50]; static char tty = 3; - int n, m = ARRAY_SIZE(colours) - 1; + int n, m = LWS_ARRAY_SIZE(colours) - 1; if (!tty) tty = isatty(2) | 2; lwsl_timestamp(level, buf, sizeof(buf)); if (tty == 3) { - n = 1 << (ARRAY_SIZE(colours) - 1); + n = 1 << (LWS_ARRAY_SIZE(colours) - 1); while (n) { if (level & n) break; @@ -2060,7 +2090,9 @@ lwsl_hexdump_level(int hexdump_level, const void *vbuf, size_t len) LWS_VISIBLE void lwsl_hexdump(const void *vbuf, size_t len) { +#if defined(_DEBUG) lwsl_hexdump_level(LLL_DEBUG, vbuf, len); +#endif } LWS_VISIBLE int @@ -2091,6 +2123,8 @@ lws_partial_buffered(struct lws *wsi) LWS_VISIBLE lws_fileofs_t lws_get_peer_write_allowance(struct lws *wsi) { + if (!wsi->role_ops->tx_credit) + return -1; return wsi->role_ops->tx_credit(wsi); } @@ -3364,7 +3398,7 @@ lws_stats_log_dump(struct lws_context *context) wl = pt->http.ah_wait_list; while (wl) { m++; - wl = wl->ah_wait_list; + wl = wl->http.ah_wait_list; } lwsl_notice(" AH wait list count / actual: %d / %d\n", @@ -3401,7 +3435,8 @@ lws_stats_log_dump(struct lws_context *context) strcpy(buf, "unknown"); #if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2) lwsl_notice(" peer %s: count wsi: %d, count ah: %d\n", - buf, df->count_wsi, df->count_ah); + buf, df->count_wsi, + df->http.count_ah); #else lwsl_notice(" peer %s: count wsi: %d\n", buf, df->count_wsi); diff --git a/thirdparty/libwebsockets/core/output.c b/thirdparty/libwebsockets/core/output.c index e2ff18ef27..11965a06b9 100644 --- a/thirdparty/libwebsockets/core/output.c +++ b/thirdparty/libwebsockets/core/output.c @@ -126,6 +126,18 @@ int lws_issue_raw(struct lws *wsi, unsigned char *buf, size_t len) lwsl_info("** %p signalling to close now\n", wsi); return -1; /* retry closing now */ } + +#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2) +#if !defined(LWS_WITHOUT_SERVER) + if (wsi->http.deferred_transaction_completed) { + lwsl_notice("%s: partial completed, doing " + "deferred transaction completed\n", + __func__); + wsi->http.deferred_transaction_completed = 0; + return lws_http_transaction_completed(wsi); + } +#endif +#endif } /* always callback on writeable */ lws_callback_on_writable(wsi); diff --git a/thirdparty/libwebsockets/core/private.h b/thirdparty/libwebsockets/core/private.h index d6b494ac8c..73748b0498 100644 --- a/thirdparty/libwebsockets/core/private.h +++ b/thirdparty/libwebsockets/core/private.h @@ -232,7 +232,7 @@ #endif #else /* not windows */ - static inline int compatible_close(int fd) { return close(fd); } + static LWS_INLINE int compatible_close(int fd) { return close(fd); } #include <sys/stat.h> #include <sys/time.h> @@ -351,7 +351,15 @@ extern "C" { #define LWS_H2_RX_SCRATCH_SIZE 512 - +#if defined(WIN32) || defined(_WIN32) + // Visual studio older than 2015 and WIN_CE has only _stricmp + #if (defined(_MSC_VER) && _MSC_VER < 1900) || defined(_WIN32_WCE) + #define strcasecmp _stricmp + #elif !defined(__MINGW32__) + #define strcasecmp stricmp + #endif + #define getdtablesize() 30000 +#endif /* * All lws_tls...() functions must return this type, converting the @@ -863,6 +871,7 @@ struct lws_context { unsigned int timeout_secs; unsigned int pt_serv_buf_size; int max_http_header_data; + int max_http_header_pool; int simultaneous_ssl_restriction; int simultaneous_ssl; #if defined(LWS_WITH_PEER_LIMITS) @@ -889,7 +898,6 @@ struct lws_context { volatile int service_tid; int service_tid_detected; - short max_http_header_pool; short count_threads; short plugin_protocol_count; short plugin_extension_count; @@ -1216,7 +1224,7 @@ LWS_EXTERN int lws_rxflow_cache(struct lws *wsi, unsigned char *buf, int n, int len); #ifndef LWS_LATENCY -static inline void +static LWS_INLINE void lws_latency(struct lws_context *context, struct lws *wsi, const char *action, int ret, int completion) { do { @@ -1224,7 +1232,7 @@ lws_latency(struct lws_context *context, struct lws *wsi, const char *action, (void)completion; } while (0); } -static inline void +static LWS_INLINE void lws_latency_pre(struct lws_context *context, struct lws *wsi) { do { (void)context; (void)wsi; } while (0); } @@ -1597,6 +1605,9 @@ void lws_free(void *p); #define lws_free_set_NULL(P) do { lws_realloc(P, 0, "free"); (P) = NULL; } while(0) #endif +char * +lws_strdup(const char *s); + int lws_plat_pipe_create(struct lws *wsi); int @@ -1606,6 +1617,9 @@ lws_plat_pipe_close(struct lws *wsi); int lws_create_event_pipes(struct lws_context *context); +int lws_open(const char *__file, int __oflag, ...); +void lws_plat_apply_FD_CLOEXEC(int n); + const struct lws_plat_file_ops * lws_vfs_select_fops(const struct lws_plat_file_ops *fops, const char *vfs_path, const char **vpath); @@ -1673,10 +1687,10 @@ lws_broadcast(struct lws_context *context, int reason, void *in, size_t len); lws_stats_atomic_max(struct lws_context * context, struct lws_context_per_thread *pt, int index, uint64_t val); #else - static inline uint64_t lws_stats_atomic_bump(struct lws_context * context, + static LWS_INLINE uint64_t lws_stats_atomic_bump(struct lws_context * context, struct lws_context_per_thread *pt, int index, uint64_t bump) { (void)context; (void)pt; (void)index; (void)bump; return 0; } - static inline uint64_t lws_stats_atomic_max(struct lws_context * context, + static LWS_INLINE uint64_t lws_stats_atomic_max(struct lws_context * context, struct lws_context_per_thread *pt, int index, uint64_t val) { (void)context; (void)pt; (void)index; (void)val; return 0; } #endif @@ -1703,6 +1717,11 @@ void lws_peer_dump_from_wsi(struct lws *wsi); #endif +#ifdef LWS_WITH_HTTP_PROXY +hubbub_error +html_parser_cb(const hubbub_token *token, void *pw); +#endif + void __lws_remove_from_timeout_list(struct lws *wsi); diff --git a/thirdparty/libwebsockets/libwebsockets.h b/thirdparty/libwebsockets/libwebsockets.h index 7ae563d582..2c01696404 100644 --- a/thirdparty/libwebsockets/libwebsockets.h +++ b/thirdparty/libwebsockets/libwebsockets.h @@ -66,14 +66,6 @@ typedef unsigned long long lws_intptr_t; #define O_RDONLY _O_RDONLY #endif -// Visual studio older than 2015 and WIN_CE has only _stricmp -#if (defined(_MSC_VER) && _MSC_VER < 1900) || defined(_WIN32_WCE) -#define strcasecmp _stricmp -#elif !defined(__MINGW32__) -#define strcasecmp stricmp -#endif -#define getdtablesize() 30000 - #define LWS_INLINE __inline #define LWS_VISIBLE #define LWS_WARN_UNUSED_RESULT @@ -150,6 +142,7 @@ typedef unsigned long long lws_intptr_t; #endif #if defined(__ANDROID__) +#include <netinet/in.h> #include <unistd.h> #define getdtablesize() sysconf(_SC_OPEN_MAX) #endif @@ -164,6 +157,9 @@ typedef unsigned long long lws_intptr_t; #ifdef LWS_HAVE_UV_VERSION_H #include <uv-version.h> #endif +#ifdef LWS_HAVE_NEW_UV_VERSION_H +#include <uv/version.h> +#endif #endif /* LWS_WITH_LIBUV */ #if defined(LWS_WITH_LIBEVENT) #include <event2/event.h> @@ -456,9 +452,6 @@ lwsl_visible(int level); #endif struct lws; -#ifndef ARRAY_SIZE -#define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0])) -#endif typedef int64_t lws_usec_t; @@ -540,7 +533,7 @@ struct timer_mapping { #define lws_uv_getloop(a, b) (NULL) -static inline void uv_timer_init(void *l, uv_timer_t *t) +static LWS_INLINE void uv_timer_init(void *l, uv_timer_t *t) { (void)l; *t = NULL; @@ -548,7 +541,7 @@ static inline void uv_timer_init(void *l, uv_timer_t *t) extern void esp32_uvtimer_cb(TimerHandle_t t); -static inline void uv_timer_start(uv_timer_t *t, uv_cb_t *cb, int first, int rep) +static LWS_INLINE void uv_timer_start(uv_timer_t *t, uv_cb_t *cb, int first, int rep) { struct timer_mapping *tm = (struct timer_mapping *)malloc(sizeof(*tm)); @@ -563,12 +556,12 @@ static inline void uv_timer_start(uv_timer_t *t, uv_cb_t *cb, int first, int rep xTimerStart(*t, 0); } -static inline void uv_timer_stop(uv_timer_t *t) +static LWS_INLINE void uv_timer_stop(uv_timer_t *t) { xTimerStop(*t, 0); } -static inline void uv_close(uv_handle_t *h, void *v) +static LWS_INLINE void uv_close(uv_handle_t *h, void *v) { free(pvTimerGetTimerID((uv_timer_t)h)); xTimerDelete(*(uv_timer_t *)h, 0); @@ -1757,7 +1750,7 @@ lws_genhash_destroy(struct lws_genhash_ctx *ctx, void *result); * If the return is nonzero, it failed and there is nothing needing to be * destroyed. */ -int +LWS_VISIBLE LWS_EXTERN int LWS_WARN_UNUSED_RESULT lws_genhmac_init(struct lws_genhmac_ctx *ctx, enum lws_genhmac_types type, const uint8_t *key, size_t key_len); @@ -1771,7 +1764,7 @@ lws_genhmac_init(struct lws_genhmac_ctx *ctx, enum lws_genhmac_types type, * * If the return is nonzero, it failed and needs destroying. */ -int +LWS_VISIBLE LWS_EXTERN int LWS_WARN_UNUSED_RESULT lws_genhmac_update(struct lws_genhmac_ctx *ctx, const void *in, size_t len); /** lws_genhmac_destroy() - copy out the result digest and destroy the ctx @@ -1785,7 +1778,7 @@ lws_genhmac_update(struct lws_genhmac_ctx *ctx, const void *in, size_t len); * NULL result is supported so that you can destroy the ctx cleanly on error * conditions, where there is no valid result. */ -int +LWS_VISIBLE LWS_EXTERN int lws_genhmac_destroy(struct lws_genhmac_ctx *ctx, void *result); ///@} @@ -2812,7 +2805,7 @@ struct lws_context_creation_info { /**< VHOST: pointer to optional linked list of per-vhost * options made accessible to protocols */ int keepalive_timeout; - /**< VHOST: (default = 0 = 60s) seconds to allow remote + /**< VHOST: (default = 0 = 5s) seconds to allow remote * client to hold on to an idle HTTP/1.1 connection */ const char *log_filepath; /**< VHOST: filepath to append logs to... this is opened before @@ -4943,7 +4936,7 @@ lws_write(struct lws *wsi, unsigned char *buf, size_t len, lws_write(wsi, (unsigned char *)(buf), len, LWS_WRITE_HTTP) /* helper for multi-frame ws message flags */ -static inline int +static LWS_INLINE int lws_write_ws_flags(int initial, int is_start, int is_end) { int r; @@ -5621,13 +5614,13 @@ struct lws_dll_lws { /* typed as struct lws * */ #define lws_dll_is_null(___dll) (!(___dll)->prev && !(___dll)->next) -static inline void +static LWS_INLINE void lws_dll_lws_add_front(struct lws_dll_lws *_a, struct lws_dll_lws *_head) { lws_dll_add_front((struct lws_dll *)_a, (struct lws_dll *)_head); } -static inline void +static LWS_INLINE void lws_dll_lws_remove(struct lws_dll_lws *_a) { lws_dll_remove((struct lws_dll *)_a); @@ -7064,9 +7057,6 @@ lws_email_destroy(struct lws_email *email); //@{ struct lejp_ctx; -#ifndef ARRAY_SIZE -#define ARRAY_SIZE(_x) (sizeof(_x) / sizeof(_x[0])) -#endif #define LWS_ARRAY_SIZE(_x) (sizeof(_x) / sizeof(_x[0])) #define LEJP_FLAG_WS_KEEP 64 #define LEJP_FLAG_WS_COMMENTLINE 32 @@ -7219,7 +7209,7 @@ typedef signed char (*lejp_callback)(struct lejp_ctx *ctx, char reason); #endif #ifndef LEJP_STRING_CHUNK /* must be >= 30 to assemble floats */ -#define LEJP_STRING_CHUNK 255 +#define LEJP_STRING_CHUNK 254 #endif enum num_flags { @@ -7253,7 +7243,7 @@ struct lejp_ctx { uint16_t i[LEJP_MAX_INDEX_DEPTH]; /* index array */ uint16_t wild[LEJP_MAX_INDEX_DEPTH]; /* index array */ char path[LEJP_MAX_PATH]; - char buf[LEJP_STRING_CHUNK]; + char buf[LEJP_STRING_CHUNK + 1]; /* int */ diff --git a/thirdparty/libwebsockets/lws_config.h b/thirdparty/libwebsockets/lws_config.h index e5e15cc2fd..86ce9ac38a 100644 --- a/thirdparty/libwebsockets/lws_config.h +++ b/thirdparty/libwebsockets/lws_config.h @@ -174,7 +174,7 @@ #define LWS_HAVE_MALLOC_H #endif -#if !defined(IPHONE_ENABLED) && !defined(OSX_ENABLED) && !defined(__HAIKU__) +#if !defined(__APPLE__) && !defined(__HAIKU__) #define LWS_HAVE_PIPE2 #endif diff --git a/thirdparty/libwebsockets/lws_config_private.h b/thirdparty/libwebsockets/lws_config_private.h index b26d225afa..e531777624 100644 --- a/thirdparty/libwebsockets/lws_config_private.h +++ b/thirdparty/libwebsockets/lws_config_private.h @@ -81,7 +81,7 @@ /* Define to 1 if you have the <sys/prctl.h> header file. */ #define LWS_HAVE_SYS_PRCTL_H -#if defined(OSX_ENABLED) || defined(IPHONE_ENABLED) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__HAIKU__) +#if defined(__APPLE__) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__HAIKU__) #undef LWS_HAVE_SYS_PRCTL_H #endif diff --git a/thirdparty/libwebsockets/misc/lejp.c b/thirdparty/libwebsockets/misc/lejp.c index 00d350f81e..99142b9553 100644 --- a/thirdparty/libwebsockets/misc/lejp.c +++ b/thirdparty/libwebsockets/misc/lejp.c @@ -20,6 +20,7 @@ */ #include <libwebsockets.h> +#include "core/private.h" #include <string.h> #include <stdio.h> @@ -30,7 +31,7 @@ * \param callback: your user callback which will received parsed tokens * \param user: optional user data pointer untouched by lejp * \param paths: your array of name elements you are interested in - * \param count_paths: ARRAY_SIZE() of @paths + * \param count_paths: LWS_ARRAY_SIZE() of @paths * * Prepares your context struct for use with lejp */ @@ -250,7 +251,7 @@ lejp_parse(struct lejp_ctx *ctx, const unsigned char *json, int len) case LEJP_MP_STRING: if (c == '\"') { - if (!ctx->sp) { + if (!ctx->sp) { /* JSON can't end on quote */ ret = LEJP_REJECT_MP_STRING_UNDERRUN; goto reject; } @@ -417,7 +418,7 @@ lejp_parse(struct lejp_ctx *ctx, const unsigned char *json, int len) goto reject; } ctx->i[ctx->ipos++] = 0; - if (ctx->ipos > ARRAY_SIZE(ctx->i)) { + if (ctx->ipos > LWS_ARRAY_SIZE(ctx->i)) { ret = LEJP_REJECT_MP_DELIM_ISTACK; goto reject; } @@ -425,17 +426,23 @@ lejp_parse(struct lejp_ctx *ctx, const unsigned char *json, int len) case ']': /* pop */ + if (!ctx->sp) { /* JSON can't end on ] */ + ret = LEJP_REJECT_MP_C_OR_E_UNDERF; + goto reject; + } ctx->sp--; if (ctx->st[ctx->sp].s != LEJP_MP_ARRAY_END) { ret = LEJP_REJECT_MP_C_OR_E_NOTARRAY; goto reject; } /* drop the path [n] bit */ - ctx->ppos = ctx->st[ctx->sp - 1].p; - ctx->ipos = ctx->st[ctx->sp - 1].i; + if (ctx->sp) { + ctx->ppos = ctx->st[ctx->sp - 1].p; + ctx->ipos = ctx->st[ctx->sp - 1].i; + } ctx->path[ctx->ppos] = '\0'; if (ctx->path_match && - ctx->ppos <= ctx->path_match_len) + ctx->ppos <= ctx->path_match_len) /* * we shrank the path to be * smaller than the matching point @@ -603,7 +610,7 @@ lejp_parse(struct lejp_ctx *ctx, const unsigned char *json, int len) break; } if (c == ']') { - if (!ctx->sp) { + if (!ctx->sp) { /* JSON can't end on ] */ ret = LEJP_REJECT_MP_C_OR_E_UNDERF; goto reject; } @@ -631,7 +638,7 @@ lejp_parse(struct lejp_ctx *ctx, const unsigned char *json, int len) goto redo_character; } if (c == '}') { - if (ctx->sp == 0) { + if (!ctx->sp) { lejp_check_path_match(ctx); if (ctx->callback(ctx, LEJPCB_OBJECT_END)) { ret = LEJP_REJECT_CALLBACK; @@ -716,7 +723,7 @@ add_stack_level: ctx->st[ctx->sp].p = ctx->ppos; ctx->st[ctx->sp].i = ctx->ipos; - if (++ctx->sp == ARRAY_SIZE(ctx->st)) { + if (++ctx->sp == LWS_ARRAY_SIZE(ctx->st)) { ret = LEJP_REJECT_STACK_OVERFLOW; goto reject; } diff --git a/thirdparty/libwebsockets/plat/lws-plat-unix.c b/thirdparty/libwebsockets/plat/lws-plat-unix.c index bacc6af647..7dba3bd82f 100644 --- a/thirdparty/libwebsockets/plat/lws-plat-unix.c +++ b/thirdparty/libwebsockets/plat/lws-plat-unix.c @@ -30,6 +30,12 @@ #endif #include <dirent.h> +void lws_plat_apply_FD_CLOEXEC(int n) +{ + if (n != -1) + fcntl(n, F_SETFD, FD_CLOEXEC ); +} + int lws_plat_socket_offset(void) { @@ -330,6 +336,8 @@ lws_plat_set_socket_options(struct lws_vhost *vhost, int fd) struct protoent *tcp_proto; #endif + fcntl(fd, F_SETFD, FD_CLOEXEC); + if (vhost->ka_time) { /* enable keepalive on this socket */ optval = 1; @@ -952,7 +960,7 @@ lws_plat_write_file(const char *filename, void *buf, int len) LWS_VISIBLE int lws_plat_read_file(const char *filename, void *buf, int len) { - int n, fd = open(filename, O_RDONLY); + int n, fd = lws_open(filename, O_RDONLY); if (fd == -1) return -1; diff --git a/thirdparty/libwebsockets/plat/lws-plat-win.c b/thirdparty/libwebsockets/plat/lws-plat-win.c index 8a43081ef1..1850b64250 100644 --- a/thirdparty/libwebsockets/plat/lws-plat-win.c +++ b/thirdparty/libwebsockets/plat/lws-plat-win.c @@ -3,6 +3,10 @@ #endif #include "core/private.h" +void lws_plat_apply_FD_CLOEXEC(int n) +{ +} + int lws_plat_socket_offset(void) { @@ -54,7 +58,7 @@ time_in_microseconds() memcpy(&datetime, &filetime, sizeof(datetime)); /* Windows file times are in 100s of nanoseconds. */ - return (datetime.QuadPart - DELTA_EPOCH_IN_MICROSECS) / 10; + return (datetime.QuadPart / 10) - DELTA_EPOCH_IN_MICROSECS; } #ifdef _WIN32_WCE @@ -229,23 +233,21 @@ _lws_plat_service_tsi(struct lws_context *context, int timeout_ms, int tsi) continue; wsi = wsi_from_fd(context, pfd->fd); - if (wsi->listener) + if (!wsi || wsi->listener) continue; - if (!wsi || wsi->sock_send_blocking) + if (wsi->sock_send_blocking) continue; pfd->revents = LWS_POLLOUT; n = lws_service_fd(context, pfd); if (n < 0) return -1; + + /* Force WSAWaitForMultipleEvents() to check events and then return immediately. */ + timeout_ms = 0; + /* if something closed, retry this slot */ if (n) i--; - - /* - * any wsi has truncated, force him signalled - */ - if (wsi->trunc_len) - WSASetEvent(pt->events[0]); } /* @@ -261,9 +263,11 @@ _lws_plat_service_tsi(struct lws_context *context, int timeout_ms, int tsi) } if (timeout_ms) { + lws_usec_t t; + lws_pt_lock(pt, __func__); /* don't stay in poll wait longer than next hr timeout */ - lws_usec_t t = __lws_hrtimer_service(pt); + t = __lws_hrtimer_service(pt); if ((lws_usec_t)timeout_ms * 1000 > t) timeout_ms = (int)(t / 1000); @@ -310,8 +314,10 @@ _lws_plat_service_tsi(struct lws_context *context, int timeout_ms, int tsi) if (pfd->revents & LWS_POLLHUP) --eIdx; - if (pfd->revents) + if (pfd->revents) { + recv(pfd->fd, NULL, 0, 0); lws_service_fd_tsi(context, pfd, tsi); + } } } @@ -637,7 +643,7 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, lws_fop_fd_t fop_fd; FILE_STANDARD_INFO fInfo = {0}; - MultiByteToWideChar(CP_UTF8, 0, filename, -1, buf, ARRAY_SIZE(buf)); + MultiByteToWideChar(CP_UTF8, 0, filename, -1, buf, LWS_ARRAY_SIZE(buf)); #if defined(_WIN32_WINNT) && _WIN32_WINNT >= 0x0602 // Windows 8 (minimum when UWP_ENABLED, but can be used in Windows builds) CREATEFILE2_EXTENDED_PARAMETERS extParams = {0}; @@ -810,7 +816,7 @@ lws_plat_write_file(const char *filename, void *buf, int len) { int m, fd; - fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC, 0600); + fd = lws_open(filename, O_WRONLY | O_CREAT | O_TRUNC, 0600); if (fd == -1) return -1; @@ -824,7 +830,7 @@ lws_plat_write_file(const char *filename, void *buf, int len) LWS_VISIBLE int lws_plat_read_file(const char *filename, void *buf, int len) { - int n, fd = open(filename, O_RDONLY); + int n, fd = lws_open(filename, O_RDONLY); if (fd == -1) return -1; diff --git a/thirdparty/libwebsockets/roles/h1/ops-h1.c b/thirdparty/libwebsockets/roles/h1/ops-h1.c index d3b16f4d1f..9001c864ea 100644 --- a/thirdparty/libwebsockets/roles/h1/ops-h1.c +++ b/thirdparty/libwebsockets/roles/h1/ops-h1.c @@ -195,6 +195,7 @@ postbody_completion: } break; + case LRS_RETURNED_CLOSE: case LRS_AWAITING_CLOSE_ACK: case LRS_WAITING_TO_SEND_CLOSE: case LRS_SHUTDOWN: @@ -444,6 +445,16 @@ try_pollout: if (lwsi_state(wsi) != LRS_ISSUING_FILE) { + if (wsi->trunc_len) { + //lwsl_notice("%s: completing partial\n", __func__); + if (lws_issue_raw(wsi, wsi->trunc_alloc + wsi->trunc_offset, + wsi->trunc_len) < 0) { + lwsl_info("%s signalling to close\n", __func__); + goto fail; + } + return LWS_HPI_RET_HANDLED; + } + lws_stats_atomic_bump(wsi->context, pt, LWSSTATS_C_WRITEABLE_CB, 1); #if defined(LWS_WITH_STATS) @@ -655,6 +666,9 @@ rops_destroy_role_h1(struct lws *wsi) ah = ah->next; } +#ifdef LWS_ROLE_WS + lws_free_set_NULL(wsi->ws); +#endif return 0; } diff --git a/thirdparty/libwebsockets/roles/http/client/client-handshake.c b/thirdparty/libwebsockets/roles/http/client/client-handshake.c index 4830fc9eca..0095c79a69 100644 --- a/thirdparty/libwebsockets/roles/http/client/client-handshake.c +++ b/thirdparty/libwebsockets/roles/http/client/client-handshake.c @@ -162,7 +162,7 @@ create_new_conn: if (!wsi->client_hostname_copy) wsi->client_hostname_copy = - strdup(lws_hdr_simple_ptr(wsi, + lws_strdup(lws_hdr_simple_ptr(wsi, _WSI_TOKEN_CLIENT_PEER_ADDRESS)); /* @@ -654,13 +654,13 @@ lws_client_reset(struct lws **pwsi, int ssl, const char *address, int port, lws_ssl_close(wsi); #endif + __remove_wsi_socket_from_fds(wsi); + if (wsi->context->event_loop_ops->close_handle_manually) wsi->context->event_loop_ops->close_handle_manually(wsi); else compatible_close(wsi->desc.sockfd); - __remove_wsi_socket_from_fds(wsi); - #if defined(LWS_WITH_TLS) wsi->tls.use_ssl = ssl; #else @@ -717,7 +717,7 @@ lws_client_reset(struct lws **pwsi, int ssl, const char *address, int port, } #ifdef LWS_WITH_HTTP_PROXY -static hubbub_error +hubbub_error html_parser_cb(const hubbub_token *token, void *pw) { struct lws_rewrite *r = (struct lws_rewrite *)pw; @@ -846,7 +846,7 @@ html_parser_cb(const hubbub_token *token, void *pw) #endif -static char * +char * lws_strdup(const char *s) { char *d = lws_malloc(strlen(s) + 1, "strdup"); diff --git a/thirdparty/libwebsockets/roles/http/client/client.c b/thirdparty/libwebsockets/roles/http/client/client.c index ce42dc6cd3..5645fa2b7a 100644 --- a/thirdparty/libwebsockets/roles/http/client/client.c +++ b/thirdparty/libwebsockets/roles/http/client/client.c @@ -93,7 +93,7 @@ lws_client_socket_service(struct lws *wsi, struct lws_pollfd *pollfd, char *sb = p; int n = 0; #if defined(LWS_WITH_SOCKS5) - char conn_mode = 0, pending_timeout = 0; + int conn_mode = 0, pending_timeout = 0; #endif if ((pollfd->revents & LWS_POLLOUT) && @@ -252,6 +252,8 @@ socks_reply_fail: /* clear his proxy connection timeout */ lws_set_timeout(wsi, NO_PENDING_TIMEOUT, 0); goto start_ws_handshake; + default: + break; } break; #endif @@ -578,6 +580,8 @@ lws_http_transaction_completed_client(struct lws *wsi) "queued client done"); } + _lws_header_table_reset(wsi->http.ah); + /* after the first one, they can only be coming from the queue */ wsi->transaction_from_pipeline_queue = 1; @@ -629,12 +633,20 @@ lws_http_transaction_completed_client(struct lws *wsi) } LWS_VISIBLE LWS_EXTERN unsigned int -lws_http_client_http_response(struct lws *wsi) +lws_http_client_http_response(struct lws *_wsi) { - if (!wsi->http.ah) - return 0; + struct lws *wsi; + unsigned int resp; + + if (_wsi->http.ah && _wsi->http.ah->http_response) + return _wsi->http.ah->http_response; - return wsi->http.ah->http_response; + lws_vhost_lock(_wsi->vhost); + wsi = _lws_client_wsi_master(_wsi); + resp = wsi->http.ah->http_response; + lws_vhost_unlock(_wsi->vhost); + + return resp; } #endif #if defined(LWS_PLAT_OPTEE) @@ -781,7 +793,7 @@ lws_client_interpret_server_handshake(struct lws *wsi) q = strrchr(new_path, '/'); if (q) lws_strncpy(q + 1, p, sizeof(new_path) - - (q - new_path)); + (q - new_path) - 1); else path = p; } @@ -910,9 +922,9 @@ lws_client_interpret_server_handshake(struct lws *wsi) * we seem to be good to go, give client last chance to check * headers and OK it */ - if (wsi->protocol->callback(wsi, + if (w->protocol->callback(w, LWS_CALLBACK_CLIENT_FILTER_PRE_ESTABLISH, - wsi->user_space, NULL, 0)) { + w->user_space, NULL, 0)) { cce = "HS: disallowed by client filter"; goto bail2; @@ -924,9 +936,9 @@ lws_client_interpret_server_handshake(struct lws *wsi) wsi->rxflow_change_to = LWS_RXFLOW_ALLOW; /* call him back to inform him he is up */ - if (wsi->protocol->callback(wsi, + if (w->protocol->callback(w, LWS_CALLBACK_ESTABLISHED_CLIENT_HTTP, - wsi->user_space, NULL, 0)) { + w->user_space, NULL, 0)) { cce = "HS: disallowed at ESTABLISHED"; goto bail3; } @@ -964,9 +976,9 @@ bail2: n = 0; if (cce) n = (int)strlen(cce); - wsi->protocol->callback(wsi, + w->protocol->callback(w, LWS_CALLBACK_CLIENT_CONNECTION_ERROR, - wsi->user_space, (void *)cce, + w->user_space, (void *)cce, (unsigned int)n); } wsi->already_did_cce = 1; @@ -1228,4 +1240,4 @@ completed: return 0; } -#endif
\ No newline at end of file +#endif diff --git a/thirdparty/libwebsockets/roles/http/header.c b/thirdparty/libwebsockets/roles/http/header.c index 99e56f7564..dbcf27cbd1 100644 --- a/thirdparty/libwebsockets/roles/http/header.c +++ b/thirdparty/libwebsockets/roles/http/header.c @@ -26,7 +26,7 @@ const unsigned char * lws_token_to_string(enum lws_token_indexes token) { - if ((unsigned int)token >= ARRAY_SIZE(set)) + if ((unsigned int)token >= LWS_ARRAY_SIZE(set)) return NULL; return (unsigned char *)set[token]; @@ -149,9 +149,17 @@ lws_add_http_common_headers(struct lws *wsi, unsigned int code, (int)strlen(content_type), p, end)) return 1; - if (content_len != LWS_ILLEGAL_HTTP_CONTENT_LEN && - lws_add_http_header_content_length(wsi, content_len, p, end)) - return 1; + if (content_len != LWS_ILLEGAL_HTTP_CONTENT_LEN) { + if (lws_add_http_header_content_length(wsi, content_len, p, end)) + return 1; + } else { + if (lws_add_http_header_by_token(wsi, WSI_TOKEN_CONNECTION, + (unsigned char *)"close", 5, + p, end)) + return 1; + + wsi->http.connection_type = HTTP_CONNECTION_CLOSE; + } return 0; } @@ -204,34 +212,40 @@ lws_add_http_header_status(struct lws *wsi, unsigned int _code, #endif #ifdef LWS_WITH_HTTP2 - if (lwsi_role_h2(wsi) || lwsi_role_h2_ENCAPSULATION(wsi)) - return lws_add_http2_header_status(wsi, code, p, end); + if (lwsi_role_h2(wsi) || lwsi_role_h2_ENCAPSULATION(wsi)) { + n = lws_add_http2_header_status(wsi, code, p, end); + if (n) + return n; + } else #endif - if (code >= 400 && code < (400 + ARRAY_SIZE(err400))) - description = err400[code - 400]; - if (code >= 500 && code < (500 + ARRAY_SIZE(err500))) - description = err500[code - 500]; - - if (code == 100) - description = "Continue"; - if (code == 200) - description = "OK"; - if (code == 304) - description = "Not Modified"; - else - if (code >= 300 && code < 400) - description = "Redirect"; - - if (wsi->http.request_version < ARRAY_SIZE(hver)) - p1 = hver[wsi->http.request_version]; - else - p1 = hver[0]; - - n = sprintf((char *)code_and_desc, "%s %u %s", p1, code, description); - - if (lws_add_http_header_by_name(wsi, NULL, code_and_desc, n, p, end)) - return 1; - + { + if (code >= 400 && code < (400 + LWS_ARRAY_SIZE(err400))) + description = err400[code - 400]; + if (code >= 500 && code < (500 + LWS_ARRAY_SIZE(err500))) + description = err500[code - 500]; + + if (code == 100) + description = "Continue"; + if (code == 200) + description = "OK"; + if (code == 304) + description = "Not Modified"; + else + if (code >= 300 && code < 400) + description = "Redirect"; + + if (wsi->http.request_version < LWS_ARRAY_SIZE(hver)) + p1 = hver[wsi->http.request_version]; + else + p1 = hver[0]; + + n = sprintf((char *)code_and_desc, "%s %u %s", p1, code, + description); + + if (lws_add_http_header_by_name(wsi, NULL, code_and_desc, n, p, + end)) + return 1; + } headers = wsi->vhost->headers; while (headers) { if (lws_add_http_header_by_name(wsi, diff --git a/thirdparty/libwebsockets/roles/http/private.h b/thirdparty/libwebsockets/roles/http/private.h index 2aa7a92f75..5699914742 100644 --- a/thirdparty/libwebsockets/roles/http/private.h +++ b/thirdparty/libwebsockets/roles/http/private.h @@ -227,6 +227,7 @@ struct _lws_http_mode_related { #if defined(LWS_WITH_HTTP_PROXY) unsigned int perform_rewrite:1; #endif + unsigned int deferred_transaction_completed:1; }; diff --git a/thirdparty/libwebsockets/roles/http/server/lejp-conf.c b/thirdparty/libwebsockets/roles/http/server/lejp-conf.c index e9ce854cfc..fbf10c288e 100644 --- a/thirdparty/libwebsockets/roles/http/server/lejp-conf.c +++ b/thirdparty/libwebsockets/roles/http/server/lejp-conf.c @@ -205,6 +205,7 @@ struct jpargs { unsigned int enable_client_ssl:1; unsigned int fresh_mount:1; unsigned int any_vhosts:1; + unsigned int chunk:1; }; static void * @@ -213,6 +214,8 @@ lwsws_align(struct jpargs *a) if ((lws_intptr_t)(a->p) & 15) a->p += 16 - ((lws_intptr_t)(a->p) & 15); + a->chunk = 0; + return a->p; } @@ -225,7 +228,7 @@ arg_to_bool(const char *s) if (n) return 1; - for (n = 0; n < (int)ARRAY_SIZE(on); n++) + for (n = 0; n < (int)LWS_ARRAY_SIZE(on); n++) if (!strcasecmp(s, on[n])) return 1; @@ -413,25 +416,30 @@ lejp_vhosts_cb(struct lejp_ctx *ctx, char reason) } /* this catches, eg, vhosts[].headers[].xxx */ - if (reason == LEJPCB_VAL_STR_END && + if ((reason == LEJPCB_VAL_STR_END || reason == LEJPCB_VAL_STR_CHUNK) && ctx->path_match == LEJPVP_HEADERS_NAME + 1) { - headers = lwsws_align(a); - a->p += sizeof(*headers); - - n = lejp_get_wildcard(ctx, 0, a->p, a->end - a->p); - /* ie, enable this protocol, no options yet */ - headers->next = a->info->headers; - a->info->headers = headers; - headers->name = a->p; - // lwsl_notice(" adding header %s=%s\n", a->p, ctx->buf); - a->p += n - 1; - *(a->p++) = ':'; - if (a->p < a->end) - *(a->p++) = '\0'; - else - *(a->p - 1) = '\0'; - headers->value = a->p; - headers->options = NULL; + if (!a->chunk) { + headers = lwsws_align(a); + a->p += sizeof(*headers); + + n = lejp_get_wildcard(ctx, 0, a->p, + lws_ptr_diff(a->end, a->p)); + /* ie, add this header */ + headers->next = a->info->headers; + a->info->headers = headers; + headers->name = a->p; + + lwsl_notice(" adding header %s=%s\n", a->p, ctx->buf); + a->p += n - 1; + *(a->p++) = ':'; + if (a->p < a->end) + *(a->p++) = '\0'; + else + *(a->p - 1) = '\0'; + headers->value = a->p; + headers->options = NULL; + } + a->chunk = reason == LEJPCB_VAL_STR_CHUNK; goto dostring; } @@ -502,7 +510,7 @@ lejp_vhosts_cb(struct lejp_ctx *ctx, char reason) if (a->last) a->last->mount_next = m; - for (n = 0; n < (int)ARRAY_SIZE(mount_protocols); n++) + for (n = 0; n < (int)LWS_ARRAY_SIZE(mount_protocols); n++) if (!strncmp(a->m.origin, mount_protocols[n], strlen(mount_protocols[n]))) { lwsl_info("----%s\n", a->m.origin); @@ -512,7 +520,7 @@ lejp_vhosts_cb(struct lejp_ctx *ctx, char reason) break; } - if (n == (int)ARRAY_SIZE(mount_protocols)) { + if (n == (int)LWS_ARRAY_SIZE(mount_protocols)) { lwsl_err("unsupported protocol:// %s\n", a->m.origin); return 1; } @@ -750,6 +758,7 @@ lejp_vhosts_cb(struct lejp_ctx *ctx, char reason) dostring: p = ctx->buf; + p[LEJP_STRING_CHUNK] = '\0'; p1 = strstr(p, ESC_INSTALL_DATADIR); if (p1) { n = p1 - p; @@ -762,7 +771,8 @@ dostring: } a->p += lws_snprintf(a->p, a->end - a->p, "%s", p); - *(a->p)++ = '\0'; + if (reason == LEJPCB_VAL_STR_END) + *(a->p)++ = '\0'; return 0; } @@ -779,7 +789,7 @@ lwsws_get_config(void *user, const char *f, const char * const *paths, struct lejp_ctx ctx; int n, m, fd; - fd = open(f, O_RDONLY); + fd = lws_open(f, O_RDONLY); if (fd < 0) { lwsl_err("Cannot open %s\n", f); return 2; @@ -927,11 +937,11 @@ lwsws_get_config_globals(struct lws_context_creation_info *info, const char *d, lws_snprintf(dd, sizeof(dd) - 1, "%s/conf", d); if (lwsws_get_config(&a, dd, paths_global, - ARRAY_SIZE(paths_global), lejp_globals_cb) > 1) + LWS_ARRAY_SIZE(paths_global), lejp_globals_cb) > 1) return 1; lws_snprintf(dd, sizeof(dd) - 1, "%s/conf.d", d); if (lwsws_get_config_d(&a, dd, paths_global, - ARRAY_SIZE(paths_global), lejp_globals_cb) > 1) + LWS_ARRAY_SIZE(paths_global), lejp_globals_cb) > 1) return 1; a.plugin_dirs[a.count_plugin_dirs] = NULL; @@ -962,11 +972,11 @@ lwsws_get_config_vhosts(struct lws_context *context, lws_snprintf(dd, sizeof(dd) - 1, "%s/conf", d); if (lwsws_get_config(&a, dd, paths_vhosts, - ARRAY_SIZE(paths_vhosts), lejp_vhosts_cb) > 1) + LWS_ARRAY_SIZE(paths_vhosts), lejp_vhosts_cb) > 1) return 1; lws_snprintf(dd, sizeof(dd) - 1, "%s/conf.d", d); if (lwsws_get_config_d(&a, dd, paths_vhosts, - ARRAY_SIZE(paths_vhosts), lejp_vhosts_cb) > 1) + LWS_ARRAY_SIZE(paths_vhosts), lejp_vhosts_cb) > 1) return 1; *cs = a.p; diff --git a/thirdparty/libwebsockets/roles/http/server/parsers.c b/thirdparty/libwebsockets/roles/http/server/parsers.c index cb022e362b..482bdc676a 100644 --- a/thirdparty/libwebsockets/roles/http/server/parsers.c +++ b/thirdparty/libwebsockets/roles/http/server/parsers.c @@ -563,7 +563,7 @@ int LWS_WARN_UNUSED_RESULT lws_hdr_simple_create(struct lws *wsi, enum lws_token_indexes h, const char *s) { wsi->http.ah->nfrag++; - if (wsi->http.ah->nfrag == ARRAY_SIZE(wsi->http.ah->frags)) { + if (wsi->http.ah->nfrag == LWS_ARRAY_SIZE(wsi->http.ah->frags)) { lwsl_warn("More hdr frags than we can deal with, dropping\n"); return -1; } @@ -677,18 +677,16 @@ lws_parse_urldecode(struct lws *wsi, uint8_t *_c) return -1; /* genuine delimiter */ if ((c == '&' || c == ';') && !enc) { - if (issue_char(wsi, c) < 0) + if (issue_char(wsi, '\0') < 0) return -1; - /* swallow the terminator */ - ah->frags[ah->nfrag].len--; /* link to next fragment */ ah->frags[ah->nfrag].nfrag = ah->nfrag + 1; ah->nfrag++; - if (ah->nfrag >= ARRAY_SIZE(ah->frags)) + if (ah->nfrag >= LWS_ARRAY_SIZE(ah->frags)) goto excessive; /* start next fragment after the & */ ah->post_literal_equal = 0; - ah->frags[ah->nfrag].offset = ah->pos; + ah->frags[ah->nfrag].offset = ++ah->pos; ah->frags[ah->nfrag].len = 0; ah->frags[ah->nfrag].nfrag = 0; goto swallow; @@ -787,9 +785,9 @@ lws_parse_urldecode(struct lws *wsi, uint8_t *_c) /* move to using WSI_TOKEN_HTTP_URI_ARGS */ ah->nfrag++; - if (ah->nfrag >= ARRAY_SIZE(ah->frags)) + if (ah->nfrag >= LWS_ARRAY_SIZE(ah->frags)) goto excessive; - ah->frags[ah->nfrag].offset = ah->pos; + ah->frags[ah->nfrag].offset = ++ah->pos; ah->frags[ah->nfrag].len = 0; ah->frags[ah->nfrag].nfrag = 0; @@ -852,10 +850,10 @@ lws_parse(struct lws *wsi, unsigned char *buf, int *len) c == ' ') break; - for (m = 0; m < ARRAY_SIZE(methods); m++) + for (m = 0; m < LWS_ARRAY_SIZE(methods); m++) if (ah->parser_state == methods[m]) break; - if (m == ARRAY_SIZE(methods)) + if (m == LWS_ARRAY_SIZE(methods)) /* it was not any of the methods */ goto check_eol; @@ -983,7 +981,7 @@ nope: if (ah->lextable_pos < 0 && lwsi_role_h1(wsi) && lwsi_role_server(wsi)) { /* this is not a header we know about */ - for (m = 0; m < ARRAY_SIZE(methods); m++) + for (m = 0; m < LWS_ARRAY_SIZE(methods); m++) if (ah->frag_index[methods[m]]) { /* * already had the method, no idea what @@ -996,7 +994,7 @@ nope: * hm it's an unknown http method from a client in fact, * it cannot be valid http */ - if (m == ARRAY_SIZE(methods)) { + if (m == LWS_ARRAY_SIZE(methods)) { /* * are we set up to accept raw in these cases? */ @@ -1025,7 +1023,7 @@ nope: lextable[ah->lextable_pos + 1]; lwsl_parser("known hdr %d\n", n); - for (m = 0; m < ARRAY_SIZE(methods); m++) + for (m = 0; m < LWS_ARRAY_SIZE(methods); m++) if (n == methods[m] && ah->frag_index[methods[m]]) { lwsl_warn("Duplicated method\n"); @@ -1061,7 +1059,7 @@ nope: start_fragment: ah->nfrag++; excessive: - if (ah->nfrag == ARRAY_SIZE(ah->frags)) { + if (ah->nfrag == LWS_ARRAY_SIZE(ah->frags)) { lwsl_warn("More hdr frags than we can deal with\n"); return -1; } diff --git a/thirdparty/libwebsockets/roles/http/server/server.c b/thirdparty/libwebsockets/roles/http/server/server.c index 350af3cd7e..abd86dc9b5 100644 --- a/thirdparty/libwebsockets/roles/http/server/server.c +++ b/thirdparty/libwebsockets/roles/http/server/server.c @@ -131,6 +131,17 @@ done_list: (void)n; #if defined(__linux__) +#ifdef LWS_WITH_UNIX_SOCK + /* + * A Unix domain sockets cannot be bound for several times, even if we set + * the SO_REUSE* options on. + * However, fortunately, each thread is able to independently listen when + * running on a reasonably new Linux kernel. So we can safely assume + * creating just one listening socket for a multi-threaded environment won't + * fail in most cases. + */ + if (!LWS_UNIX_SOCK_ENABLED(vhost)) +#endif limit = vhost->context->count_threads; #endif @@ -694,7 +705,7 @@ lws_find_string_in_file(const char *filename, const char *string, int stringlen) char buf[128]; int fd, match = 0, pos = 0, n = 0, hit = 0; - fd = open(filename, O_RDONLY); + fd = lws_open(filename, O_RDONLY); if (fd < 0) { lwsl_err("can't open auth file: %s\n", filename); return 0; @@ -812,7 +823,7 @@ lws_http_get_uri_and_method(struct lws *wsi, char **puri_ptr, int *puri_len) { int n, count = 0; - for (n = 0; n < (int)ARRAY_SIZE(methods); n++) + for (n = 0; n < (int)LWS_ARRAY_SIZE(methods); n++) if (lws_hdr_total_length(wsi, methods[n])) count++; if (!count) { @@ -827,7 +838,7 @@ lws_http_get_uri_and_method(struct lws *wsi, char **puri_ptr, int *puri_len) return -1; } - for (n = 0; n < (int)ARRAY_SIZE(methods); n++) + for (n = 0; n < (int)LWS_ARRAY_SIZE(methods); n++) if (lws_hdr_total_length(wsi, methods[n])) { *puri_ptr = lws_hdr_simple_ptr(wsi, methods[n]); *puri_len = lws_hdr_total_length(wsi, methods[n]); @@ -857,7 +868,7 @@ lws_http_action(struct lws *wsi) }; meth = lws_http_get_uri_and_method(wsi, &uri_ptr, &uri_len); - if (meth < 0 || meth >= (int)ARRAY_SIZE(method_names)) + if (meth < 0 || meth >= (int)LWS_ARRAY_SIZE(method_names)) goto bail_nuke_ah; /* we insist on absolute paths */ @@ -1128,7 +1139,7 @@ lws_http_action(struct lws *wsi) } if (pcolon > pslash) pcolon = NULL; - + if (pcolon) n = pcolon - hit->origin; else @@ -1142,13 +1153,13 @@ lws_http_action(struct lws *wsi) i.address = ads; i.port = 80; - if (hit->origin_protocol == LWSMPRO_HTTPS) { + if (hit->origin_protocol == LWSMPRO_HTTPS) { i.port = 443; i.ssl_connection = 1; } if (pcolon) i.port = atoi(pcolon + 1); - + lws_snprintf(rpath, sizeof(rpath) - 1, "/%s/%s", pslash + 1, uri_ptr + hit->mountpoint_len); lws_clean_url(rpath); @@ -1164,7 +1175,7 @@ lws_http_action(struct lws *wsi) p++; } } - + i.path = rpath; i.host = i.address; @@ -1178,7 +1189,7 @@ lws_http_action(struct lws *wsi) "from %s, to %s\n", i.address, i.port, i.path, i.ssl_connection, i.uri_replace_from, i.uri_replace_to); - + if (!lws_client_connect_via_info(&i)) { lwsl_err("proxy connect fail\n"); return 1; @@ -1714,12 +1725,31 @@ lws_http_transaction_completed(struct lws *wsi) { int n = NO_PENDING_TIMEOUT; + if (wsi->trunc_len) { + /* + * ...so he tried to send something large as the http reply, + * it went as a partial, but he immediately said the + * transaction was completed. + * + * Defer the transaction completed until the last part of the + * partial is sent. + */ + lwsl_notice("%s: deferring due to partial\n", __func__); + wsi->http.deferred_transaction_completed = 1; + + return 0; + } + lwsl_info("%s: wsi %p\n", __func__, wsi); lws_access_log(wsi); if (!wsi->hdr_parsing_completed) { - lwsl_notice("%s: ignoring, ah parsing incomplete\n", __func__); + char peer[64]; + lws_get_peer_simple(wsi, peer, sizeof(peer) - 1); + peer[sizeof(peer) - 1] = '\0'; + lwsl_notice("%s: (from %s) ignoring, ah parsing incomplete\n", + __func__, peer); return 0; } diff --git a/thirdparty/libwebsockets/roles/ws/client-parser-ws.c b/thirdparty/libwebsockets/roles/ws/client-parser-ws.c index aa561ce034..7287fb1590 100644 --- a/thirdparty/libwebsockets/roles/ws/client-parser-ws.c +++ b/thirdparty/libwebsockets/roles/ws/client-parser-ws.c @@ -450,7 +450,7 @@ ping_drop: break; case LWSWSOPC_PONG: - lwsl_info("client receied pong\n"); + lwsl_info("client received pong\n"); lwsl_hexdump(&wsi->ws->rx_ubuf[LWS_PRE], wsi->ws->rx_ubuf_head); @@ -488,9 +488,6 @@ ping_drop: ebuf.token = &wsi->ws->rx_ubuf[LWS_PRE]; ebuf.len = wsi->ws->rx_ubuf_head; - if (wsi->ws->opcode == LWSWSOPC_PONG && !ebuf.len) - goto already_done; - #if !defined(LWS_WITHOUT_EXTENSIONS) drain_extension: lwsl_ext("%s: passing %d to ext\n", __func__, ebuf.len); @@ -504,14 +501,12 @@ drain_extension: #endif lwsl_debug("post inflate ebuf len %d\n", ebuf.len); - if ( #if !defined(LWS_WITHOUT_EXTENSIONS) - rx_draining_ext && -#endif - !ebuf.len) { + if (rx_draining_ext && !ebuf.len) { lwsl_debug(" --- ending drain on 0 read result\n"); goto already_done; } +#endif if (wsi->ws->check_utf8 && !wsi->ws->defeat_check_utf8) { if (lws_check_utf8(&wsi->ws->utf8, diff --git a/thirdparty/libwebsockets/roles/ws/ops-ws.c b/thirdparty/libwebsockets/roles/ws/ops-ws.c index 5ddaba9e18..665b2c9b74 100644 --- a/thirdparty/libwebsockets/roles/ws/ops-ws.c +++ b/thirdparty/libwebsockets/roles/ws/ops-ws.c @@ -1246,8 +1246,7 @@ int rops_handle_POLLOUT_ws(struct lws *wsi) return LWS_HP_RET_BAIL_OK; } - if (lwsi_role_client(wsi) && !wsi->socket_is_permanently_unusable && - wsi->ws->send_check_ping) { + if (!wsi->socket_is_permanently_unusable && wsi->ws->send_check_ping) { lwsl_info("issuing ping on wsi %p\n", wsi); wsi->ws->send_check_ping = 0; @@ -1282,7 +1281,7 @@ int rops_handle_POLLOUT_ws(struct lws *wsi) * payload ordering, but since they are always complete * fragments control packets can interleave OK. */ - if (lwsi_role_client(wsi) && wsi->ws->tx_draining_ext) { + if (wsi->ws->tx_draining_ext) { lwsl_ext("SERVICING TX EXT DRAINING\n"); if (lws_write(wsi, NULL, 0, LWS_WRITE_CONTINUATION) < 0) return LWS_HP_RET_BAIL_DIE; @@ -1292,8 +1291,10 @@ int rops_handle_POLLOUT_ws(struct lws *wsi) /* Priority 6: extensions */ - if (!wsi->ws->extension_data_pending) + if (!wsi->ws->extension_data_pending && !wsi->ws->tx_draining_ext) { + lwsl_ext("%s: !wsi->ws->extension_data_pending\n", __func__); return LWS_HP_RET_USER_SERVICE; + } /* * check in on the active extensions, see if they @@ -1412,15 +1413,13 @@ rops_periodic_checks_ws(struct lws_context *context, int tsi, time_t now) wsi->ws->time_next_ping_check) > context->ws_ping_pong_interval) { - lwsl_info("req pp on wsi %p\n", - wsi); + lwsl_info("req pp on wsi %p\n", wsi); wsi->ws->send_check_ping = 1; lws_set_timeout(wsi, PENDING_TIMEOUT_WS_PONG_CHECK_SEND_PING, context->timeout_secs); lws_callback_on_writable(wsi); - wsi->ws->time_next_ping_check = - now; + wsi->ws->time_next_ping_check = now; } wsi = wsi->same_vh_protocol_next; } @@ -1466,6 +1465,9 @@ rops_service_flag_pending_ws(struct lws_context *context, int tsi) static int rops_close_via_role_protocol_ws(struct lws *wsi, enum lws_close_status reason) { + if (!wsi->ws) + return 0; + if (!wsi->ws->close_in_ping_buffer_len && /* already a reason */ (reason == LWS_CLOSE_STATUS_NOSTATUS || reason == LWS_CLOSE_STATUS_NOSTATUS_CONTEXT_DESTROY)) @@ -1512,7 +1514,7 @@ rops_close_role_ws(struct lws_context_per_thread *pt, struct lws *wsi) if (wsi->ws->tx_draining_ext) { struct lws **w = &pt->ws.tx_draining_ext_list; - lwsl_notice("%s: CLEARING tx_draining_ext\n", __func__); + lwsl_ext("%s: CLEARING tx_draining_ext\n", __func__); wsi->ws->tx_draining_ext = 0; /* remove us from context draining ext list */ while (*w) { @@ -1563,7 +1565,7 @@ rops_write_role_protocol_ws(struct lws *wsi, unsigned char *buf, size_t len, /* remove us from the list */ struct lws **w = &pt->ws.tx_draining_ext_list; - lwsl_notice("%s: CLEARING tx_draining_ext\n", __func__); + lwsl_ext("%s: CLEARING tx_draining_ext\n", __func__); wsi->ws->tx_draining_ext = 0; /* remove us from context draining ext list */ while (*w) { @@ -1588,7 +1590,7 @@ rops_write_role_protocol_ws(struct lws *wsi, unsigned char *buf, size_t len, if (!(wpt & LWS_WRITE_NO_FIN) && len) *wp &= ~LWS_WRITE_NO_FIN; - lwsl_notice("FORCED draining wp to 0x%02X (stashed 0x%02X, incoming 0x%02X)\n", *wp, + lwsl_ext("FORCED draining wp to 0x%02X (stashed 0x%02X, incoming 0x%02X)\n", *wp, wsi->ws->tx_draining_stashed_wp, wpt); // assert(0); } @@ -1644,7 +1646,7 @@ rops_write_role_protocol_ws(struct lws *wsi, unsigned char *buf, size_t len, // lwsl_notice("ext processed %d plaintext into %d compressed (wp 0x%x)\n", m, (int)ebuf.len, *wp); if (n && ebuf.len) { - lwsl_notice("write drain len %d (wp 0x%x) SETTING tx_draining_ext\n", (int)ebuf.len, *wp); + lwsl_ext("write drain len %d (wp 0x%x) SETTING tx_draining_ext\n", (int)ebuf.len, *wp); /* extension requires further draining */ wsi->ws->tx_draining_ext = 1; wsi->ws->tx_draining_ext_list = pt->ws.tx_draining_ext_list; diff --git a/thirdparty/libwebsockets/tls/mbedtls/mbedtls-server.c b/thirdparty/libwebsockets/tls/mbedtls/mbedtls-server.c index 2de6d422e3..f17c7e5494 100644 --- a/thirdparty/libwebsockets/tls/mbedtls/mbedtls-server.c +++ b/thirdparty/libwebsockets/tls/mbedtls/mbedtls-server.c @@ -631,7 +631,7 @@ lws_tls_acme_sni_csr_create(struct lws_context *context, const char *elements[], /* subject must be formatted like "C=TW,O=warmcat,CN=myserver" */ - for (n = 0; n < (int)ARRAY_SIZE(x5); n++) { + for (n = 0; n < (int)LWS_ARRAY_SIZE(x5); n++) { if (p != subject) *p++ = ','; if (elements[n]) diff --git a/thirdparty/libwebsockets/tls/mbedtls/ssl.c b/thirdparty/libwebsockets/tls/mbedtls/ssl.c index 6ae9d2556b..f311ef50e3 100644 --- a/thirdparty/libwebsockets/tls/mbedtls/ssl.c +++ b/thirdparty/libwebsockets/tls/mbedtls/ssl.c @@ -121,8 +121,6 @@ lws_ssl_capable_read(struct lws *wsi, unsigned char *buf, int len) if (wsi->vhost) wsi->vhost->conn_stats.rx += n; - lws_restart_ws_ping_pong_timer(wsi); - /* * if it was our buffer that limited what we read, * check if SSL has additional data pending inside SSL buffers. diff --git a/thirdparty/libwebsockets/tls/mbedtls/wrapper/include/internal/ssl_types.h b/thirdparty/libwebsockets/tls/mbedtls/wrapper/include/internal/ssl_types.h index ba19663d9e..68ac748a28 100644 --- a/thirdparty/libwebsockets/tls/mbedtls/wrapper/include/internal/ssl_types.h +++ b/thirdparty/libwebsockets/tls/mbedtls/wrapper/include/internal/ssl_types.h @@ -37,7 +37,11 @@ typedef void RSA; typedef void STACK; typedef void BIO; +#if defined(WIN32) || defined(_WIN32) +#define ossl_inline __inline +#else #define ossl_inline inline +#endif #define SSL_METHOD_CALL(f, s, ...) s->method->func->ssl_##f(s, ##__VA_ARGS__) #define X509_METHOD_CALL(f, x, ...) x->method->x509_##f(x, ##__VA_ARGS__) diff --git a/thirdparty/libwebsockets/uwp_fixes.diff b/thirdparty/libwebsockets/uwp_fixes.diff index 5b9d8724ed..21c3275bba 100644 --- a/thirdparty/libwebsockets/uwp_fixes.diff +++ b/thirdparty/libwebsockets/uwp_fixes.diff @@ -1,15 +1,15 @@ diff --git a/thirdparty/libwebsockets/plat/lws-plat-win.c b/thirdparty/libwebsockets/plat/lws-plat-win.c -index 948db6289..511e29739 100644 +index bd513b494..1850b6425 100644 --- a/thirdparty/libwebsockets/plat/lws-plat-win.c +++ b/thirdparty/libwebsockets/plat/lws-plat-win.c -@@ -635,9 +635,20 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, +@@ -641,9 +641,20 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, HANDLE ret; WCHAR buf[MAX_PATH]; lws_fop_fd_t fop_fd; - LARGE_INTEGER llFileSize = {0}; + FILE_STANDARD_INFO fInfo = {0}; - MultiByteToWideChar(CP_UTF8, 0, filename, -1, buf, ARRAY_SIZE(buf)); + MultiByteToWideChar(CP_UTF8, 0, filename, -1, buf, LWS_ARRAY_SIZE(buf)); + +#if defined(_WIN32_WINNT) && _WIN32_WINNT >= 0x0602 // Windows 8 (minimum when UWP_ENABLED, but can be used in Windows builds) + CREATEFILE2_EXTENDED_PARAMETERS extParams = {0}; @@ -24,7 +24,7 @@ index 948db6289..511e29739 100644 if (((*flags) & 7) == _O_RDONLY) { ret = CreateFileW(buf, GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); -@@ -645,6 +656,7 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, +@@ -651,6 +662,7 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, ret = CreateFileW(buf, GENERIC_WRITE, 0, NULL, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL); } @@ -32,7 +32,7 @@ index 948db6289..511e29739 100644 if (ret == LWS_INVALID_FILE) goto bail; -@@ -657,9 +669,9 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, +@@ -663,9 +675,9 @@ _lws_plat_file_open(const struct lws_plat_file_ops *fops, const char *filename, fop_fd->fd = ret; fop_fd->filesystem_priv = NULL; /* we don't use it */ fop_fd->flags = *flags; diff --git a/thirdparty/libwebsockets/win32helpers/getopt.c b/thirdparty/libwebsockets/win32helpers/getopt.c index 2181f1cb12..3bb21f6f28 100644 --- a/thirdparty/libwebsockets/win32helpers/getopt.c +++ b/thirdparty/libwebsockets/win32helpers/getopt.c @@ -1,153 +1,153 @@ -/* $NetBSD: getopt.c,v 1.16 1999/12/02 13:15:56 kleink Exp $ */
-
-/*
- * Copyright (c) 1987, 1993, 1994
- * The Regents of the University of California. All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in the
- * documentation and/or other materials provided with the distribution.
- * 3. All advertising materials mentioning features or use of this software
- * must display the following acknowledgement:
- * This product includes software developed by the University of
- * California, Berkeley and its contributors.
- * 4. Neither the name of the University nor the names of its contributors
- * may be used to endorse or promote products derived from this software
- * without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#if 0
-static char sccsid[] = "@(#)getopt.c 8.3 (Berkeley) 4/27/95";
-#endif
-
-#include <assert.h>
-#include <errno.h>
-#include <stdio.h>
-#include <string.h>
-
-#define __P(x) x
-#define _DIAGASSERT(x) assert(x)
-
-#ifdef __weak_alias
-__weak_alias(getopt,_getopt);
-#endif
-
-
-int opterr = 1, /* if error message should be printed */
- optind = 1, /* index into parent argv vector */
- optopt, /* character checked for validity */
- optreset; /* reset getopt */
-char *optarg; /* argument associated with option */
-
-static char * _progname __P((char *));
-int getopt_internal __P((int, char * const *, const char *));
-
-static char *
-_progname(nargv0)
- char * nargv0;
-{
- char * tmp;
-
- _DIAGASSERT(nargv0 != NULL);
-
- tmp = strrchr(nargv0, '/');
- if (tmp)
- tmp++;
- else
- tmp = nargv0;
- return(tmp);
-}
-
-#define BADCH (int)'?'
-#define BADARG (int)':'
-#define EMSG ""
-
-/*
- * getopt --
- * Parse argc/argv argument vector.
- */
-int
-getopt(nargc, nargv, ostr)
- int nargc;
- char * const nargv[];
- const char *ostr;
-{
- static char *__progname = 0;
- static char *place = EMSG; /* option letter processing */
- char *oli; /* option letter list index */
- __progname = __progname?__progname:_progname(*nargv);
-
- _DIAGASSERT(nargv != NULL);
- _DIAGASSERT(ostr != NULL);
-
- if (optreset || !*place) { /* update scanning pointer */
- optreset = 0;
- if (optind >= nargc || *(place = nargv[optind]) != '-') {
- place = EMSG;
- return (-1);
- }
- if (place[1] && *++place == '-' /* found "--" */
- && place[1] == '\0') {
- ++optind;
- place = EMSG;
- return (-1);
- }
- } /* option letter okay? */
- if ((optopt = (int)*place++) == (int)':' ||
- !(oli = strchr(ostr, optopt))) {
- /*
- * if the user didn't specify '-' as an option,
- * assume it means -1.
- */
- if (optopt == (int)'-')
- return (-1);
- if (!*place)
- ++optind;
- if (opterr && *ostr != ':')
- (void)fprintf(stderr,
- "%s: illegal option -- %c\n", __progname, optopt);
- return (BADCH);
- }
- if (*++oli != ':') { /* don't need argument */
- optarg = NULL;
- if (!*place)
- ++optind;
- }
- else { /* need an argument */
- if (*place) /* no white space */
- optarg = place;
- else if (nargc <= ++optind) { /* no arg */
- place = EMSG;
- if (*ostr == ':')
- return (BADARG);
- if (opterr)
- (void)fprintf(stderr,
- "%s: option requires an argument -- %c\n",
- __progname, optopt);
- return (BADCH);
- }
- else /* white space */
- optarg = nargv[optind];
- place = EMSG;
- ++optind;
- }
- return (optopt); /* dump back option letter */
-}
-
+/* $NetBSD: getopt.c,v 1.16 1999/12/02 13:15:56 kleink Exp $ */ + +/* + * Copyright (c) 1987, 1993, 1994 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. All advertising materials mentioning features or use of this software + * must display the following acknowledgement: + * This product includes software developed by the University of + * California, Berkeley and its contributors. + * 4. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#if 0 +static char sccsid[] = "@(#)getopt.c 8.3 (Berkeley) 4/27/95"; +#endif + +#include <assert.h> +#include <errno.h> +#include <stdio.h> +#include <string.h> + +#define __P(x) x +#define _DIAGASSERT(x) assert(x) + +#ifdef __weak_alias +__weak_alias(getopt,_getopt); +#endif + + +int opterr = 1, /* if error message should be printed */ + optind = 1, /* index into parent argv vector */ + optopt, /* character checked for validity */ + optreset; /* reset getopt */ +char *optarg; /* argument associated with option */ + +static char * _progname __P((char *)); +int getopt_internal __P((int, char * const *, const char *)); + +static char * +_progname(nargv0) + char * nargv0; +{ + char * tmp; + + _DIAGASSERT(nargv0 != NULL); + + tmp = strrchr(nargv0, '/'); + if (tmp) + tmp++; + else + tmp = nargv0; + return(tmp); +} + +#define BADCH (int)'?' +#define BADARG (int)':' +#define EMSG "" + +/* + * getopt -- + * Parse argc/argv argument vector. + */ +int +getopt(nargc, nargv, ostr) + int nargc; + char * const nargv[]; + const char *ostr; +{ + static char *__progname = 0; + static char *place = EMSG; /* option letter processing */ + char *oli; /* option letter list index */ + __progname = __progname?__progname:_progname(*nargv); + + _DIAGASSERT(nargv != NULL); + _DIAGASSERT(ostr != NULL); + + if (optreset || !*place) { /* update scanning pointer */ + optreset = 0; + if (optind >= nargc || *(place = nargv[optind]) != '-') { + place = EMSG; + return (-1); + } + if (place[1] && *++place == '-' /* found "--" */ + && place[1] == '\0') { + ++optind; + place = EMSG; + return (-1); + } + } /* option letter okay? */ + if ((optopt = (int)*place++) == (int)':' || + !(oli = strchr(ostr, optopt))) { + /* + * if the user didn't specify '-' as an option, + * assume it means -1. + */ + if (optopt == (int)'-') + return (-1); + if (!*place) + ++optind; + if (opterr && *ostr != ':') + (void)fprintf(stderr, + "%s: illegal option -- %c\n", __progname, optopt); + return (BADCH); + } + if (*++oli != ':') { /* don't need argument */ + optarg = NULL; + if (!*place) + ++optind; + } + else { /* need an argument */ + if (*place) /* no white space */ + optarg = place; + else if (nargc <= ++optind) { /* no arg */ + place = EMSG; + if (*ostr == ':') + return (BADARG); + if (opterr) + (void)fprintf(stderr, + "%s: option requires an argument -- %c\n", + __progname, optopt); + return (BADCH); + } + else /* white space */ + optarg = nargv[optind]; + place = EMSG; + ++optind; + } + return (optopt); /* dump back option letter */ +} + diff --git a/thirdparty/libwebsockets/win32helpers/getopt_long.c b/thirdparty/libwebsockets/win32helpers/getopt_long.c index 22e5fa8945..6dfccf367d 100644 --- a/thirdparty/libwebsockets/win32helpers/getopt_long.c +++ b/thirdparty/libwebsockets/win32helpers/getopt_long.c @@ -1,240 +1,240 @@ -
-/*
- * Copyright (c) 1987, 1993, 1994, 1996
- * The Regents of the University of California. All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in the
- * documentation and/or other materials provided with the distribution.
- * 3. All advertising materials mentioning features or use of this software
- * must display the following acknowledgement:
- * This product includes software developed by the University of
- * California, Berkeley and its contributors.
- * 4. Neither the name of the University nor the names of its contributors
- * may be used to endorse or promote products derived from this software
- * without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-#include <assert.h>
-#include <errno.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include "getopt.h"
-
-#define lws_ptr_diff(head, tail) \
- ((int)((char *)(head) - (char *)(tail)))
-
-extern int opterr; /* if error message should be printed */
-extern int optind; /* index into parent argv vector */
-extern int optopt; /* character checked for validity */
-extern int optreset; /* reset getopt */
-extern char *optarg; /* argument associated with option */
-
-#define __P(x) x
-#define _DIAGASSERT(x) assert(x)
-
-static char * __progname __P((char *));
-int getopt_internal __P((int, char * const *, const char *));
-
-static char *
-__progname(nargv0)
- char * nargv0;
-{
- char * tmp;
-
- _DIAGASSERT(nargv0 != NULL);
-
- tmp = strrchr(nargv0, '/');
- if (tmp)
- tmp++;
- else
- tmp = nargv0;
- return(tmp);
-}
-
-#define BADCH (int)'?'
-#define BADARG (int)':'
-#define EMSG ""
-
-/*
- * getopt --
- * Parse argc/argv argument vector.
- */
-int
-getopt_internal(nargc, nargv, ostr)
- int nargc;
- char * const *nargv;
- const char *ostr;
-{
- static char *place = EMSG; /* option letter processing */
- char *oli; /* option letter list index */
-
- _DIAGASSERT(nargv != NULL);
- _DIAGASSERT(ostr != NULL);
-
- if (optreset || !*place) { /* update scanning pointer */
- optreset = 0;
- if (optind >= nargc || *(place = nargv[optind]) != '-') {
- place = EMSG;
- return (-1);
- }
- if (place[1] && *++place == '-') { /* found "--" */
- /* ++optind; */
- place = EMSG;
- return (-2);
- }
- } /* option letter okay? */
- if ((optopt = (int)*place++) == (int)':' ||
- !(oli = strchr(ostr, optopt))) {
- /*
- * if the user didn't specify '-' as an option,
- * assume it means -1.
- */
- if (optopt == (int)'-')
- return (-1);
- if (!*place)
- ++optind;
- if (opterr && *ostr != ':')
- (void)fprintf(stderr,
- "%s: illegal option -- %c\n", __progname(nargv[0]), optopt);
- return (BADCH);
- }
- if (*++oli != ':') { /* don't need argument */
- optarg = NULL;
- if (!*place)
- ++optind;
- } else { /* need an argument */
- if (*place) /* no white space */
- optarg = place;
- else if (nargc <= ++optind) { /* no arg */
- place = EMSG;
- if ((opterr) && (*ostr != ':'))
- (void)fprintf(stderr,
- "%s: option requires an argument -- %c\n",
- __progname(nargv[0]), optopt);
- return (BADARG);
- } else /* white space */
- optarg = nargv[optind];
- place = EMSG;
- ++optind;
- }
- return (optopt); /* dump back option letter */
-}
-
-#if 0
-/*
- * getopt --
- * Parse argc/argv argument vector.
- */
-int
-getopt2(nargc, nargv, ostr)
- int nargc;
- char * const *nargv;
- const char *ostr;
-{
- int retval;
-
- if ((retval = getopt_internal(nargc, nargv, ostr)) == -2) {
- retval = -1;
- ++optind;
- }
- return(retval);
-}
-#endif
-
-/*
- * getopt_long --
- * Parse argc/argv argument vector.
- */
-int
-getopt_long(nargc, nargv, options, long_options, index)
- int nargc;
- char ** nargv;
- char * options;
- struct option * long_options;
- int * index;
-{
- int retval;
-
- _DIAGASSERT(nargv != NULL);
- _DIAGASSERT(options != NULL);
- _DIAGASSERT(long_options != NULL);
- /* index may be NULL */
-
- if ((retval = getopt_internal(nargc, nargv, options)) == -2) {
- char *current_argv = nargv[optind++] + 2, *has_equal;
- int i, current_argv_len, match = -1;
-
- if (*current_argv == '\0') {
- return(-1);
- }
- if ((has_equal = strchr(current_argv, '=')) != NULL) {
- current_argv_len = lws_ptr_diff(has_equal, current_argv);
- has_equal++;
- } else
- current_argv_len = (int)strlen(current_argv);
-
- for (i = 0; long_options[i].name; i++) {
- if (strncmp(current_argv, long_options[i].name, current_argv_len))
- continue;
-
- if (strlen(long_options[i].name) == (unsigned)current_argv_len) {
- match = i;
- break;
- }
- if (match == -1)
- match = i;
- }
- if (match != -1) {
- if (long_options[match].has_arg == required_argument ||
- long_options[match].has_arg == optional_argument) {
- if (has_equal)
- optarg = has_equal;
- else
- optarg = nargv[optind++];
- }
- if ((long_options[match].has_arg == required_argument)
- && (optarg == NULL)) {
- /*
- * Missing argument, leading :
- * indicates no error should be generated
- */
- if ((opterr) && (*options != ':'))
- (void)fprintf(stderr,
- "%s: option requires an argument -- %s\n",
- __progname(nargv[0]), current_argv);
- return (BADARG);
- }
- } else { /* No matching argument */
- if ((opterr) && (*options != ':'))
- (void)fprintf(stderr,
- "%s: illegal option -- %s\n", __progname(nargv[0]), current_argv);
- return (BADCH);
- }
- if (long_options[match].flag) {
- *long_options[match].flag = long_options[match].val;
- retval = 0;
- } else
- retval = long_options[match].val;
- if (index)
- *index = match;
- }
- return(retval);
-}
+ +/* + * Copyright (c) 1987, 1993, 1994, 1996 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. All advertising materials mentioning features or use of this software + * must display the following acknowledgement: + * This product includes software developed by the University of + * California, Berkeley and its contributors. + * 4. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ +#include <assert.h> +#include <errno.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include "getopt.h" + +#define lws_ptr_diff(head, tail) \ + ((int)((char *)(head) - (char *)(tail))) + +extern int opterr; /* if error message should be printed */ +extern int optind; /* index into parent argv vector */ +extern int optopt; /* character checked for validity */ +extern int optreset; /* reset getopt */ +extern char *optarg; /* argument associated with option */ + +#define __P(x) x +#define _DIAGASSERT(x) assert(x) + +static char * __progname __P((char *)); +int getopt_internal __P((int, char * const *, const char *)); + +static char * +__progname(nargv0) + char * nargv0; +{ + char * tmp; + + _DIAGASSERT(nargv0 != NULL); + + tmp = strrchr(nargv0, '/'); + if (tmp) + tmp++; + else + tmp = nargv0; + return(tmp); +} + +#define BADCH (int)'?' +#define BADARG (int)':' +#define EMSG "" + +/* + * getopt -- + * Parse argc/argv argument vector. + */ +int +getopt_internal(nargc, nargv, ostr) + int nargc; + char * const *nargv; + const char *ostr; +{ + static char *place = EMSG; /* option letter processing */ + char *oli; /* option letter list index */ + + _DIAGASSERT(nargv != NULL); + _DIAGASSERT(ostr != NULL); + + if (optreset || !*place) { /* update scanning pointer */ + optreset = 0; + if (optind >= nargc || *(place = nargv[optind]) != '-') { + place = EMSG; + return (-1); + } + if (place[1] && *++place == '-') { /* found "--" */ + /* ++optind; */ + place = EMSG; + return (-2); + } + } /* option letter okay? */ + if ((optopt = (int)*place++) == (int)':' || + !(oli = strchr(ostr, optopt))) { + /* + * if the user didn't specify '-' as an option, + * assume it means -1. + */ + if (optopt == (int)'-') + return (-1); + if (!*place) + ++optind; + if (opterr && *ostr != ':') + (void)fprintf(stderr, + "%s: illegal option -- %c\n", __progname(nargv[0]), optopt); + return (BADCH); + } + if (*++oli != ':') { /* don't need argument */ + optarg = NULL; + if (!*place) + ++optind; + } else { /* need an argument */ + if (*place) /* no white space */ + optarg = place; + else if (nargc <= ++optind) { /* no arg */ + place = EMSG; + if ((opterr) && (*ostr != ':')) + (void)fprintf(stderr, + "%s: option requires an argument -- %c\n", + __progname(nargv[0]), optopt); + return (BADARG); + } else /* white space */ + optarg = nargv[optind]; + place = EMSG; + ++optind; + } + return (optopt); /* dump back option letter */ +} + +#if 0 +/* + * getopt -- + * Parse argc/argv argument vector. + */ +int +getopt2(nargc, nargv, ostr) + int nargc; + char * const *nargv; + const char *ostr; +{ + int retval; + + if ((retval = getopt_internal(nargc, nargv, ostr)) == -2) { + retval = -1; + ++optind; + } + return(retval); +} +#endif + +/* + * getopt_long -- + * Parse argc/argv argument vector. + */ +int +getopt_long(nargc, nargv, options, long_options, index) + int nargc; + char ** nargv; + char * options; + struct option * long_options; + int * index; +{ + int retval; + + _DIAGASSERT(nargv != NULL); + _DIAGASSERT(options != NULL); + _DIAGASSERT(long_options != NULL); + /* index may be NULL */ + + if ((retval = getopt_internal(nargc, nargv, options)) == -2) { + char *current_argv = nargv[optind++] + 2, *has_equal; + int i, current_argv_len, match = -1; + + if (*current_argv == '\0') { + return(-1); + } + if ((has_equal = strchr(current_argv, '=')) != NULL) { + current_argv_len = lws_ptr_diff(has_equal, current_argv); + has_equal++; + } else + current_argv_len = (int)strlen(current_argv); + + for (i = 0; long_options[i].name; i++) { + if (strncmp(current_argv, long_options[i].name, current_argv_len)) + continue; + + if (strlen(long_options[i].name) == (unsigned)current_argv_len) { + match = i; + break; + } + if (match == -1) + match = i; + } + if (match != -1) { + if (long_options[match].has_arg == required_argument || + long_options[match].has_arg == optional_argument) { + if (has_equal) + optarg = has_equal; + else + optarg = nargv[optind++]; + } + if ((long_options[match].has_arg == required_argument) + && (optarg == NULL)) { + /* + * Missing argument, leading : + * indicates no error should be generated + */ + if ((opterr) && (*options != ':')) + (void)fprintf(stderr, + "%s: option requires an argument -- %s\n", + __progname(nargv[0]), current_argv); + return (BADARG); + } + } else { /* No matching argument */ + if ((opterr) && (*options != ':')) + (void)fprintf(stderr, + "%s: illegal option -- %s\n", __progname(nargv[0]), current_argv); + return (BADCH); + } + if (long_options[match].flag) { + *long_options[match].flag = long_options[match].val; + retval = 0; + } else + retval = long_options[match].val; + if (index) + *index = match; + } + return(retval); +} diff --git a/thirdparty/libwebsockets/win32helpers/gettimeofday.c b/thirdparty/libwebsockets/win32helpers/gettimeofday.c index 08385c2320..35dd73531d 100644 --- a/thirdparty/libwebsockets/win32helpers/gettimeofday.c +++ b/thirdparty/libwebsockets/win32helpers/gettimeofday.c @@ -1,36 +1,36 @@ -#include <time.h>
-#include <windows.h> //I've omitted context line
-
-#include "gettimeofday.h"
-
-int gettimeofday(struct timeval *tv, struct timezone *tz)
-{
- FILETIME ft;
- unsigned __int64 tmpres = 0;
- static int tzflag;
-
- if (NULL != tv) {
- GetSystemTimeAsFileTime(&ft);
-
- tmpres |= ft.dwHighDateTime;
- tmpres <<= 32;
- tmpres |= ft.dwLowDateTime;
-
- /*converting file time to unix epoch*/
- tmpres /= 10; /*convert into microseconds*/
+#include <time.h> +#include <windows.h> //I've omitted context line + +#include "gettimeofday.h" + +int gettimeofday(struct timeval *tv, struct timezone *tz) +{ + FILETIME ft; + unsigned __int64 tmpres = 0; + static int tzflag; + + if (NULL != tv) { + GetSystemTimeAsFileTime(&ft); + + tmpres |= ft.dwHighDateTime; + tmpres <<= 32; + tmpres |= ft.dwLowDateTime; + + /*converting file time to unix epoch*/ + tmpres /= 10; /*convert into microseconds*/ tmpres -= DELTA_EPOCH_IN_MICROSECS; - tv->tv_sec = (long)(tmpres / 1000000UL);
- tv->tv_usec = (long)(tmpres % 1000000UL);
- }
-
- if (NULL != tz) {
- if (!tzflag) {
- _tzset();
- tzflag++;
- }
- tz->tz_minuteswest = _timezone / 60;
- tz->tz_dsttime = _daylight;
- }
-
- return 0;
-}
+ tv->tv_sec = (long)(tmpres / 1000000UL); + tv->tv_usec = (long)(tmpres % 1000000UL); + } + + if (NULL != tz) { + if (!tzflag) { + _tzset(); + tzflag++; + } + tz->tz_minuteswest = _timezone / 60; + tz->tz_dsttime = _daylight; + } + + return 0; +} diff --git a/thirdparty/misc/easing_equations.cpp b/thirdparty/misc/easing_equations.cpp new file mode 100644 index 0000000000..bc84564b19 --- /dev/null +++ b/thirdparty/misc/easing_equations.cpp @@ -0,0 +1,308 @@ +/** + * Adapted from Penner Easing equations' C++ port. + * Source: https://github.com/jesusgollonet/ofpennereasing + * License: BSD-3-clause + */ + +#include "scene/animation/tween.h" + +const real_t pi = 3.1415926535898; + +/////////////////////////////////////////////////////////////////////////// +// linear +/////////////////////////////////////////////////////////////////////////// +namespace linear { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return c * t / d + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + return c * t / d + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + return c * t / d + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return c * t / d + b; +} +}; // namespace linear +/////////////////////////////////////////////////////////////////////////// +// sine +/////////////////////////////////////////////////////////////////////////// +namespace sine { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return -c * cos(t / d * (pi / 2)) + c + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + return c * sin(t / d * (pi / 2)) + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + return -c / 2 * (cos(pi * t / d) - 1) + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace sine +/////////////////////////////////////////////////////////////////////////// +// quint +/////////////////////////////////////////////////////////////////////////// +namespace quint { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return c * pow(t / d, 5) + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + return c * (pow(t / d - 1, 5) + 1) + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + t = t / d * 2; + if (t < 1) return c / 2 * pow(t, 5) + b; + return c / 2 * (pow(t - 2, 5) + 2) + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace quint +/////////////////////////////////////////////////////////////////////////// +// quart +/////////////////////////////////////////////////////////////////////////// +namespace quart { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return c * pow(t / d, 4) + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + return -c * (pow(t / d - 1, 4) - 1) + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + t = t / d * 2; + if (t < 1) return c / 2 * pow(t, 4) + b; + return -c / 2 * (pow(t - 2, 4) - 2) + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace quart +/////////////////////////////////////////////////////////////////////////// +// quad +/////////////////////////////////////////////////////////////////////////// +namespace quad { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return c * pow(t / d, 2) + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + t = t / d; + return -c * t * (t - 2) + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + t = t / d * 2; + if (t < 1) return c / 2 * pow(t, 2) + b; + return -c / 2 * ((t - 1) * (t - 3) - 1) + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace quad +/////////////////////////////////////////////////////////////////////////// +// expo +/////////////////////////////////////////////////////////////////////////// +namespace expo { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + if (t == 0) return b; + return c * pow(2, 10 * (t / d - 1)) + b - c * 0.001; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + if (t == d) return b + c; + return c * 1.001 * (-pow(2, -10 * t / d) + 1) + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + if (t == 0) return b; + if (t == d) return b + c; + t = t / d * 2; + if (t < 1) return c / 2 * pow(2, 10 * (t - 1)) + b - c * 0.0005; + return c / 2 * 1.0005 * (-pow(2, -10 * (t - 1)) + 2) + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace expo +/////////////////////////////////////////////////////////////////////////// +// elastic +/////////////////////////////////////////////////////////////////////////// +namespace elastic { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + if (t == 0) return b; + if ((t /= d) == 1) return b + c; + float p = d * 0.3f; + float a = c; + float s = p / 4; + float postFix = a * pow(2, 10 * (t -= 1)); // this is a fix, again, with post-increment operators + return -(postFix * sin((t * d - s) * (2 * pi) / p)) + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + if (t == 0) return b; + if ((t /= d) == 1) return b + c; + float p = d * 0.3f; + float a = c; + float s = p / 4; + return (a * pow(2, -10 * t) * sin((t * d - s) * (2 * pi) / p) + c + b); +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + if (t == 0) return b; + if ((t /= d / 2) == 2) return b + c; + float p = d * (0.3f * 1.5f); + float a = c; + float s = p / 4; + + if (t < 1) { + float postFix = a * pow(2, 10 * (t -= 1)); // postIncrement is evil + return -0.5f * (postFix * sin((t * d - s) * (2 * pi) / p)) + b; + } + float postFix = a * pow(2, -10 * (t -= 1)); // postIncrement is evil + return postFix * sin((t * d - s) * (2 * pi) / p) * 0.5f + c + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace elastic +/////////////////////////////////////////////////////////////////////////// +// cubic +/////////////////////////////////////////////////////////////////////////// +namespace cubic { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return c * (t /= d) * t * t + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + t = t / d - 1; + return c * (t * t * t + 1) + b; +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + if ((t /= d / 2) < 1) return c / 2 * t * t * t + b; + return c / 2 * ((t -= 2) * t * t + 2) + b; +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace cubic +/////////////////////////////////////////////////////////////////////////// +// circ +/////////////////////////////////////////////////////////////////////////// +namespace circ { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return -c * (sqrt(1 - (t /= d) * t) - 1) + b; // TODO: ehrich: operation with t is undefined +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + return c * sqrt(1 - (t = t / d - 1) * t) + b; // TODO: ehrich: operation with t is undefined +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + if ((t /= d / 2) < 1) return -c / 2 * (sqrt(1 - t * t) - 1) + b; + return c / 2 * (sqrt(1 - t * (t -= 2)) + 1) + b; // TODO: ehrich: operation with t is undefined +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace circ +/////////////////////////////////////////////////////////////////////////// +// bounce +/////////////////////////////////////////////////////////////////////////// +namespace bounce { +static real_t out(real_t t, real_t b, real_t c, real_t d); + +static real_t in(real_t t, real_t b, real_t c, real_t d) { + return c - out(d - t, 0, c, d) + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + if ((t /= d) < (1 / 2.75f)) { + return c * (7.5625f * t * t) + b; + } else if (t < (2 / 2.75f)) { + float postFix = t -= (1.5f / 2.75f); + return c * (7.5625f * (postFix)*t + .75f) + b; + } else if (t < (2.5 / 2.75)) { + float postFix = t -= (2.25f / 2.75f); + return c * (7.5625f * (postFix)*t + .9375f) + b; + } else { + float postFix = t -= (2.625f / 2.75f); + return c * (7.5625f * (postFix)*t + .984375f) + b; + } +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? in(t * 2, b, c / 2, d) : out((t * 2) - d, b + c / 2, c / 2, d); +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace bounce +/////////////////////////////////////////////////////////////////////////// +// back +/////////////////////////////////////////////////////////////////////////// +namespace back { +static real_t in(real_t t, real_t b, real_t c, real_t d) { + float s = 1.70158f; + float postFix = t /= d; + return c * (postFix)*t * ((s + 1) * t - s) + b; +} + +static real_t out(real_t t, real_t b, real_t c, real_t d) { + float s = 1.70158f; + return c * ((t = t / d - 1) * t * ((s + 1) * t + s) + 1) + b; // TODO: ehrich: operation with t is undefined +} + +static real_t in_out(real_t t, real_t b, real_t c, real_t d) { + float s = 1.70158f; + if ((t /= d / 2) < 1) return c / 2 * (t * t * (((s *= (1.525f)) + 1) * t - s)) + b; // TODO: ehrich: operation with s is undefined + float postFix = t -= 2; + return c / 2 * ((postFix)*t * (((s *= (1.525f)) + 1) * t + s) + 2) + b; // TODO: ehrich: operation with s is undefined +} + +static real_t out_in(real_t t, real_t b, real_t c, real_t d) { + return (t < d / 2) ? out(t * 2, b, c / 2, d) : in((t * 2) - d, b + c / 2, c / 2, d); +} +}; // namespace back + +Tween::interpolater Tween::interpolaters[Tween::TRANS_COUNT][Tween::EASE_COUNT] = { + { &linear::in, &linear::out, &linear::in_out, &linear::out_in }, + { &sine::in, &sine::out, &sine::in_out, &sine::out_in }, + { &quint::in, &quint::out, &quint::in_out, &quint::out_in }, + { &quart::in, &quart::out, &quart::in_out, &quart::out_in }, + { &quad::in, &quad::out, &quad::in_out, &quad::out_in }, + { &expo::in, &expo::out, &expo::in_out, &expo::out_in }, + { &elastic::in, &elastic::out, &elastic::in_out, &elastic::out_in }, + { &cubic::in, &cubic::out, &cubic::in_out, &cubic::out_in }, + { &circ::in, &circ::out, &circ::in_out, &circ::out_in }, + { &bounce::in, &bounce::out, &bounce::in_out, &bounce::out_in }, + { &back::in, &back::out, &back::in_out, &back::out_in }, +}; + +real_t Tween::_run_equation(TransitionType p_trans_type, EaseType p_ease_type, real_t t, real_t b, real_t c, real_t d) { + + interpolater cb = interpolaters[p_trans_type][p_ease_type]; + ERR_FAIL_COND_V(cb == NULL, b); + return cb(t, b, c, d); +} diff --git a/thirdparty/misc/stb_vorbis.h b/thirdparty/misc/stb_vorbis.h new file mode 100644 index 0000000000..357efcd5fc --- /dev/null +++ b/thirdparty/misc/stb_vorbis.h @@ -0,0 +1,2 @@ +#define STB_VORBIS_HEADER_ONLY +#include "stb_vorbis.c" diff --git a/thirdparty/squish/Add-Decompress-Bc5-to-Squish.patch b/thirdparty/squish/Add-Decompress-Bc5-to-Squish.patch deleted file mode 100644 index 1e06a8d318..0000000000 --- a/thirdparty/squish/Add-Decompress-Bc5-to-Squish.patch +++ /dev/null @@ -1,143 +0,0 @@ -From 7b64cc4c8b0be0443741483bf65909f5140179c0 Mon Sep 17 00:00:00 2001 -From: Orkun <orkuntezerm@gmail.com> -Date: Sun, 19 Nov 2017 02:24:31 +0300 -Subject: [PATCH] Fix #12220: Add Decompress Bc5 to Squish - -This Commit fixes the corrupted file preview described in #12220. -Added DecompressColourBc5 function to squish. ---- - thirdparty/squish/colourblock.cpp | 85 +++++++++++++++++++++++++++++++++++++++ - thirdparty/squish/colourblock.h | 3 ++ - thirdparty/squish/squish.cpp | 8 +++- - 3 files changed, 95 insertions(+), 1 deletion(-) - -diff --git a/thirdparty/squish/colourblock.cpp b/thirdparty/squish/colourblock.cpp -index af8b98036..3de46382c 100644 ---- a/thirdparty/squish/colourblock.cpp -+++ b/thirdparty/squish/colourblock.cpp -@@ -211,4 +211,89 @@ void DecompressColour( u8* rgba, void const* block, bool isDxt1 ) - } - } - -+// -- Godot start -- -+void DecompressColourBc5( u8* rgba, void const* block) -+{ -+ // get the block bytes -+ u8 const* bytes = reinterpret_cast< u8 const* >( block ); -+ -+ // unpack the endpoints -+ u8 codes[16]; -+ int red_0 = bytes[0]; -+ int red_1 = bytes[1]; -+ -+ codes[0] = red_0; -+ codes[1] = red_1; -+ codes[6] = 0.0f; -+ codes[7] = 1.0f; -+ // generate the midpoints -+ if(red_0 > red_1) -+ { -+ for( int i = 2; i < 8; ++i ) -+ { -+ codes[i] = ((8-i)*red_0 + (i-1)*red_1)/7; -+ } -+ } -+ else -+ { -+ for( int i = 2; i < 6; ++i ) -+ { -+ codes[i] = ((6-i)*red_0 + (i-1)*red_1)/5; -+ } -+ } -+ -+ int green_0 = bytes[8]; -+ int green_1 = bytes[9]; -+ -+ codes[0 + 8] = green_0; -+ codes[1 + 8] = green_1; -+ codes[6 + 8] = 0.0f; -+ codes[7 + 8] = 1.0f; -+ // generate the midpoints -+ if(green_0 > green_1) -+ { -+ for( int i = 2; i < 8; ++i ) -+ { -+ codes[i + 8] = ((8-i)*green_0 + (i-1)*green_1)/7; -+ } -+ } -+ else -+ { -+ for( int i = 2; i < 6; ++i ) -+ { -+ codes[i + 8] = ((6-i)*green_0 + (i-1)*green_1)/5; -+ } -+ } -+ -+ u8 indices[32]; -+ for( int i = 0; i < 4; ++i ) -+ { -+ u8 packed = bytes[2 + i]; -+ u8* red_ind = indices + 4*i; -+ -+ red_ind[0] = packed & 0x3; -+ red_ind[1] = ( packed >> 2 ) & 0x3; -+ red_ind[2] = ( packed >> 4 ) & 0x3; -+ red_ind[3] = ( packed >> 6 ) & 0x3; -+ -+ packed = bytes[8 + i]; -+ u8* green_ind = indices + 4*i + 16; -+ green_ind[0] = packed & 0x3; -+ green_ind[1] = ( packed >> 2 ) & 0x3; -+ green_ind[2] = ( packed >> 4 ) & 0x3; -+ green_ind[3] = ( packed >> 6 ) & 0x3; -+ } -+ -+ // store out the colours -+ for( int i = 0; i < 16; ++i ) -+ { -+ rgba[4*i] = codes[indices[i]]; -+ rgba[4*i +1] = codes[indices[i + 16] + 8]; -+ rgba[4*i +2] = 0; -+ rgba[4*i +3] = 255; -+ } -+} -+// -- GODOT end -- -+ -+ - } // namespace squish -diff --git a/thirdparty/squish/colourblock.h b/thirdparty/squish/colourblock.h -index fee2cd7c5..3cb9b7e3b 100644 ---- a/thirdparty/squish/colourblock.h -+++ b/thirdparty/squish/colourblock.h -@@ -35,6 +35,9 @@ void WriteColourBlock3( Vec3::Arg start, Vec3::Arg end, u8 const* indices, void* - void WriteColourBlock4( Vec3::Arg start, Vec3::Arg end, u8 const* indices, void* block ); - - void DecompressColour( u8* rgba, void const* block, bool isDxt1 ); -+// -- GODOT start -- -+void DecompressColourBc5( u8* rgba, void const* block ); -+// -- GODOT end -- - - } // namespace squish - -diff --git a/thirdparty/squish/squish.cpp b/thirdparty/squish/squish.cpp -index 1d22a64ad..fd11a147d 100644 ---- a/thirdparty/squish/squish.cpp -+++ b/thirdparty/squish/squish.cpp -@@ -135,7 +135,13 @@ void Decompress( u8* rgba, void const* block, int flags ) - colourBlock = reinterpret_cast< u8 const* >( block ) + 8; - - // decompress colour -- DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 ); -+ // -- GODOT start -- -+ //DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 ); -+ if(( flags & ( kBc5 ) ) != 0) -+ DecompressColourBc5( rgba, colourBlock); -+ else -+ DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 ); -+ // -- GODOT end -- - - // decompress alpha separately if necessary - if( ( flags & kDxt3 ) != 0 ) --- -2.13.6 - diff --git a/thirdparty/squish/colourblock.cpp b/thirdparty/squish/colourblock.cpp index 3de46382c0..3d87adaa77 100644 --- a/thirdparty/squish/colourblock.cpp +++ b/thirdparty/squish/colourblock.cpp @@ -24,6 +24,9 @@ -------------------------------------------------------------------------- */ #include "colourblock.h" +// -- Godot start -- +#include "alpha.h" +// -- Godot end -- namespace squish { @@ -214,83 +217,17 @@ void DecompressColour( u8* rgba, void const* block, bool isDxt1 ) // -- Godot start -- void DecompressColourBc5( u8* rgba, void const* block) { - // get the block bytes - u8 const* bytes = reinterpret_cast< u8 const* >( block ); - - // unpack the endpoints - u8 codes[16]; - int red_0 = bytes[0]; - int red_1 = bytes[1]; - - codes[0] = red_0; - codes[1] = red_1; - codes[6] = 0.0f; - codes[7] = 1.0f; - // generate the midpoints - if(red_0 > red_1) - { - for( int i = 2; i < 8; ++i ) - { - codes[i] = ((8-i)*red_0 + (i-1)*red_1)/7; - } - } - else - { - for( int i = 2; i < 6; ++i ) - { - codes[i] = ((6-i)*red_0 + (i-1)*red_1)/5; - } - } - - int green_0 = bytes[8]; - int green_1 = bytes[9]; - - codes[0 + 8] = green_0; - codes[1 + 8] = green_1; - codes[6 + 8] = 0.0f; - codes[7 + 8] = 1.0f; - // generate the midpoints - if(green_0 > green_1) - { - for( int i = 2; i < 8; ++i ) - { - codes[i + 8] = ((8-i)*green_0 + (i-1)*green_1)/7; - } - } - else - { - for( int i = 2; i < 6; ++i ) - { - codes[i + 8] = ((6-i)*green_0 + (i-1)*green_1)/5; - } - } - - u8 indices[32]; - for( int i = 0; i < 4; ++i ) - { - u8 packed = bytes[2 + i]; - u8* red_ind = indices + 4*i; - - red_ind[0] = packed & 0x3; - red_ind[1] = ( packed >> 2 ) & 0x3; - red_ind[2] = ( packed >> 4 ) & 0x3; - red_ind[3] = ( packed >> 6 ) & 0x3; - - packed = bytes[8 + i]; - u8* green_ind = indices + 4*i + 16; - green_ind[0] = packed & 0x3; - green_ind[1] = ( packed >> 2 ) & 0x3; - green_ind[2] = ( packed >> 4 ) & 0x3; - green_ind[3] = ( packed >> 6 ) & 0x3; + void const* rblock = block; + void const* gblock = reinterpret_cast< u8 const* >( block ) + 8; + DecompressAlphaDxt5(rgba,rblock); + for ( int i = 0; i < 16; ++i ) { + rgba[i*4] = rgba[i*4 + 3]; } - - // store out the colours - for( int i = 0; i < 16; ++i ) - { - rgba[4*i] = codes[indices[i]]; - rgba[4*i +1] = codes[indices[i + 16] + 8]; - rgba[4*i +2] = 0; - rgba[4*i +3] = 255; + DecompressAlphaDxt5(rgba,gblock); + for ( int i = 0; i < 16; ++i ) { + rgba[i*4+1] = rgba[i*4 + 3]; + rgba[i*4 + 2] = 0; + rgba[i*4 + 3] = 255; } } // -- GODOT end -- diff --git a/thirdparty/squish/godot-changes.patch b/thirdparty/squish/godot-changes.patch new file mode 100644 index 0000000000..ef7bafb4b4 --- /dev/null +++ b/thirdparty/squish/godot-changes.patch @@ -0,0 +1,102 @@ +diff --git a/thirdparty/squish/colourblock.cpp b/thirdparty/squish/colourblock.cpp +index af8b98036..3d87adaa7 100644 +--- a/thirdparty/squish/colourblock.cpp ++++ b/thirdparty/squish/colourblock.cpp +@@ -24,6 +24,9 @@ + -------------------------------------------------------------------------- */ + + #include "colourblock.h" ++// -- Godot start -- ++#include "alpha.h" ++// -- Godot end -- + + namespace squish { + +@@ -211,4 +214,23 @@ void DecompressColour( u8* rgba, void const* block, bool isDxt1 ) + } + } + ++// -- Godot start -- ++void DecompressColourBc5( u8* rgba, void const* block) ++{ ++ void const* rblock = block; ++ void const* gblock = reinterpret_cast< u8 const* >( block ) + 8; ++ DecompressAlphaDxt5(rgba,rblock); ++ for ( int i = 0; i < 16; ++i ) { ++ rgba[i*4] = rgba[i*4 + 3]; ++ } ++ DecompressAlphaDxt5(rgba,gblock); ++ for ( int i = 0; i < 16; ++i ) { ++ rgba[i*4+1] = rgba[i*4 + 3]; ++ rgba[i*4 + 2] = 0; ++ rgba[i*4 + 3] = 255; ++ } ++} ++// -- GODOT end -- ++ ++ + } // namespace squish +diff --git a/thirdparty/squish/colourblock.h b/thirdparty/squish/colourblock.h +index fee2cd7c5..3cb9b7e3b 100644 +--- a/thirdparty/squish/colourblock.h ++++ b/thirdparty/squish/colourblock.h +@@ -35,6 +35,9 @@ void WriteColourBlock3( Vec3::Arg start, Vec3::Arg end, u8 const* indices, void* + void WriteColourBlock4( Vec3::Arg start, Vec3::Arg end, u8 const* indices, void* block ); + + void DecompressColour( u8* rgba, void const* block, bool isDxt1 ); ++// -- GODOT start -- ++void DecompressColourBc5( u8* rgba, void const* block ); ++// -- GODOT end -- + + } // namespace squish + +diff --git a/thirdparty/squish/config.h b/thirdparty/squish/config.h +index 92edefe96..05f8d7259 100644 +--- a/thirdparty/squish/config.h ++++ b/thirdparty/squish/config.h +@@ -32,6 +32,26 @@ + #endif + + // Set to 1 or 2 when building squish to use SSE or SSE2 instructions. ++// -- GODOT start -- ++#ifdef _MSC_VER ++ #if defined(_M_IX86_FP) ++ #if _M_IX86_FP >= 2 ++ #define SQUISH_USE_SSE 2 ++ #elif _M_IX86_FP >= 1 ++ #define SQUISH_USE_SSE 1 ++ #endif ++ #elif defined(_M_X64) ++ #define SQUISH_USE_SSE 2 ++ #endif ++#else ++ #if defined(__SSE2__) ++ #define SQUISH_USE_SSE 2 ++ #elif defined(__SSE__) ++ #define SQUISH_USE_SSE 1 ++ #endif ++#endif ++// -- GODOT end -- ++ + #ifndef SQUISH_USE_SSE + #define SQUISH_USE_SSE 0 + #endif +diff --git a/thirdparty/squish/squish.cpp b/thirdparty/squish/squish.cpp +index 1d22a64ad..fd11a147d 100644 +--- a/thirdparty/squish/squish.cpp ++++ b/thirdparty/squish/squish.cpp +@@ -135,7 +135,13 @@ void Decompress( u8* rgba, void const* block, int flags ) + colourBlock = reinterpret_cast< u8 const* >( block ) + 8; + + // decompress colour +- DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 ); ++ // -- GODOT start -- ++ //DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 ); ++ if(( flags & ( kBc5 ) ) != 0) ++ DecompressColourBc5( rgba, colourBlock); ++ else ++ DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 ); ++ // -- GODOT end -- + + // decompress alpha separately if necessary + if( ( flags & kDxt3 ) != 0 ) diff --git a/thirdparty/tinyexr/tinyexr.h b/thirdparty/tinyexr/tinyexr.h index 990c8ee142..b3a7ee00c2 100644 --- a/thirdparty/tinyexr/tinyexr.h +++ b/thirdparty/tinyexr/tinyexr.h @@ -116,6 +116,8 @@ extern "C" { #define TINYEXR_ERROR_UNSUPPORTED_FORMAT (-7) #define TINYEXR_ERROR_INVALID_HEADER (-8) #define TINYEXR_ERROR_UNSUPPORTED_FEATURE (-9) +#define TINYEXR_ERROR_CANT_WRITE_FILE (-10) +#define TINYEXR_ERROR_SERIALZATION_FAILED (-11) // @note { OpenEXR file format: http://www.openexr.com/openexrfilelayout.pdf } @@ -279,9 +281,12 @@ extern int LoadEXR(float **out_rgba, int *width, int *height, // Save image as fp16(HALF) format when `save_as_fp16` is positive non-zero // value. // Save image as fp32(FLOAT) format when `save_as_fp16` is 0. +// Use ZIP compression by default. +// Returns negative value and may set error string in `err` when there's an +// error extern int SaveEXR(const float *data, const int width, const int height, const int components, const int save_as_fp16, - const char *filename); + const char *filename, const char **err); // Initialize EXRHeader struct extern void InitEXRHeader(EXRHeader *exr_header); @@ -400,9 +405,9 @@ extern int SaveEXRImageToFile(const EXRImage *image, // Saves multi-channel, single-frame OpenEXR image to a memory. // Image is compressed using EXRImage.compression value. -// Return the number of bytes if succes. -// Returns negative value and may set error string in `err` when there's an -// error +// Return the number of bytes if success. +// Return zero and will set error string in `err` when there's an +// error. // When there was an error message, Application must free `err` with // FreeEXRErrorMessage() extern size_t SaveEXRImageToMemory(const EXRImage *image, @@ -524,15 +529,23 @@ namespace miniz { #if __has_warning("-Wcomma") #pragma clang diagnostic ignored "-Wcomma" #endif + #if __has_warning("-Wmacro-redefined") #pragma clang diagnostic ignored "-Wmacro-redefined" #endif + #if __has_warning("-Wcast-qual") #pragma clang diagnostic ignored "-Wcast-qual" #endif + #if __has_warning("-Wzero-as-null-pointer-constant") #pragma clang diagnostic ignored "-Wzero-as-null-pointer-constant" #endif + +#if __has_warning("-Wtautological-constant-compare") +#pragma clang diagnostic ignored "-Wtautological-constant-compare" +#endif + #endif /* miniz.c v1.15 - public domain deflate/inflate, zlib-subset, ZIP @@ -2518,10 +2531,10 @@ tinfl_status tinfl_decompress(tinfl_decompressor *r, tinfl_status status = TINFL_STATUS_FAILED; mz_uint32 num_bits, dist, counter, num_extra; tinfl_bit_buf_t bit_buf; - const mz_uint8 *pIn_buf_cur = pIn_buf_next, *const pIn_buf_end = - pIn_buf_next + *pIn_buf_size; - mz_uint8 *pOut_buf_cur = pOut_buf_next, *const pOut_buf_end = - pOut_buf_next + *pOut_buf_size; + const mz_uint8 *pIn_buf_cur = pIn_buf_next, + *const pIn_buf_end = pIn_buf_next + *pIn_buf_size; + mz_uint8 *pOut_buf_cur = pOut_buf_next, + *const pOut_buf_end = pOut_buf_next + *pOut_buf_size; size_t out_buf_size_mask = (decomp_flags & TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF) ? (size_t)-1 @@ -2938,9 +2951,8 @@ void *tinfl_decompress_mem_to_heap(const void *pSrc_buf, size_t src_buf_len, tinfl_status status = tinfl_decompress( &decomp, (const mz_uint8 *)pSrc_buf + src_buf_ofs, &src_buf_size, (mz_uint8 *)pBuf, pBuf ? (mz_uint8 *)pBuf + *pOut_len : NULL, - &dst_buf_size, - (flags & ~TINFL_FLAG_HAS_MORE_INPUT) | - TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF); + &dst_buf_size, (flags & ~TINFL_FLAG_HAS_MORE_INPUT) | + TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF); if ((status < 0) || (status == TINFL_STATUS_NEEDS_MORE_INPUT)) { MZ_FREE(pBuf); *pOut_len = 0; @@ -2993,8 +3005,9 @@ int tinfl_decompress_mem_to_callback(const void *pIn_buf, size_t *pIn_buf_size, tinfl_status status = tinfl_decompress(&decomp, (const mz_uint8 *)pIn_buf + in_buf_ofs, &in_buf_size, pDict, pDict + dict_ofs, &dst_buf_size, - (flags & ~(TINFL_FLAG_HAS_MORE_INPUT | - TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF))); + (flags & + ~(TINFL_FLAG_HAS_MORE_INPUT | + TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF))); in_buf_ofs += in_buf_size; if ((dst_buf_size) && (!(*pPut_buf_func)(pDict + dict_ofs, (int)dst_buf_size, pPut_buf_user))) @@ -3119,9 +3132,7 @@ static const mz_uint8 s_tdefl_large_dist_extra[128] = { // Radix sorts tdefl_sym_freq[] array by 16-bit key m_key. Returns ptr to sorted // values. -typedef struct { - mz_uint16 m_key, m_sym_index; -} tdefl_sym_freq; +typedef struct { mz_uint16 m_key, m_sym_index; } tdefl_sym_freq; static tdefl_sym_freq *tdefl_radix_sort_syms(mz_uint num_syms, tdefl_sym_freq *pSyms0, tdefl_sym_freq *pSyms1) { @@ -5265,10 +5276,9 @@ mz_bool mz_zip_reader_file_stat(mz_zip_archive *pZip, mz_uint file_index, n = MZ_READ_LE16(p + MZ_ZIP_CDH_COMMENT_LEN_OFS); n = MZ_MIN(n, MZ_ZIP_MAX_ARCHIVE_FILE_COMMENT_SIZE - 1); pStat->m_comment_size = n; - memcpy(pStat->m_comment, - p + MZ_ZIP_CENTRAL_DIR_HEADER_SIZE + - MZ_READ_LE16(p + MZ_ZIP_CDH_FILENAME_LEN_OFS) + - MZ_READ_LE16(p + MZ_ZIP_CDH_EXTRA_LEN_OFS), + memcpy(pStat->m_comment, p + MZ_ZIP_CENTRAL_DIR_HEADER_SIZE + + MZ_READ_LE16(p + MZ_ZIP_CDH_FILENAME_LEN_OFS) + + MZ_READ_LE16(p + MZ_ZIP_CDH_EXTRA_LEN_OFS), n); pStat->m_comment[n] = '\0'; @@ -10087,9 +10097,10 @@ static bool DecodePixelData(/* out */ unsigned char **out_images, unsigned short *outLine = reinterpret_cast<unsigned short *>(out_images[c]); if (line_order == 0) { - outLine += (y + v) * x_stride; + outLine += (size_t(y) + v) * size_t(x_stride); } else { - outLine += (height - 1 - (y + v)) * x_stride; + outLine += + (size_t(height) - 1 - (size_t(y) + v)) * size_t(x_stride); } for (int u = 0; u < width; u++) { @@ -10105,9 +10116,10 @@ static bool DecodePixelData(/* out */ unsigned char **out_images, } else if (requested_pixel_types[c] == TINYEXR_PIXELTYPE_FLOAT) { float *outLine = reinterpret_cast<float *>(out_images[c]); if (line_order == 0) { - outLine += (y + v) * x_stride; + outLine += (size_t(y) + v) * size_t(x_stride); } else { - outLine += (height - 1 - (y + v)) * x_stride; + outLine += + (size_t(height) - 1 - (size_t(y) + v)) * size_t(x_stride); } if (reinterpret_cast<const unsigned char *>(line_ptr + width) > @@ -10140,9 +10152,10 @@ static bool DecodePixelData(/* out */ unsigned char **out_images, float *outLine = reinterpret_cast<float *>(out_images[c]); if (line_order == 0) { - outLine += (y + v) * x_stride; + outLine += (size_t(y) + v) * size_t(x_stride); } else { - outLine += (height - 1 - (y + v)) * x_stride; + outLine += + (size_t(height) - 1 - (size_t(y) + v)) * size_t(x_stride); } if (reinterpret_cast<const unsigned char *>(line_ptr + width) > @@ -10167,9 +10180,10 @@ static bool DecodePixelData(/* out */ unsigned char **out_images, unsigned int *outLine = reinterpret_cast<unsigned int *>(out_images[c]); if (line_order == 0) { - outLine += (y + v) * x_stride; + outLine += (size_t(y) + v) * size_t(x_stride); } else { - outLine += (height - 1 - (y + v)) * x_stride; + outLine += + (size_t(height) - 1 - (size_t(y) + v)) * size_t(x_stride); } for (int u = 0; u < width; u++) { @@ -11133,21 +11147,53 @@ int LoadEXR(float **out_rgba, int *width, int *height, const char *filename, } } - if ((idxA == 0) && (idxR == -1) && (idxG == -1) && (idxB == -1)) { - // Alpha channel only. + if (exr_header.num_channels == 1) { + // Grayscale channel only. - if (exr_header.tiled) { - // todo.implement this - } (*out_rgba) = reinterpret_cast<float *>( malloc(4 * sizeof(float) * static_cast<size_t>(exr_image.width) * static_cast<size_t>(exr_image.height))); - for (int i = 0; i < exr_image.width * exr_image.height; i++) { - const float val = reinterpret_cast<float **>(exr_image.images)[0][i]; - (*out_rgba)[4 * i + 0] = val; - (*out_rgba)[4 * i + 1] = val; - (*out_rgba)[4 * i + 2] = val; - (*out_rgba)[4 * i + 3] = val; + + if (exr_header.tiled) { + // todo.implement this + + for (int it = 0; it < exr_image.num_tiles; it++) { + for (int j = 0; j < exr_header.tile_size_y; j++) { + for (int i = 0; i < exr_header.tile_size_x; i++) { + const int ii = + exr_image.tiles[it].offset_x * exr_header.tile_size_x + i; + const int jj = + exr_image.tiles[it].offset_y * exr_header.tile_size_y + j; + const int idx = ii + jj * exr_image.width; + + // out of region check. + if (ii >= exr_image.width) { + continue; + } + if (jj >= exr_image.height) { + continue; + } + const int srcIdx = i + j * exr_header.tile_size_x; + unsigned char **src = exr_image.tiles[it].images; + (*out_rgba)[4 * idx + 0] = + reinterpret_cast<float **>(src)[0][srcIdx]; + (*out_rgba)[4 * idx + 1] = + reinterpret_cast<float **>(src)[0][srcIdx]; + (*out_rgba)[4 * idx + 2] = + reinterpret_cast<float **>(src)[0][srcIdx]; + (*out_rgba)[4 * idx + 3] = + reinterpret_cast<float **>(src)[0][srcIdx]; + } + } + } + } else { + for (int i = 0; i < exr_image.width * exr_image.height; i++) { + const float val = reinterpret_cast<float **>(exr_image.images)[0][i]; + (*out_rgba)[4 * i + 0] = val; + (*out_rgba)[4 * i + 1] = val; + (*out_rgba)[4 * i + 2] = val; + (*out_rgba)[4 * i + 3] = val; + } } } else { // Assume RGB(A) @@ -11179,7 +11225,7 @@ int LoadEXR(float **out_rgba, int *width, int *height, const char *filename, static_cast<size_t>(exr_image.height))); if (exr_header.tiled) { for (int it = 0; it < exr_image.num_tiles; it++) { - for (int j = 0; j < exr_header.tile_size_y; j++) + for (int j = 0; j < exr_header.tile_size_y; j++) { for (int i = 0; i < exr_header.tile_size_x; i++) { const int ii = exr_image.tiles[it].offset_x * exr_header.tile_size_x + i; @@ -11209,6 +11255,7 @@ int LoadEXR(float **out_rgba, int *width, int *height, const char *filename, (*out_rgba)[4 * idx + 3] = 1.0; } } + } } } else { for (int i = 0; i < exr_image.width * exr_image.height; i++) { @@ -11356,18 +11403,53 @@ int LoadEXRFromMemory(float **out_rgba, int *width, int *height, malloc(4 * sizeof(float) * static_cast<size_t>(exr_image.width) * static_cast<size_t>(exr_image.height))); - for (int i = 0; i < exr_image.width * exr_image.height; i++) { - (*out_rgba)[4 * i + 0] = - reinterpret_cast<float **>(exr_image.images)[idxR][i]; - (*out_rgba)[4 * i + 1] = - reinterpret_cast<float **>(exr_image.images)[idxG][i]; - (*out_rgba)[4 * i + 2] = - reinterpret_cast<float **>(exr_image.images)[idxB][i]; - if (idxA != -1) { - (*out_rgba)[4 * i + 3] = - reinterpret_cast<float **>(exr_image.images)[idxA][i]; - } else { - (*out_rgba)[4 * i + 3] = 1.0; + if (exr_header.tiled) { + for (int it = 0; it < exr_image.num_tiles; it++) { + for (int j = 0; j < exr_header.tile_size_y; j++) + for (int i = 0; i < exr_header.tile_size_x; i++) { + const int ii = + exr_image.tiles[it].offset_x * exr_header.tile_size_x + i; + const int jj = + exr_image.tiles[it].offset_y * exr_header.tile_size_y + j; + const int idx = ii + jj * exr_image.width; + + // out of region check. + if (ii >= exr_image.width) { + continue; + } + if (jj >= exr_image.height) { + continue; + } + const int srcIdx = i + j * exr_header.tile_size_x; + unsigned char **src = exr_image.tiles[it].images; + (*out_rgba)[4 * idx + 0] = + reinterpret_cast<float **>(src)[idxR][srcIdx]; + (*out_rgba)[4 * idx + 1] = + reinterpret_cast<float **>(src)[idxG][srcIdx]; + (*out_rgba)[4 * idx + 2] = + reinterpret_cast<float **>(src)[idxB][srcIdx]; + if (idxA != -1) { + (*out_rgba)[4 * idx + 3] = + reinterpret_cast<float **>(src)[idxA][srcIdx]; + } else { + (*out_rgba)[4 * idx + 3] = 1.0; + } + } + } + } else { + for (int i = 0; i < exr_image.width * exr_image.height; i++) { + (*out_rgba)[4 * i + 0] = + reinterpret_cast<float **>(exr_image.images)[idxR][i]; + (*out_rgba)[4 * i + 1] = + reinterpret_cast<float **>(exr_image.images)[idxG][i]; + (*out_rgba)[4 * i + 2] = + reinterpret_cast<float **>(exr_image.images)[idxB][i]; + if (idxA != -1) { + (*out_rgba)[4 * i + 3] = + reinterpret_cast<float **>(exr_image.images)[idxA][i]; + } else { + (*out_rgba)[4 * i + 3] = 1.0; + } } } @@ -11452,7 +11534,7 @@ size_t SaveEXRImageToMemory(const EXRImage *exr_image, if (exr_image == NULL || memory_out == NULL || exr_header->compression_type < 0) { tinyexr::SetErrorMessage("Invalid argument for SaveEXRImageToMemory", err); - return 0; // @fixme + return 0; } #if !TINYEXR_USE_PIZ @@ -11623,8 +11705,6 @@ size_t SaveEXRImageToMemory(const EXRImage *exr_image, sizeof( tinyexr::tinyexr_int64); // sizeof(header) + sizeof(offsetTable) - std::vector<unsigned char> data; - std::vector<std::vector<unsigned char> > data_list( static_cast<size_t>(num_blocks)); std::vector<size_t> channel_offset_list( @@ -11863,9 +11943,9 @@ size_t SaveEXRImageToMemory(const EXRImage *exr_image, } else if (exr_header->compression_type == TINYEXR_COMPRESSIONTYPE_PIZ) { #if TINYEXR_USE_PIZ unsigned int bufLen = - 1024 + static_cast<unsigned int>( - 1.2 * static_cast<unsigned int>( - buf.size())); // @fixme { compute good bound. } + 8192 + static_cast<unsigned int>( + 2 * static_cast<unsigned int>( + buf.size())); // @fixme { compute good bound. } std::vector<unsigned char> block(bufLen); unsigned int outSize = static_cast<unsigned int>(block.size()); @@ -11924,13 +12004,12 @@ size_t SaveEXRImageToMemory(const EXRImage *exr_image, } // omp parallel for (size_t i = 0; i < static_cast<size_t>(num_blocks); i++) { - data.insert(data.end(), data_list[i].begin(), data_list[i].end()); - offsets[i] = offset; tinyexr::swap8(reinterpret_cast<tinyexr::tinyexr_uint64 *>(&offsets[i])); offset += data_list[i].size(); } + size_t totalSize = static_cast<size_t>(offset); { memory.insert( memory.end(), reinterpret_cast<unsigned char *>(&offsets.at(0)), @@ -11938,14 +12017,21 @@ size_t SaveEXRImageToMemory(const EXRImage *exr_image, sizeof(tinyexr::tinyexr_uint64) * static_cast<size_t>(num_blocks)); } - { memory.insert(memory.end(), data.begin(), data.end()); } - - assert(memory.size() > 0); + if ( memory.size() == 0 ) { + tinyexr::SetErrorMessage("Output memory size is zero", err); + return 0; + } - (*memory_out) = static_cast<unsigned char *>(malloc(memory.size())); + (*memory_out) = static_cast<unsigned char *>(malloc(totalSize)); memcpy((*memory_out), &memory.at(0), memory.size()); + unsigned char *memory_ptr = *memory_out + memory.size(); - return memory.size(); // OK + for (size_t i = 0; i < static_cast<size_t>(num_blocks); i++) { + memcpy(memory_ptr, &data_list[i].at(0), data_list[i].size()); + memory_ptr += data_list[i].size(); + } + + return totalSize; // OK } int SaveEXRImageToFile(const EXRImage *exr_image, const EXRHeader *exr_header, @@ -11960,7 +12046,7 @@ int SaveEXRImageToFile(const EXRImage *exr_image, const EXRHeader *exr_header, if (exr_header->compression_type == TINYEXR_COMPRESSIONTYPE_PIZ) { tinyexr::SetErrorMessage("PIZ compression is not supported in this build", err); - return 0; + return TINYEXR_ERROR_UNSUPPORTED_FEATURE; } #endif @@ -11968,7 +12054,7 @@ int SaveEXRImageToFile(const EXRImage *exr_image, const EXRHeader *exr_header, if (exr_header->compression_type == TINYEXR_COMPRESSIONTYPE_ZFP) { tinyexr::SetErrorMessage("ZFP compression is not supported in this build", err); - return 0; + return TINYEXR_ERROR_UNSUPPORTED_FEATURE; } #endif @@ -11980,19 +12066,28 @@ int SaveEXRImageToFile(const EXRImage *exr_image, const EXRHeader *exr_header, #endif if (!fp) { tinyexr::SetErrorMessage("Cannot write a file", err); - return TINYEXR_ERROR_CANT_OPEN_FILE; + return TINYEXR_ERROR_CANT_WRITE_FILE; } unsigned char *mem = NULL; size_t mem_size = SaveEXRImageToMemory(exr_image, exr_header, &mem, err); + if (mem_size == 0) { + return TINYEXR_ERROR_SERIALZATION_FAILED; + } + size_t written_size = 0; if ((mem_size > 0) && mem) { - fwrite(mem, 1, mem_size, fp); + written_size = fwrite(mem, 1, mem_size, fp); } free(mem); fclose(fp); + if (written_size != mem_size) { + tinyexr::SetErrorMessage("Cannot write a file", err); + return TINYEXR_ERROR_CANT_WRITE_FILE; + } + return TINYEXR_SUCCESS; } @@ -12861,20 +12956,27 @@ int LoadEXRMultipartImageFromFile(EXRImage *exr_images, } int SaveEXR(const float *data, int width, int height, int components, - const int save_as_fp16, const char *outfilename) { + const int save_as_fp16, const char *outfilename, const char **err) { if ((components == 1) || components == 3 || components == 4) { // OK } else { + std::stringstream ss; + ss << "Unsupported component value : " << components << std::endl; + + tinyexr::SetErrorMessage(ss.str(), err); return TINYEXR_ERROR_INVALID_ARGUMENT; } - // Assume at least 16x16 pixels. - if (width < 16) return TINYEXR_ERROR_INVALID_ARGUMENT; - if (height < 16) return TINYEXR_ERROR_INVALID_ARGUMENT; - EXRHeader header; InitEXRHeader(&header); + if ((width < 16) && (height < 16)) { + // No compression for small image. + header.compression_type = TINYEXR_COMPRESSIONTYPE_NONE; + } else { + header.compression_type = TINYEXR_COMPRESSIONTYPE_ZIP; + } + EXRImage image; InitEXRImage(&image); @@ -12980,8 +13082,7 @@ int SaveEXR(const float *data, int width, int height, int components, } } - const char *err; - int ret = SaveEXRImageToFile(&image, &header, outfilename, &err); + int ret = SaveEXRImageToFile(&image, &header, outfilename, err); if (ret != TINYEXR_SUCCESS) { return ret; } diff --git a/thirdparty/xatlas/xatlas.cpp b/thirdparty/xatlas/xatlas.cpp new file mode 100644 index 0000000000..f6a9ce64dc --- /dev/null +++ b/thirdparty/xatlas/xatlas.cpp @@ -0,0 +1,7384 @@ +// This code is in the public domain -- castanyo@yahoo.es +#include "xatlas.h" +#include <assert.h> +#include <float.h> +#include <math.h> +#include <stdarg.h> +#include <stdint.h> +#include <stdio.h> +#include <string.h> +#include <time.h> +#include <algorithm> +#include <cmath> +#include <memory> +#include <unordered_map> +#include <vector> + +#undef min +#undef max + +#ifndef xaAssert +#define xaAssert(exp) \ + if (!(exp)) { \ + xaPrint("%s %s %s\n", #exp, __FILE__, __LINE__); \ + } +#endif +#ifndef xaDebugAssert +#define xaDebugAssert(exp) assert(exp) +#endif +#ifndef xaPrint +#define xaPrint(...) \ + if (xatlas::internal::s_print) { \ + xatlas::internal::s_print(__VA_ARGS__); \ + } +#endif + +#ifdef _MSC_VER +// Ignore gcc attributes. +#define __attribute__(X) +#endif + +#ifdef _MSC_VER +#define restrict +#define NV_FORCEINLINE __forceinline +#else +#define restrict __restrict__ +#define NV_FORCEINLINE __attribute__((always_inline)) inline +#endif + +#define NV_UINT32_MAX 0xffffffff +#define NV_FLOAT_MAX 3.402823466e+38F + +#ifndef PI +#define PI float(3.1415926535897932384626433833) +#endif + +#define NV_EPSILON (0.0001f) +#define NV_NORMAL_EPSILON (0.001f) + +namespace xatlas { +namespace internal { + +static PrintFunc s_print = NULL; + +static int align(int x, int a) { + return (x + a - 1) & ~(a - 1); +} + +static bool isAligned(int x, int a) { + return (x & (a - 1)) == 0; +} + +/// Return the maximum of the three arguments. +template <typename T> +static T max3(const T &a, const T &b, const T &c) { + return std::max(a, std::max(b, c)); +} + +/// Return the maximum of the three arguments. +template <typename T> +static T min3(const T &a, const T &b, const T &c) { + return std::min(a, std::min(b, c)); +} + +/// Clamp between two values. +template <typename T> +static T clamp(const T &x, const T &a, const T &b) { + return std::min(std::max(x, a), b); +} + +static float saturate(float f) { + return clamp(f, 0.0f, 1.0f); +} + +// Robust floating point comparisons: +// http://realtimecollisiondetection.net/blog/?p=89 +static bool equal(const float f0, const float f1, const float epsilon = NV_EPSILON) { + //return fabs(f0-f1) <= epsilon; + return fabs(f0 - f1) <= epsilon * max3(1.0f, fabsf(f0), fabsf(f1)); +} + +NV_FORCEINLINE static int ftoi_floor(float val) { + return (int)val; +} + +NV_FORCEINLINE static int ftoi_ceil(float val) { + return (int)ceilf(val); +} + +NV_FORCEINLINE static int ftoi_round(float f) { + return int(floorf(f + 0.5f)); +} + +static bool isZero(const float f, const float epsilon = NV_EPSILON) { + return fabs(f) <= epsilon; +} + +static float lerp(float f0, float f1, float t) { + const float s = 1.0f - t; + return f0 * s + f1 * t; +} + +static float square(float f) { + return f * f; +} + +static int square(int i) { + return i * i; +} + +/** Return the next power of two. +* @see http://graphics.stanford.edu/~seander/bithacks.html +* @warning Behaviour for 0 is undefined. +* @note isPowerOfTwo(x) == true -> nextPowerOfTwo(x) == x +* @note nextPowerOfTwo(x) = 2 << log2(x-1) +*/ +static uint32_t nextPowerOfTwo(uint32_t x) { + xaDebugAssert(x != 0); + // On modern CPUs this is supposed to be as fast as using the bsr instruction. + x--; + x |= x >> 1; + x |= x >> 2; + x |= x >> 4; + x |= x >> 8; + x |= x >> 16; + return x + 1; +} + +static uint64_t nextPowerOfTwo(uint64_t x) { + xaDebugAssert(x != 0); + uint32_t p = 1; + while (x > p) { + p += p; + } + return p; +} + +static uint32_t sdbmHash(const void *data_in, uint32_t size, uint32_t h = 5381) { + const uint8_t *data = (const uint8_t *)data_in; + uint32_t i = 0; + while (i < size) { + h = (h << 16) + (h << 6) - h + (uint32_t)data[i++]; + } + return h; +} + +// Note that this hash does not handle NaN properly. +static uint32_t sdbmFloatHash(const float *f, uint32_t count, uint32_t h = 5381) { + for (uint32_t i = 0; i < count; i++) { + union { + float f; + uint32_t i; + } x = { f[i] }; + if (x.i == 0x80000000) x.i = 0; + h = sdbmHash(&x, 4, h); + } + return h; +} + +template <typename T> +static uint32_t hash(const T &t, uint32_t h = 5381) { + return sdbmHash(&t, sizeof(T), h); +} + +static uint32_t hash(const float &f, uint32_t h) { + return sdbmFloatHash(&f, 1, h); +} + +// Functors for hash table: +template <typename Key> +struct Hash { + uint32_t operator()(const Key &k) const { return hash(k); } +}; + +template <typename Key> +struct Equal { + bool operator()(const Key &k0, const Key &k1) const { return k0 == k1; } +}; + +class Vector2 { +public: + typedef Vector2 const &Arg; + + Vector2() {} + explicit Vector2(float f) : + x(f), + y(f) {} + Vector2(float x, float y) : + x(x), + y(y) {} + Vector2(Vector2::Arg v) : + x(v.x), + y(v.y) {} + + const Vector2 &operator=(Vector2::Arg v) { + x = v.x; + y = v.y; + return *this; + } + const float *ptr() const { return &x; } + + void set(float _x, float _y) { + x = _x; + y = _y; + } + + Vector2 operator-() const { + return Vector2(-x, -y); + } + + void operator+=(Vector2::Arg v) { + x += v.x; + y += v.y; + } + + void operator-=(Vector2::Arg v) { + x -= v.x; + y -= v.y; + } + + void operator*=(float s) { + x *= s; + y *= s; + } + + void operator*=(Vector2::Arg v) { + x *= v.x; + y *= v.y; + } + + friend bool operator==(Vector2::Arg a, Vector2::Arg b) { + return a.x == b.x && a.y == b.y; + } + + friend bool operator!=(Vector2::Arg a, Vector2::Arg b) { + return a.x != b.x || a.y != b.y; + } + + union { +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4201) +#endif + struct + { + float x, y; + }; +#ifdef _MSC_VER +#pragma warning(pop) +#endif + + float component[2]; + }; +}; + +Vector2 operator+(Vector2::Arg a, Vector2::Arg b) { + return Vector2(a.x + b.x, a.y + b.y); +} + +Vector2 operator-(Vector2::Arg a, Vector2::Arg b) { + return Vector2(a.x - b.x, a.y - b.y); +} + +Vector2 operator*(Vector2::Arg v, float s) { + return Vector2(v.x * s, v.y * s); +} + +Vector2 operator*(Vector2::Arg v1, Vector2::Arg v2) { + return Vector2(v1.x * v2.x, v1.y * v2.y); +} + +Vector2 operator/(Vector2::Arg v, float s) { + return Vector2(v.x / s, v.y / s); +} + +Vector2 lerp(Vector2::Arg v1, Vector2::Arg v2, float t) { + const float s = 1.0f - t; + return Vector2(v1.x * s + t * v2.x, v1.y * s + t * v2.y); +} + +float dot(Vector2::Arg a, Vector2::Arg b) { + return a.x * b.x + a.y * b.y; +} + +float lengthSquared(Vector2::Arg v) { + return v.x * v.x + v.y * v.y; +} + +float length(Vector2::Arg v) { + return sqrtf(lengthSquared(v)); +} + +float distance(Vector2::Arg a, Vector2::Arg b) { + return length(a - b); +} + +bool isNormalized(Vector2::Arg v, float epsilon = NV_NORMAL_EPSILON) { + return equal(length(v), 1, epsilon); +} + +Vector2 normalize(Vector2::Arg v, float epsilon = NV_EPSILON) { + float l = length(v); + xaDebugAssert(!isZero(l, epsilon)); +#ifdef NDEBUG + epsilon = 0; // silence unused parameter warning +#endif + Vector2 n = v * (1.0f / l); + xaDebugAssert(isNormalized(n)); + return n; +} + +Vector2 normalizeSafe(Vector2::Arg v, Vector2::Arg fallback, float epsilon = NV_EPSILON) { + float l = length(v); + if (isZero(l, epsilon)) { + return fallback; + } + return v * (1.0f / l); +} + +bool equal(Vector2::Arg v1, Vector2::Arg v2, float epsilon = NV_EPSILON) { + return equal(v1.x, v2.x, epsilon) && equal(v1.y, v2.y, epsilon); +} + +Vector2 max(Vector2::Arg a, Vector2::Arg b) { + return Vector2(std::max(a.x, b.x), std::max(a.y, b.y)); +} + +bool isFinite(Vector2::Arg v) { + return std::isfinite(v.x) && std::isfinite(v.y); +} + +// Note, this is the area scaled by 2! +float triangleArea(Vector2::Arg v0, Vector2::Arg v1) { + return (v0.x * v1.y - v0.y * v1.x); // * 0.5f; +} +float triangleArea(Vector2::Arg a, Vector2::Arg b, Vector2::Arg c) { + // IC: While it may be appealing to use the following expression: + //return (c.x * a.y + a.x * b.y + b.x * c.y - b.x * a.y - c.x * b.y - a.x * c.y); // * 0.5f; + // That's actually a terrible idea. Small triangles far from the origin can end up producing fairly large floating point + // numbers and the results becomes very unstable and dependent on the order of the factors. + // Instead, it's preferable to subtract the vertices first, and multiply the resulting small values together. The result + // in this case is always much more accurate (as long as the triangle is small) and less dependent of the location of + // the triangle. + //return ((a.x - c.x) * (b.y - c.y) - (a.y - c.y) * (b.x - c.x)); // * 0.5f; + return triangleArea(a - c, b - c); +} + +float triangleArea2(Vector2::Arg v1, Vector2::Arg v2, Vector2::Arg v3) { + return 0.5f * (v3.x * v1.y + v1.x * v2.y + v2.x * v3.y - v2.x * v1.y - v3.x * v2.y - v1.x * v3.y); +} + +static uint32_t hash(const Vector2 &v, uint32_t h) { + return sdbmFloatHash(v.component, 2, h); +} + +class Vector3 { +public: + typedef Vector3 const &Arg; + + Vector3() {} + explicit Vector3(float f) : + x(f), + y(f), + z(f) {} + Vector3(float x, float y, float z) : + x(x), + y(y), + z(z) {} + Vector3(Vector2::Arg v, float z) : + x(v.x), + y(v.y), + z(z) {} + Vector3(Vector3::Arg v) : + x(v.x), + y(v.y), + z(v.z) {} + + const Vector3 &operator=(Vector3::Arg v) { + x = v.x; + y = v.y; + z = v.z; + return *this; + } + + Vector2 xy() const { + return Vector2(x, y); + } + + const float *ptr() const { return &x; } + + void set(float _x, float _y, float _z) { + x = _x; + y = _y; + z = _z; + } + + Vector3 operator-() const { + return Vector3(-x, -y, -z); + } + + void operator+=(Vector3::Arg v) { + x += v.x; + y += v.y; + z += v.z; + } + + void operator-=(Vector3::Arg v) { + x -= v.x; + y -= v.y; + z -= v.z; + } + + void operator*=(float s) { + x *= s; + y *= s; + z *= s; + } + + void operator/=(float s) { + float is = 1.0f / s; + x *= is; + y *= is; + z *= is; + } + + void operator*=(Vector3::Arg v) { + x *= v.x; + y *= v.y; + z *= v.z; + } + + void operator/=(Vector3::Arg v) { + x /= v.x; + y /= v.y; + z /= v.z; + } + + friend bool operator==(Vector3::Arg a, Vector3::Arg b) { + return a.x == b.x && a.y == b.y && a.z == b.z; + } + + friend bool operator!=(Vector3::Arg a, Vector3::Arg b) { + return a.x != b.x || a.y != b.y || a.z != b.z; + } + + union { +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4201) +#endif + struct + { + float x, y, z; + }; +#ifdef _MSC_VER +#pragma warning(pop) +#endif + + float component[3]; + }; +}; + +Vector3 add(Vector3::Arg a, Vector3::Arg b) { + return Vector3(a.x + b.x, a.y + b.y, a.z + b.z); +} +Vector3 add(Vector3::Arg a, float b) { + return Vector3(a.x + b, a.y + b, a.z + b); +} +Vector3 operator+(Vector3::Arg a, Vector3::Arg b) { + return add(a, b); +} +Vector3 operator+(Vector3::Arg a, float b) { + return add(a, b); +} + +Vector3 sub(Vector3::Arg a, Vector3::Arg b) { + return Vector3(a.x - b.x, a.y - b.y, a.z - b.z); +} + +Vector3 sub(Vector3::Arg a, float b) { + return Vector3(a.x - b, a.y - b, a.z - b); +} + +Vector3 operator-(Vector3::Arg a, Vector3::Arg b) { + return sub(a, b); +} + +Vector3 operator-(Vector3::Arg a, float b) { + return sub(a, b); +} + +Vector3 cross(Vector3::Arg a, Vector3::Arg b) { + return Vector3(a.y * b.z - a.z * b.y, a.z * b.x - a.x * b.z, a.x * b.y - a.y * b.x); +} + +Vector3 operator*(Vector3::Arg v, float s) { + return Vector3(v.x * s, v.y * s, v.z * s); +} + +Vector3 operator*(float s, Vector3::Arg v) { + return Vector3(v.x * s, v.y * s, v.z * s); +} + +Vector3 operator*(Vector3::Arg v, Vector3::Arg s) { + return Vector3(v.x * s.x, v.y * s.y, v.z * s.z); +} + +Vector3 operator/(Vector3::Arg v, float s) { + return v * (1.0f / s); +} + +Vector3 lerp(Vector3::Arg v1, Vector3::Arg v2, float t) { + const float s = 1.0f - t; + return Vector3(v1.x * s + t * v2.x, v1.y * s + t * v2.y, v1.z * s + t * v2.z); +} + +float dot(Vector3::Arg a, Vector3::Arg b) { + return a.x * b.x + a.y * b.y + a.z * b.z; +} + +float lengthSquared(Vector3::Arg v) { + return v.x * v.x + v.y * v.y + v.z * v.z; +} + +float length(Vector3::Arg v) { + return sqrtf(lengthSquared(v)); +} + +float distance(Vector3::Arg a, Vector3::Arg b) { + return length(a - b); +} + +float distanceSquared(Vector3::Arg a, Vector3::Arg b) { + return lengthSquared(a - b); +} + +bool isNormalized(Vector3::Arg v, float epsilon = NV_NORMAL_EPSILON) { + return equal(length(v), 1, epsilon); +} + +Vector3 normalize(Vector3::Arg v, float epsilon = NV_EPSILON) { + float l = length(v); + xaDebugAssert(!isZero(l, epsilon)); +#ifdef NDEBUG + epsilon = 0; // silence unused parameter warning +#endif + Vector3 n = v * (1.0f / l); + xaDebugAssert(isNormalized(n)); + return n; +} + +Vector3 normalizeSafe(Vector3::Arg v, Vector3::Arg fallback, float epsilon = NV_EPSILON) { + float l = length(v); + if (isZero(l, epsilon)) { + return fallback; + } + return v * (1.0f / l); +} + +bool equal(Vector3::Arg v1, Vector3::Arg v2, float epsilon = NV_EPSILON) { + return equal(v1.x, v2.x, epsilon) && equal(v1.y, v2.y, epsilon) && equal(v1.z, v2.z, epsilon); +} + +Vector3 min(Vector3::Arg a, Vector3::Arg b) { + return Vector3(std::min(a.x, b.x), std::min(a.y, b.y), std::min(a.z, b.z)); +} + +Vector3 max(Vector3::Arg a, Vector3::Arg b) { + return Vector3(std::max(a.x, b.x), std::max(a.y, b.y), std::max(a.z, b.z)); +} + +Vector3 clamp(Vector3::Arg v, float min, float max) { + return Vector3(clamp(v.x, min, max), clamp(v.y, min, max), clamp(v.z, min, max)); +} + +Vector3 saturate(Vector3::Arg v) { + return Vector3(saturate(v.x), saturate(v.y), saturate(v.z)); +} + +Vector3 floor(Vector3::Arg v) { + return Vector3(floorf(v.x), floorf(v.y), floorf(v.z)); +} + +bool isFinite(Vector3::Arg v) { + return std::isfinite(v.x) && std::isfinite(v.y) && std::isfinite(v.z); +} + +static uint32_t hash(const Vector3 &v, uint32_t h) { + return sdbmFloatHash(v.component, 3, h); +} + +/// Basis class to compute tangent space basis, ortogonalizations and to +/// transform vectors from one space to another. +class Basis { +public: + /// Create a null basis. + Basis() : + tangent(0, 0, 0), + bitangent(0, 0, 0), + normal(0, 0, 0) {} + + void buildFrameForDirection(Vector3::Arg d, float angle = 0) { + xaAssert(isNormalized(d)); + normal = d; + // Choose minimum axis. + if (fabsf(normal.x) < fabsf(normal.y) && fabsf(normal.x) < fabsf(normal.z)) { + tangent = Vector3(1, 0, 0); + } else if (fabsf(normal.y) < fabsf(normal.z)) { + tangent = Vector3(0, 1, 0); + } else { + tangent = Vector3(0, 0, 1); + } + // Ortogonalize + tangent -= normal * dot(normal, tangent); + tangent = normalize(tangent); + bitangent = cross(normal, tangent); + // Rotate frame around normal according to angle. + if (angle != 0.0f) { + float c = cosf(angle); + float s = sinf(angle); + Vector3 tmp = c * tangent - s * bitangent; + bitangent = s * tangent + c * bitangent; + tangent = tmp; + } + } + + Vector3 tangent; + Vector3 bitangent; + Vector3 normal; +}; + +// Simple bit array. +class BitArray { +public: + BitArray() : + m_size(0) {} + BitArray(uint32_t sz) { + resize(sz); + } + + uint32_t size() const { + return m_size; + } + + void clear() { + resize(0); + } + + void resize(uint32_t new_size) { + m_size = new_size; + m_wordArray.resize((m_size + 31) >> 5); + } + + /// Get bit. + bool bitAt(uint32_t b) const { + xaDebugAssert(b < m_size); + return (m_wordArray[b >> 5] & (1 << (b & 31))) != 0; + } + + // Set a bit. + void setBitAt(uint32_t idx) { + xaDebugAssert(idx < m_size); + m_wordArray[idx >> 5] |= (1 << (idx & 31)); + } + + // Toggle a bit. + void toggleBitAt(uint32_t idx) { + xaDebugAssert(idx < m_size); + m_wordArray[idx >> 5] ^= (1 << (idx & 31)); + } + + // Set a bit to the given value. @@ Rename modifyBitAt? + void setBitAt(uint32_t idx, bool b) { + xaDebugAssert(idx < m_size); + m_wordArray[idx >> 5] = setBits(m_wordArray[idx >> 5], 1 << (idx & 31), b); + xaDebugAssert(bitAt(idx) == b); + } + + // Clear all the bits. + void clearAll() { + memset(m_wordArray.data(), 0, m_wordArray.size() * sizeof(uint32_t)); + } + + // Set all the bits. + void setAll() { + memset(m_wordArray.data(), 0xFF, m_wordArray.size() * sizeof(uint32_t)); + } + +private: + // See "Conditionally set or clear bits without branching" at http://graphics.stanford.edu/~seander/bithacks.html + uint32_t setBits(uint32_t w, uint32_t m, bool b) { + return (w & ~m) | (-int(b) & m); + } + + // Number of bits stored. + uint32_t m_size; + + // Array of bits. + std::vector<uint32_t> m_wordArray; +}; + +/// Bit map. This should probably be called BitImage. +class BitMap { +public: + BitMap() : + m_width(0), + m_height(0) {} + BitMap(uint32_t w, uint32_t h) : + m_width(w), + m_height(h), + m_bitArray(w * h) {} + + uint32_t width() const { + return m_width; + } + uint32_t height() const { + return m_height; + } + + void resize(uint32_t w, uint32_t h, bool initValue) { + BitArray tmp(w * h); + if (initValue) + tmp.setAll(); + else + tmp.clearAll(); + // @@ Copying one bit at a time. This could be much faster. + for (uint32_t y = 0; y < m_height; y++) { + for (uint32_t x = 0; x < m_width; x++) { + //tmp.setBitAt(y*w + x, bitAt(x, y)); + if (bitAt(x, y) != initValue) tmp.toggleBitAt(y * w + x); + } + } + std::swap(m_bitArray, tmp); + m_width = w; + m_height = h; + } + + bool bitAt(uint32_t x, uint32_t y) const { + xaDebugAssert(x < m_width && y < m_height); + return m_bitArray.bitAt(y * m_width + x); + } + + void setBitAt(uint32_t x, uint32_t y) { + xaDebugAssert(x < m_width && y < m_height); + m_bitArray.setBitAt(y * m_width + x); + } + + void clearAll() { + m_bitArray.clearAll(); + } + +private: + uint32_t m_width; + uint32_t m_height; + BitArray m_bitArray; +}; + +// Axis Aligned Bounding Box. +class Box { +public: + Box() {} + Box(const Box &b) : + minCorner(b.minCorner), + maxCorner(b.maxCorner) {} + Box(const Vector3 &mins, const Vector3 &maxs) : + minCorner(mins), + maxCorner(maxs) {} + + operator const float *() const { + return reinterpret_cast<const float *>(this); + } + + // Clear the bounds. + void clearBounds() { + minCorner.set(FLT_MAX, FLT_MAX, FLT_MAX); + maxCorner.set(-FLT_MAX, -FLT_MAX, -FLT_MAX); + } + + // Return extents of the box. + Vector3 extents() const { + return (maxCorner - minCorner) * 0.5f; + } + + // Add a point to this box. + void addPointToBounds(const Vector3 &p) { + minCorner = min(minCorner, p); + maxCorner = max(maxCorner, p); + } + + // Get the volume of the box. + float volume() const { + Vector3 d = extents(); + return 8.0f * (d.x * d.y * d.z); + } + + Vector3 minCorner; + Vector3 maxCorner; +}; + +class Fit { +public: + static Vector3 computeCentroid(int n, const Vector3 *__restrict points) { + Vector3 centroid(0.0f); + for (int i = 0; i < n; i++) { + centroid += points[i]; + } + centroid /= float(n); + return centroid; + } + + static Vector3 computeCovariance(int n, const Vector3 *__restrict points, float *__restrict covariance) { + // compute the centroid + Vector3 centroid = computeCentroid(n, points); + // compute covariance matrix + for (int i = 0; i < 6; i++) { + covariance[i] = 0.0f; + } + for (int i = 0; i < n; i++) { + Vector3 v = points[i] - centroid; + covariance[0] += v.x * v.x; + covariance[1] += v.x * v.y; + covariance[2] += v.x * v.z; + covariance[3] += v.y * v.y; + covariance[4] += v.y * v.z; + covariance[5] += v.z * v.z; + } + return centroid; + } + + static bool isPlanar(int n, const Vector3 *points, float epsilon = NV_EPSILON) { + // compute the centroid and covariance + float matrix[6]; + computeCovariance(n, points, matrix); + float eigenValues[3]; + Vector3 eigenVectors[3]; + if (!eigenSolveSymmetric3(matrix, eigenValues, eigenVectors)) { + return false; + } + return eigenValues[2] < epsilon; + } + + // Tridiagonal solver from Charles Bloom. + // Householder transforms followed by QL decomposition. + // Seems to be based on the code from Numerical Recipes in C. + static bool eigenSolveSymmetric3(const float matrix[6], float eigenValues[3], Vector3 eigenVectors[3]) { + xaDebugAssert(matrix != NULL && eigenValues != NULL && eigenVectors != NULL); + float subd[3]; + float diag[3]; + float work[3][3]; + work[0][0] = matrix[0]; + work[0][1] = work[1][0] = matrix[1]; + work[0][2] = work[2][0] = matrix[2]; + work[1][1] = matrix[3]; + work[1][2] = work[2][1] = matrix[4]; + work[2][2] = matrix[5]; + EigenSolver3_Tridiagonal(work, diag, subd); + if (!EigenSolver3_QLAlgorithm(work, diag, subd)) { + for (int i = 0; i < 3; i++) { + eigenValues[i] = 0; + eigenVectors[i] = Vector3(0); + } + return false; + } + for (int i = 0; i < 3; i++) { + eigenValues[i] = (float)diag[i]; + } + // eigenvectors are the columns; make them the rows : + for (int i = 0; i < 3; i++) { + for (int j = 0; j < 3; j++) { + eigenVectors[j].component[i] = (float)work[i][j]; + } + } + // shuffle to sort by singular value : + if (eigenValues[2] > eigenValues[0] && eigenValues[2] > eigenValues[1]) { + std::swap(eigenValues[0], eigenValues[2]); + std::swap(eigenVectors[0], eigenVectors[2]); + } + if (eigenValues[1] > eigenValues[0]) { + std::swap(eigenValues[0], eigenValues[1]); + std::swap(eigenVectors[0], eigenVectors[1]); + } + if (eigenValues[2] > eigenValues[1]) { + std::swap(eigenValues[1], eigenValues[2]); + std::swap(eigenVectors[1], eigenVectors[2]); + } + xaDebugAssert(eigenValues[0] >= eigenValues[1] && eigenValues[0] >= eigenValues[2]); + xaDebugAssert(eigenValues[1] >= eigenValues[2]); + return true; + } + +private: + static void EigenSolver3_Tridiagonal(float mat[3][3], float *diag, float *subd) { + // Householder reduction T = Q^t M Q + // Input: + // mat, symmetric 3x3 matrix M + // Output: + // mat, orthogonal matrix Q + // diag, diagonal entries of T + // subd, subdiagonal entries of T (T is symmetric) + const float epsilon = 1e-08f; + float a = mat[0][0]; + float b = mat[0][1]; + float c = mat[0][2]; + float d = mat[1][1]; + float e = mat[1][2]; + float f = mat[2][2]; + diag[0] = a; + subd[2] = 0.f; + if (fabsf(c) >= epsilon) { + const float ell = sqrtf(b * b + c * c); + b /= ell; + c /= ell; + const float q = 2 * b * e + c * (f - d); + diag[1] = d + c * q; + diag[2] = f - c * q; + subd[0] = ell; + subd[1] = e - b * q; + mat[0][0] = 1; + mat[0][1] = 0; + mat[0][2] = 0; + mat[1][0] = 0; + mat[1][1] = b; + mat[1][2] = c; + mat[2][0] = 0; + mat[2][1] = c; + mat[2][2] = -b; + } else { + diag[1] = d; + diag[2] = f; + subd[0] = b; + subd[1] = e; + mat[0][0] = 1; + mat[0][1] = 0; + mat[0][2] = 0; + mat[1][0] = 0; + mat[1][1] = 1; + mat[1][2] = 0; + mat[2][0] = 0; + mat[2][1] = 0; + mat[2][2] = 1; + } + } + + static bool EigenSolver3_QLAlgorithm(float mat[3][3], float *diag, float *subd) { + // QL iteration with implicit shifting to reduce matrix from tridiagonal + // to diagonal + const int maxiter = 32; + for (int ell = 0; ell < 3; ell++) { + int iter; + for (iter = 0; iter < maxiter; iter++) { + int m; + for (m = ell; m <= 1; m++) { + float dd = fabsf(diag[m]) + fabsf(diag[m + 1]); + if (fabsf(subd[m]) + dd == dd) + break; + } + if (m == ell) + break; + float g = (diag[ell + 1] - diag[ell]) / (2 * subd[ell]); + float r = sqrtf(g * g + 1); + if (g < 0) + g = diag[m] - diag[ell] + subd[ell] / (g - r); + else + g = diag[m] - diag[ell] + subd[ell] / (g + r); + float s = 1, c = 1, p = 0; + for (int i = m - 1; i >= ell; i--) { + float f = s * subd[i], b = c * subd[i]; + if (fabsf(f) >= fabsf(g)) { + c = g / f; + r = sqrtf(c * c + 1); + subd[i + 1] = f * r; + c *= (s = 1 / r); + } else { + s = f / g; + r = sqrtf(s * s + 1); + subd[i + 1] = g * r; + s *= (c = 1 / r); + } + g = diag[i + 1] - p; + r = (diag[i] - g) * s + 2 * b * c; + p = s * r; + diag[i + 1] = g + p; + g = c * r - b; + for (int k = 0; k < 3; k++) { + f = mat[k][i + 1]; + mat[k][i + 1] = s * mat[k][i] + c * f; + mat[k][i] = c * mat[k][i] - s * f; + } + } + diag[ell] -= p; + subd[ell] = g; + subd[m] = 0; + } + if (iter == maxiter) + // should not get here under normal circumstances + return false; + } + return true; + } +}; + +/// Fixed size vector class. +class FullVector { +public: + FullVector(uint32_t dim) { m_array.resize(dim); } + FullVector(const FullVector &v) : + m_array(v.m_array) {} + + const FullVector &operator=(const FullVector &v) { + xaAssert(dimension() == v.dimension()); + m_array = v.m_array; + return *this; + } + + uint32_t dimension() const { return m_array.size(); } + const float &operator[](uint32_t index) const { return m_array[index]; } + float &operator[](uint32_t index) { return m_array[index]; } + + void fill(float f) { + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] = f; + } + } + + void operator+=(const FullVector &v) { + xaDebugAssert(dimension() == v.dimension()); + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] += v.m_array[i]; + } + } + + void operator-=(const FullVector &v) { + xaDebugAssert(dimension() == v.dimension()); + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] -= v.m_array[i]; + } + } + + void operator*=(const FullVector &v) { + xaDebugAssert(dimension() == v.dimension()); + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] *= v.m_array[i]; + } + } + + void operator+=(float f) { + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] += f; + } + } + + void operator-=(float f) { + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] -= f; + } + } + + void operator*=(float f) { + const uint32_t dim = dimension(); + for (uint32_t i = 0; i < dim; i++) { + m_array[i] *= f; + } + } + +private: + std::vector<float> m_array; +}; + +namespace halfedge { +class Face; +class Vertex; + +class Edge { +public: + uint32_t id; + Edge *next; + Edge *prev; // This is not strictly half-edge, but makes algorithms easier and faster. + Edge *pair; + Vertex *vertex; + Face *face; + + // Default constructor. + Edge(uint32_t id) : + id(id), + next(NULL), + prev(NULL), + pair(NULL), + vertex(NULL), + face(NULL) {} + + // Vertex queries. + const Vertex *from() const { + return vertex; + } + + Vertex *from() { + return vertex; + } + + const Vertex *to() const { + return pair->vertex; // This used to be 'next->vertex', but that changed often when the connectivity of the mesh changes. + } + + Vertex *to() { + return pair->vertex; + } + + // Edge queries. + void setNext(Edge *e) { + next = e; + if (e != NULL) e->prev = this; + } + void setPrev(Edge *e) { + prev = e; + if (e != NULL) e->next = this; + } + + // @@ It would be more simple to only check m_pair == NULL + // Face queries. + bool isBoundary() const { + return !(face && pair->face); + } + + // @@ This is not exactly accurate, we should compare the texture coordinates... + bool isSeam() const { + return vertex != pair->next->vertex || next->vertex != pair->vertex; + } + + bool isNormalSeam() const; + bool isTextureSeam() const; + + bool isValid() const { + // null face is OK. + if (next == NULL || prev == NULL || pair == NULL || vertex == NULL) return false; + if (next->prev != this) return false; + if (prev->next != this) return false; + if (pair->pair != this) return false; + return true; + } + + float length() const; + + // Return angle between this edge and the previous one. + float angle() const; +}; + +class Vertex { +public: + uint32_t id; + uint32_t original_id; + Edge *edge; + Vertex *next; + Vertex *prev; + Vector3 pos; + Vector3 nor; + Vector2 tex; + + Vertex(uint32_t id) : + id(id), + original_id(id), + edge(NULL), + pos(0.0f), + nor(0.0f), + tex(0.0f) { + next = this; + prev = this; + } + + // Set first edge of all colocals. + void setEdge(Edge *e) { + for (VertexIterator it(colocals()); !it.isDone(); it.advance()) { + it.current()->edge = e; + } + } + + // Update position of all colocals. + void setPos(const Vector3 &p) { + for (VertexIterator it(colocals()); !it.isDone(); it.advance()) { + it.current()->pos = p; + } + } + + bool isFirstColocal() const { + return firstColocal() == this; + } + + const Vertex *firstColocal() const { + uint32_t firstId = id; + const Vertex *vertex = this; + for (ConstVertexIterator it(colocals()); !it.isDone(); it.advance()) { + if (it.current()->id < firstId) { + firstId = vertex->id; + vertex = it.current(); + } + } + return vertex; + } + + Vertex *firstColocal() { + Vertex *vertex = this; + uint32_t firstId = id; + for (VertexIterator it(colocals()); !it.isDone(); it.advance()) { + if (it.current()->id < firstId) { + firstId = vertex->id; + vertex = it.current(); + } + } + return vertex; + } + + bool isColocal(const Vertex *v) const { + if (this == v) return true; + if (pos != v->pos) return false; + for (ConstVertexIterator it(colocals()); !it.isDone(); it.advance()) { + if (v == it.current()) { + return true; + } + } + return false; + } + + void linkColocal(Vertex *v) { + next->prev = v; + v->next = next; + next = v; + v->prev = this; + } + void unlinkColocal() { + next->prev = prev; + prev->next = next; + next = this; + prev = this; + } + + // @@ Note: This only works if linkBoundary has been called. + bool isBoundary() const { + return (edge && !edge->face); + } + + // Iterator that visits the edges around this vertex in counterclockwise order. + class EdgeIterator //: public Iterator<Edge *> + { + public: + EdgeIterator(Edge *e) : + m_end(NULL), + m_current(e) {} + + virtual void advance() { + if (m_end == NULL) m_end = m_current; + m_current = m_current->pair->next; + //m_current = m_current->prev->pair; + } + + virtual bool isDone() const { + return m_end == m_current; + } + virtual Edge *current() const { + return m_current; + } + Vertex *vertex() const { + return m_current->vertex; + } + + private: + Edge *m_end; + Edge *m_current; + }; + + EdgeIterator edges() { + return EdgeIterator(edge); + } + EdgeIterator edges(Edge *e) { + return EdgeIterator(e); + } + + // Iterator that visits the edges around this vertex in counterclockwise order. + class ConstEdgeIterator //: public Iterator<Edge *> + { + public: + ConstEdgeIterator(const Edge *e) : + m_end(NULL), + m_current(e) {} + ConstEdgeIterator(EdgeIterator it) : + m_end(NULL), + m_current(it.current()) {} + + virtual void advance() { + if (m_end == NULL) m_end = m_current; + m_current = m_current->pair->next; + //m_current = m_current->prev->pair; + } + + virtual bool isDone() const { + return m_end == m_current; + } + virtual const Edge *current() const { + return m_current; + } + const Vertex *vertex() const { + return m_current->to(); + } + + private: + const Edge *m_end; + const Edge *m_current; + }; + + ConstEdgeIterator edges() const { + return ConstEdgeIterator(edge); + } + ConstEdgeIterator edges(const Edge *e) const { + return ConstEdgeIterator(e); + } + + // Iterator that visits all the colocal vertices. + class VertexIterator //: public Iterator<Edge *> + { + public: + VertexIterator(Vertex *v) : + m_end(NULL), + m_current(v) {} + + virtual void advance() { + if (m_end == NULL) m_end = m_current; + m_current = m_current->next; + } + + virtual bool isDone() const { + return m_end == m_current; + } + virtual Vertex *current() const { + return m_current; + } + + private: + Vertex *m_end; + Vertex *m_current; + }; + + VertexIterator colocals() { + return VertexIterator(this); + } + + // Iterator that visits all the colocal vertices. + class ConstVertexIterator //: public Iterator<Edge *> + { + public: + ConstVertexIterator(const Vertex *v) : + m_end(NULL), + m_current(v) {} + + virtual void advance() { + if (m_end == NULL) m_end = m_current; + m_current = m_current->next; + } + + virtual bool isDone() const { + return m_end == m_current; + } + virtual const Vertex *current() const { + return m_current; + } + + private: + const Vertex *m_end; + const Vertex *m_current; + }; + + ConstVertexIterator colocals() const { + return ConstVertexIterator(this); + } +}; + +bool Edge::isNormalSeam() const { + return (vertex->nor != pair->next->vertex->nor || next->vertex->nor != pair->vertex->nor); +} + +bool Edge::isTextureSeam() const { + return (vertex->tex != pair->next->vertex->tex || next->vertex->tex != pair->vertex->tex); +} + +float Edge::length() const { + return internal::length(to()->pos - from()->pos); +} + +float Edge::angle() const { + Vector3 p = vertex->pos; + Vector3 a = prev->vertex->pos; + Vector3 b = next->vertex->pos; + Vector3 v0 = a - p; + Vector3 v1 = b - p; + return acosf(dot(v0, v1) / (internal::length(v0) * internal::length(v1))); +} + +class Face { +public: + uint32_t id; + uint16_t group; + uint16_t material; + Edge *edge; + + Face(uint32_t id) : + id(id), + group(uint16_t(~0)), + material(uint16_t(~0)), + edge(NULL) {} + + float area() const { + float area = 0; + const Vector3 &v0 = edge->from()->pos; + for (ConstEdgeIterator it(edges(edge->next)); it.current() != edge->prev; it.advance()) { + const Edge *e = it.current(); + const Vector3 &v1 = e->vertex->pos; + const Vector3 &v2 = e->next->vertex->pos; + area += length(cross(v1 - v0, v2 - v0)); + } + return area * 0.5f; + } + + float parametricArea() const { + float area = 0; + const Vector2 &v0 = edge->from()->tex; + for (ConstEdgeIterator it(edges(edge->next)); it.current() != edge->prev; it.advance()) { + const Edge *e = it.current(); + const Vector2 &v1 = e->vertex->tex; + const Vector2 &v2 = e->next->vertex->tex; + area += triangleArea(v0, v1, v2); + } + return area * 0.5f; + } + + Vector3 normal() const { + Vector3 n(0); + const Vertex *vertex0 = NULL; + for (ConstEdgeIterator it(edges()); !it.isDone(); it.advance()) { + const Edge *e = it.current(); + xaAssert(e != NULL); + if (vertex0 == NULL) { + vertex0 = e->vertex; + } else if (e->next->vertex != vertex0) { + const halfedge::Vertex *vertex1 = e->from(); + const halfedge::Vertex *vertex2 = e->to(); + const Vector3 &p0 = vertex0->pos; + const Vector3 &p1 = vertex1->pos; + const Vector3 &p2 = vertex2->pos; + Vector3 v10 = p1 - p0; + Vector3 v20 = p2 - p0; + n += cross(v10, v20); + } + } + return normalizeSafe(n, Vector3(0, 0, 1), 0.0f); + } + + Vector3 centroid() const { + Vector3 sum(0.0f); + uint32_t count = 0; + for (ConstEdgeIterator it(edges()); !it.isDone(); it.advance()) { + const Edge *e = it.current(); + sum += e->from()->pos; + count++; + } + return sum / float(count); + } + + // Unnormalized face normal assuming it's a triangle. + Vector3 triangleNormal() const { + Vector3 p0 = edge->vertex->pos; + Vector3 p1 = edge->next->vertex->pos; + Vector3 p2 = edge->next->next->vertex->pos; + Vector3 e0 = p2 - p0; + Vector3 e1 = p1 - p0; + return normalizeSafe(cross(e0, e1), Vector3(0), 0.0f); + } + + Vector3 triangleNormalAreaScaled() const { + Vector3 p0 = edge->vertex->pos; + Vector3 p1 = edge->next->vertex->pos; + Vector3 p2 = edge->next->next->vertex->pos; + Vector3 e0 = p2 - p0; + Vector3 e1 = p1 - p0; + return cross(e0, e1); + } + + // Average of the edge midpoints weighted by the edge length. + // I want a point inside the triangle, but closer to the cirumcenter. + Vector3 triangleCenter() const { + Vector3 p0 = edge->vertex->pos; + Vector3 p1 = edge->next->vertex->pos; + Vector3 p2 = edge->next->next->vertex->pos; + float l0 = length(p1 - p0); + float l1 = length(p2 - p1); + float l2 = length(p0 - p2); + Vector3 m0 = (p0 + p1) * l0 / (l0 + l1 + l2); + Vector3 m1 = (p1 + p2) * l1 / (l0 + l1 + l2); + Vector3 m2 = (p2 + p0) * l2 / (l0 + l1 + l2); + return m0 + m1 + m2; + } + + bool isValid() const { + uint32_t count = 0; + for (ConstEdgeIterator it(edges()); !it.isDone(); it.advance()) { + const Edge *e = it.current(); + if (e->face != this) return false; + if (!e->isValid()) return false; + if (!e->pair->isValid()) return false; + count++; + } + if (count < 3) return false; + return true; + } + + bool contains(const Edge *e) const { + for (ConstEdgeIterator it(edges()); !it.isDone(); it.advance()) { + if (it.current() == e) return true; + } + return false; + } + + uint32_t edgeCount() const { + uint32_t count = 0; + for (ConstEdgeIterator it(edges()); !it.isDone(); it.advance()) { + ++count; + } + return count; + } + + // The iterator that visits the edges of this face in clockwise order. + class EdgeIterator //: public Iterator<Edge *> + { + public: + EdgeIterator(Edge *e) : + m_end(NULL), + m_current(e) {} + + virtual void advance() { + if (m_end == NULL) m_end = m_current; + m_current = m_current->next; + } + + virtual bool isDone() const { + return m_end == m_current; + } + virtual Edge *current() const { + return m_current; + } + Vertex *vertex() const { + return m_current->vertex; + } + + private: + Edge *m_end; + Edge *m_current; + }; + + EdgeIterator edges() { + return EdgeIterator(edge); + } + EdgeIterator edges(Edge *e) { + xaDebugAssert(contains(e)); + return EdgeIterator(e); + } + + // The iterator that visits the edges of this face in clockwise order. + class ConstEdgeIterator //: public Iterator<const Edge *> + { + public: + ConstEdgeIterator(const Edge *e) : + m_end(NULL), + m_current(e) {} + ConstEdgeIterator(const EdgeIterator &it) : + m_end(NULL), + m_current(it.current()) {} + + virtual void advance() { + if (m_end == NULL) m_end = m_current; + m_current = m_current->next; + } + + virtual bool isDone() const { + return m_end == m_current; + } + virtual const Edge *current() const { + return m_current; + } + const Vertex *vertex() const { + return m_current->vertex; + } + + private: + const Edge *m_end; + const Edge *m_current; + }; + + ConstEdgeIterator edges() const { + return ConstEdgeIterator(edge); + } + ConstEdgeIterator edges(const Edge *e) const { + xaDebugAssert(contains(e)); + return ConstEdgeIterator(e); + } +}; + +/// Simple half edge mesh designed for dynamic mesh manipulation. +class Mesh { +public: + Mesh() : + m_colocalVertexCount(0) {} + + Mesh(const Mesh *mesh) { + // Copy mesh vertices. + const uint32_t vertexCount = mesh->vertexCount(); + m_vertexArray.resize(vertexCount); + for (uint32_t v = 0; v < vertexCount; v++) { + const Vertex *vertex = mesh->vertexAt(v); + xaDebugAssert(vertex->id == v); + m_vertexArray[v] = new Vertex(v); + m_vertexArray[v]->pos = vertex->pos; + m_vertexArray[v]->nor = vertex->nor; + m_vertexArray[v]->tex = vertex->tex; + } + m_colocalVertexCount = vertexCount; + // Copy mesh faces. + const uint32_t faceCount = mesh->faceCount(); + std::vector<uint32_t> indexArray; + indexArray.reserve(3); + for (uint32_t f = 0; f < faceCount; f++) { + const Face *face = mesh->faceAt(f); + for (Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const Vertex *vertex = it.current()->from(); + indexArray.push_back(vertex->id); + } + addFace(indexArray); + indexArray.clear(); + } + } + + ~Mesh() { + clear(); + } + + void clear() { + for (size_t i = 0; i < m_vertexArray.size(); i++) + delete m_vertexArray[i]; + m_vertexArray.clear(); + for (auto it = m_edgeMap.begin(); it != m_edgeMap.end(); it++) + delete it->second; + m_edgeArray.clear(); + m_edgeMap.clear(); + for (size_t i = 0; i < m_faceArray.size(); i++) + delete m_faceArray[i]; + m_faceArray.clear(); + } + + Vertex *addVertex(const Vector3 &pos) { + xaDebugAssert(isFinite(pos)); + Vertex *v = new Vertex(m_vertexArray.size()); + v->pos = pos; + m_vertexArray.push_back(v); + return v; + } + + /// Link colocal vertices based on geometric location only. + void linkColocals() { + xaPrint("--- Linking colocals:\n"); + const uint32_t vertexCount = this->vertexCount(); + std::unordered_map<Vector3, Vertex *, Hash<Vector3>, Equal<Vector3> > vertexMap; + vertexMap.reserve(vertexCount); + for (uint32_t v = 0; v < vertexCount; v++) { + Vertex *vertex = vertexAt(v); + Vertex *colocal = vertexMap[vertex->pos]; + if (colocal) { + colocal->linkColocal(vertex); + } else { + vertexMap[vertex->pos] = vertex; + } + } + m_colocalVertexCount = vertexMap.size(); + xaPrint("--- %d vertex positions.\n", m_colocalVertexCount); + // @@ Remove duplicated vertices? or just leave them as colocals? + } + + void linkColocalsWithCanonicalMap(const std::vector<uint32_t> &canonicalMap) { + xaPrint("--- Linking colocals:\n"); + uint32_t vertexMapSize = 0; + for (uint32_t i = 0; i < canonicalMap.size(); i++) { + vertexMapSize = std::max(vertexMapSize, canonicalMap[i] + 1); + } + std::vector<Vertex *> vertexMap; + vertexMap.resize(vertexMapSize, NULL); + m_colocalVertexCount = 0; + const uint32_t vertexCount = this->vertexCount(); + for (uint32_t v = 0; v < vertexCount; v++) { + Vertex *vertex = vertexAt(v); + Vertex *colocal = vertexMap[canonicalMap[v]]; + if (colocal != NULL) { + xaDebugAssert(vertex->pos == colocal->pos); + colocal->linkColocal(vertex); + } else { + vertexMap[canonicalMap[v]] = vertex; + m_colocalVertexCount++; + } + } + xaPrint("--- %d vertex positions.\n", m_colocalVertexCount); + } + + Face *addFace() { + Face *f = new Face(m_faceArray.size()); + m_faceArray.push_back(f); + return f; + } + + Face *addFace(uint32_t v0, uint32_t v1, uint32_t v2) { + uint32_t indexArray[3]; + indexArray[0] = v0; + indexArray[1] = v1; + indexArray[2] = v2; + return addFace(indexArray, 3, 0, 3); + } + + Face *addUniqueFace(uint32_t v0, uint32_t v1, uint32_t v2) { + + int base_vertex = m_vertexArray.size(); + + uint32_t ids[3] = { v0, v1, v2 }; + + Vector3 base[3] = { + m_vertexArray[v0]->pos, + m_vertexArray[v1]->pos, + m_vertexArray[v2]->pos, + }; + + //make sure its not a degenerate + bool degenerate = distanceSquared(base[0], base[1]) < NV_EPSILON || distanceSquared(base[0], base[2]) < NV_EPSILON || distanceSquared(base[1], base[2]) < NV_EPSILON; + xaDebugAssert(!degenerate); + + float min_x = 0; + + for (int i = 0; i < 3; i++) { + if (i == 0 || m_vertexArray[v0]->pos.x < min_x) { + min_x = m_vertexArray[v0]->pos.x; + } + } + + float max_x = 0; + + for (int j = 0; j < m_vertexArray.size(); j++) { + if (j == 0 || m_vertexArray[j]->pos.x > max_x) { //vertex already exists + max_x = m_vertexArray[j]->pos.x; + } + } + + //separate from everything else, in x axis + for (int i = 0; i < 3; i++) { + + base[i].x -= min_x; + base[i].x += max_x + 10.0; + } + + for (int i = 0; i < 3; i++) { + Vertex *v = new Vertex(m_vertexArray.size()); + v->pos = base[i]; + v->nor = m_vertexArray[ids[i]]->nor, + v->tex = m_vertexArray[ids[i]]->tex, + + v->original_id = ids[i]; + m_vertexArray.push_back(v); + } + + uint32_t indexArray[3]; + indexArray[0] = base_vertex + 0; + indexArray[1] = base_vertex + 1; + indexArray[2] = base_vertex + 2; + return addFace(indexArray, 3, 0, 3); + } + + Face *addFace(uint32_t v0, uint32_t v1, uint32_t v2, uint32_t v3) { + uint32_t indexArray[4]; + indexArray[0] = v0; + indexArray[1] = v1; + indexArray[2] = v2; + indexArray[3] = v3; + return addFace(indexArray, 4, 0, 4); + } + + Face *addFace(const std::vector<uint32_t> &indexArray) { + return addFace(indexArray, 0, indexArray.size()); + } + + Face *addFace(const std::vector<uint32_t> &indexArray, uint32_t first, uint32_t num) { + return addFace(indexArray.data(), (uint32_t)indexArray.size(), first, num); + } + + Face *addFace(const uint32_t *indexArray, uint32_t indexCount, uint32_t first, uint32_t num) { + xaDebugAssert(first < indexCount); + xaDebugAssert(num <= indexCount - first); + xaDebugAssert(num > 2); + if (!canAddFace(indexArray, first, num)) { + return NULL; + } + Face *f = new Face(m_faceArray.size()); + Edge *firstEdge = NULL; + Edge *last = NULL; + Edge *current = NULL; + for (uint32_t i = 0; i < num - 1; i++) { + current = addEdge(indexArray[first + i], indexArray[first + i + 1]); + xaAssert(current != NULL && current->face == NULL); + current->face = f; + if (last != NULL) + last->setNext(current); + else + firstEdge = current; + last = current; + } + current = addEdge(indexArray[first + num - 1], indexArray[first]); + xaAssert(current != NULL && current->face == NULL); + current->face = f; + last->setNext(current); + current->setNext(firstEdge); + f->edge = firstEdge; + m_faceArray.push_back(f); + return f; + } + + // These functions disconnect the given element from the mesh and delete it. + + // @@ We must always disconnect edge pairs simultaneously. + void disconnect(Edge *edge) { + xaDebugAssert(edge != NULL); + // Remove from edge list. + if ((edge->id & 1) == 0) { + xaDebugAssert(m_edgeArray[edge->id / 2] == edge); + m_edgeArray[edge->id / 2] = NULL; + } + // Remove edge from map. @@ Store map key inside edge? + xaDebugAssert(edge->from() != NULL && edge->to() != NULL); + size_t removed = m_edgeMap.erase(Key(edge->from()->id, edge->to()->id)); + xaDebugAssert(removed == 1); +#ifdef NDEBUG + removed = 0; // silence unused parameter warning +#endif + // Disconnect from vertex. + if (edge->vertex != NULL) { + if (edge->vertex->edge == edge) { + if (edge->prev && edge->prev->pair) { + edge->vertex->edge = edge->prev->pair; + } else if (edge->pair && edge->pair->next) { + edge->vertex->edge = edge->pair->next; + } else { + edge->vertex->edge = NULL; + // @@ Remove disconnected vertex? + } + } + } + // Disconnect from face. + if (edge->face != NULL) { + if (edge->face->edge == edge) { + if (edge->next != NULL && edge->next != edge) { + edge->face->edge = edge->next; + } else if (edge->prev != NULL && edge->prev != edge) { + edge->face->edge = edge->prev; + } else { + edge->face->edge = NULL; + // @@ Remove disconnected face? + } + } + } + // Disconnect from previous. + if (edge->prev) { + if (edge->prev->next == edge) { + edge->prev->setNext(NULL); + } + //edge->setPrev(NULL); + } + // Disconnect from next. + if (edge->next) { + if (edge->next->prev == edge) { + edge->next->setPrev(NULL); + } + //edge->setNext(NULL); + } + } + + void remove(Edge *edge) { + xaDebugAssert(edge != NULL); + disconnect(edge); + delete edge; + } + + void remove(Vertex *vertex) { + xaDebugAssert(vertex != NULL); + // Remove from vertex list. + m_vertexArray[vertex->id] = NULL; + // Disconnect from colocals. + vertex->unlinkColocal(); + // Disconnect from edges. + if (vertex->edge != NULL) { + // @@ Removing a connected vertex is asking for trouble... + if (vertex->edge->vertex == vertex) { + // @@ Connect edge to a colocal? + vertex->edge->vertex = NULL; + } + vertex->setEdge(NULL); + } + delete vertex; + } + + void remove(Face *face) { + xaDebugAssert(face != NULL); + // Remove from face list. + m_faceArray[face->id] = NULL; + // Disconnect from edges. + if (face->edge != NULL) { + xaDebugAssert(face->edge->face == face); + face->edge->face = NULL; + face->edge = NULL; + } + delete face; + } + + // Triangulate in place. + void triangulate() { + bool all_triangles = true; + const uint32_t faceCount = m_faceArray.size(); + for (uint32_t f = 0; f < faceCount; f++) { + Face *face = m_faceArray[f]; + if (face->edgeCount() != 3) { + all_triangles = false; + break; + } + } + if (all_triangles) { + return; + } + // Do not touch vertices, but rebuild edges and faces. + std::vector<Edge *> edgeArray; + std::vector<Face *> faceArray; + std::swap(edgeArray, m_edgeArray); + std::swap(faceArray, m_faceArray); + m_edgeMap.clear(); + for (uint32_t f = 0; f < faceCount; f++) { + Face *face = faceArray[f]; + // Trivial fan-like triangulation. + const uint32_t v0 = face->edge->vertex->id; + uint32_t v2, v1 = (uint32_t)-1; + for (Face::EdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + Edge *edge = it.current(); + v2 = edge->to()->id; + if (v2 == v0) break; + if (v1 != -1) addFace(v0, v1, v2); + v1 = v2; + } + } + xaDebugAssert(m_faceArray.size() > faceCount); // triangle count > face count + linkBoundary(); + for (size_t i = 0; i < edgeArray.size(); i++) + delete edgeArray[i]; + for (size_t i = 0; i < faceArray.size(); i++) + delete faceArray[i]; + } + + /// Link boundary edges once the mesh has been created. + void linkBoundary() { + xaPrint("--- Linking boundaries:\n"); + int num = 0; + // Create boundary edges. + uint32_t edgeCount = this->edgeCount(); + for (uint32_t e = 0; e < edgeCount; e++) { + Edge *edge = edgeAt(e); + if (edge != NULL && edge->pair == NULL) { + Edge *pair = new Edge(edge->id + 1); + uint32_t i = edge->from()->id; + uint32_t j = edge->next->from()->id; + Key key(j, i); + xaAssert(m_edgeMap.find(key) == m_edgeMap.end()); + pair->vertex = m_vertexArray[j]; + m_edgeMap[key] = pair; + edge->pair = pair; + pair->pair = edge; + num++; + } + } + // Link boundary edges. + for (uint32_t e = 0; e < edgeCount; e++) { + Edge *edge = edgeAt(e); + if (edge != NULL && edge->pair->face == NULL) { + linkBoundaryEdge(edge->pair); + } + } + xaPrint("--- %d boundary edges.\n", num); + } + + /* + Fixing T-junctions. + + - Find T-junctions. Find vertices that are on an edge. + - This test is approximate. + - Insert edges on a spatial index to speedup queries. + - Consider only open edges, that is edges that have no pairs. + - Consider only vertices on boundaries. + - Close T-junction. + - Split edge. + + */ + bool splitBoundaryEdges() // Returns true if any split was made. + { + std::vector<Vertex *> boundaryVertices; + for (uint32_t i = 0; i < m_vertexArray.size(); i++) { + Vertex *v = m_vertexArray[i]; + if (v->isBoundary()) { + boundaryVertices.push_back(v); + } + } + xaPrint("Fixing T-junctions:\n"); + int splitCount = 0; + for (uint32_t v = 0; v < boundaryVertices.size(); v++) { + Vertex *vertex = boundaryVertices[v]; + Vector3 x0 = vertex->pos; + // Find edges that this vertex overlaps with. + for (uint32_t e = 0; e < m_edgeArray.size(); e++) { + Edge *edge = m_edgeArray[e]; + if (edge != NULL && edge->isBoundary()) { + if (edge->from() == vertex || edge->to() == vertex) { + continue; + } + Vector3 x1 = edge->from()->pos; + Vector3 x2 = edge->to()->pos; + Vector3 v01 = x0 - x1; + Vector3 v21 = x2 - x1; + float l = length(v21); + float d = length(cross(v01, v21)) / l; + if (isZero(d)) { + float t = dot(v01, v21) / (l * l); + if (t > 0.0f + NV_EPSILON && t < 1.0f - NV_EPSILON) { + xaDebugAssert(equal(lerp(x1, x2, t), x0)); + Vertex *splitVertex = splitBoundaryEdge(edge, t, x0); + vertex->linkColocal(splitVertex); // @@ Should we do this here? + splitCount++; + } + } + } + } + } + xaPrint(" - %d edges split.\n", splitCount); + xaDebugAssert(isValid()); + return splitCount != 0; + } + + // Vertices + uint32_t vertexCount() const { + return m_vertexArray.size(); + } + const Vertex *vertexAt(int i) const { + return m_vertexArray[i]; + } + Vertex *vertexAt(int i) { + return m_vertexArray[i]; + } + + uint32_t colocalVertexCount() const { + return m_colocalVertexCount; + } + + // Faces + uint32_t faceCount() const { + return m_faceArray.size(); + } + const Face *faceAt(int i) const { + return m_faceArray[i]; + } + Face *faceAt(int i) { + return m_faceArray[i]; + } + + // Edges + uint32_t edgeCount() const { + return m_edgeArray.size(); + } + const Edge *edgeAt(int i) const { + return m_edgeArray[i]; + } + Edge *edgeAt(int i) { + return m_edgeArray[i]; + } + + class ConstVertexIterator; + + class VertexIterator { + friend class ConstVertexIterator; + + public: + VertexIterator(Mesh *mesh) : + m_mesh(mesh), + m_current(0) {} + + virtual void advance() { + m_current++; + } + virtual bool isDone() const { + return m_current == m_mesh->vertexCount(); + } + virtual Vertex *current() const { + return m_mesh->vertexAt(m_current); + } + + private: + halfedge::Mesh *m_mesh; + uint32_t m_current; + }; + VertexIterator vertices() { + return VertexIterator(this); + } + + class ConstVertexIterator { + public: + ConstVertexIterator(const Mesh *mesh) : + m_mesh(mesh), + m_current(0) {} + ConstVertexIterator(class VertexIterator &it) : + m_mesh(it.m_mesh), + m_current(it.m_current) {} + + virtual void advance() { + m_current++; + } + virtual bool isDone() const { + return m_current == m_mesh->vertexCount(); + } + virtual const Vertex *current() const { + return m_mesh->vertexAt(m_current); + } + + private: + const halfedge::Mesh *m_mesh; + uint32_t m_current; + }; + ConstVertexIterator vertices() const { + return ConstVertexIterator(this); + } + + class ConstFaceIterator; + + class FaceIterator { + friend class ConstFaceIterator; + + public: + FaceIterator(Mesh *mesh) : + m_mesh(mesh), + m_current(0) {} + + virtual void advance() { + m_current++; + } + virtual bool isDone() const { + return m_current == m_mesh->faceCount(); + } + virtual Face *current() const { + return m_mesh->faceAt(m_current); + } + + private: + halfedge::Mesh *m_mesh; + uint32_t m_current; + }; + FaceIterator faces() { + return FaceIterator(this); + } + + class ConstFaceIterator { + public: + ConstFaceIterator(const Mesh *mesh) : + m_mesh(mesh), + m_current(0) {} + ConstFaceIterator(const FaceIterator &it) : + m_mesh(it.m_mesh), + m_current(it.m_current) {} + + virtual void advance() { + m_current++; + } + virtual bool isDone() const { + return m_current == m_mesh->faceCount(); + } + virtual const Face *current() const { + return m_mesh->faceAt(m_current); + } + + private: + const halfedge::Mesh *m_mesh; + uint32_t m_current; + }; + ConstFaceIterator faces() const { + return ConstFaceIterator(this); + } + + class ConstEdgeIterator; + + class EdgeIterator { + friend class ConstEdgeIterator; + + public: + EdgeIterator(Mesh *mesh) : + m_mesh(mesh), + m_current(0) {} + + virtual void advance() { + m_current++; + } + virtual bool isDone() const { + return m_current == m_mesh->edgeCount(); + } + virtual Edge *current() const { + return m_mesh->edgeAt(m_current); + } + + private: + halfedge::Mesh *m_mesh; + uint32_t m_current; + }; + EdgeIterator edges() { + return EdgeIterator(this); + } + + class ConstEdgeIterator { + public: + ConstEdgeIterator(const Mesh *mesh) : + m_mesh(mesh), + m_current(0) {} + ConstEdgeIterator(const EdgeIterator &it) : + m_mesh(it.m_mesh), + m_current(it.m_current) {} + + virtual void advance() { + m_current++; + } + virtual bool isDone() const { + return m_current == m_mesh->edgeCount(); + } + virtual const Edge *current() const { + return m_mesh->edgeAt(m_current); + } + + private: + const halfedge::Mesh *m_mesh; + uint32_t m_current; + }; + ConstEdgeIterator edges() const { + return ConstEdgeIterator(this); + } + + // @@ Add half-edge iterator. + + bool isValid() const { + // Make sure all edges are valid. + const uint32_t edgeCount = m_edgeArray.size(); + for (uint32_t e = 0; e < edgeCount; e++) { + Edge *edge = m_edgeArray[e]; + if (edge != NULL) { + if (edge->id != 2 * e) { + return false; + } + if (!edge->isValid()) { + return false; + } + if (edge->pair->id != 2 * e + 1) { + return false; + } + if (!edge->pair->isValid()) { + return false; + } + } + } + // @@ Make sure all faces are valid. + // @@ Make sure all vertices are valid. + return true; + } + + // Error status: + + struct ErrorCode { + enum Enum { + AlreadyAddedEdge, + DegenerateColocalEdge, + DegenerateEdge, + DuplicateEdge + }; + }; + + mutable ErrorCode::Enum errorCode; + mutable uint32_t errorIndex0; + mutable uint32_t errorIndex1; + +private: + // Return true if the face can be added to the manifold mesh. + bool canAddFace(const std::vector<uint32_t> &indexArray, uint32_t first, uint32_t num) const { + return canAddFace(indexArray.data(), first, num); + } + + bool canAddFace(const uint32_t *indexArray, uint32_t first, uint32_t num) const { + for (uint32_t j = num - 1, i = 0; i < num; j = i++) { + if (!canAddEdge(indexArray[first + j], indexArray[first + i])) { + errorIndex0 = indexArray[first + j]; + errorIndex1 = indexArray[first + i]; + return false; + } + } + // We also have to make sure the face does not have any duplicate edge! + for (uint32_t i = 0; i < num; i++) { + int i0 = indexArray[first + i + 0]; + int i1 = indexArray[first + (i + 1) % num]; + for (uint32_t j = i + 1; j < num; j++) { + int j0 = indexArray[first + j + 0]; + int j1 = indexArray[first + (j + 1) % num]; + if (i0 == j0 && i1 == j1) { + errorCode = ErrorCode::DuplicateEdge; + errorIndex0 = i0; + errorIndex1 = i1; + return false; + } + } + } + return true; + } + + // Return true if the edge doesn't exist or doesn't have any adjacent face. + bool canAddEdge(uint32_t i, uint32_t j) const { + if (i == j) { + // Skip degenerate edges. + errorCode = ErrorCode::DegenerateEdge; + return false; + } + // Same check, but taking into account colocal vertices. + const Vertex *v0 = vertexAt(i); + const Vertex *v1 = vertexAt(j); + for (Vertex::ConstVertexIterator it(v0->colocals()); !it.isDone(); it.advance()) { + if (it.current() == v1) { + // Skip degenerate edges. + errorCode = ErrorCode::DegenerateColocalEdge; + return false; + } + } + // Make sure edge has not been added yet. + Edge *edge = findEdge(i, j); + // We ignore edges that don't have an adjacent face yet, since this face could become the edge's face. + if (!(edge == NULL || edge->face == NULL)) { + errorCode = ErrorCode::AlreadyAddedEdge; + return false; + } + return true; + } + + Edge *addEdge(uint32_t i, uint32_t j) { + xaAssert(i != j); + Edge *edge = findEdge(i, j); + if (edge != NULL) { + // Edge may already exist, but its face must not be set. + xaDebugAssert(edge->face == NULL); + // Nothing else to do! + } else { + // Add new edge. + // Lookup pair. + Edge *pair = findEdge(j, i); + if (pair != NULL) { + // Create edge with same id. + edge = new Edge(pair->id + 1); + // Link edge pairs. + edge->pair = pair; + pair->pair = edge; + // @@ I'm not sure this is necessary! + pair->vertex->setEdge(pair); + } else { + // Create edge. + edge = new Edge(2 * m_edgeArray.size()); + // Add only unpaired edges. + m_edgeArray.push_back(edge); + } + edge->vertex = m_vertexArray[i]; + m_edgeMap[Key(i, j)] = edge; + } + // Face and Next are set by addFace. + return edge; + } + + /// Find edge, test all colocals. + Edge *findEdge(uint32_t i, uint32_t j) const { + Edge *edge = NULL; + const Vertex *v0 = vertexAt(i); + const Vertex *v1 = vertexAt(j); + // Test all colocal pairs. + for (Vertex::ConstVertexIterator it0(v0->colocals()); !it0.isDone(); it0.advance()) { + for (Vertex::ConstVertexIterator it1(v1->colocals()); !it1.isDone(); it1.advance()) { + Key key(it0.current()->id, it1.current()->id); + if (edge == NULL) { + auto edgeIt = m_edgeMap.find(key); + if (edgeIt != m_edgeMap.end()) + edge = (*edgeIt).second; +#if !defined(_DEBUG) + if (edge != NULL) return edge; +#endif + } else { + // Make sure that only one edge is found. + xaDebugAssert(m_edgeMap.find(key) == m_edgeMap.end()); + } + } + } + return edge; + } + + /// Link this boundary edge. + void linkBoundaryEdge(Edge *edge) { + xaAssert(edge->face == NULL); + // Make sure next pointer has not been set. @@ We want to be able to relink boundary edges after mesh changes. + Edge *next = edge; + while (next->pair->face != NULL) { + // Get pair prev + Edge *e = next->pair->next; + while (e->next != next->pair) { + e = e->next; + } + next = e; + } + edge->setNext(next->pair); + // Adjust vertex edge, so that it's the boundary edge. (required for isBoundary()) + if (edge->vertex->edge != edge) { + // Multiple boundaries in the same edge. + edge->vertex->edge = edge; + } + } + + Vertex *splitBoundaryEdge(Edge *edge, float t, const Vector3 &pos) { + /* + We want to go from this configuration: + + + + + | ^ + edge |<->| pair + v | + + + + + To this one: + + + + + | ^ + e0 |<->| p0 + v | + vertex + + + | ^ + e1 |<->| p1 + v | + + + + + */ + Edge *pair = edge->pair; + // Make sure boundaries are linked. + xaDebugAssert(pair != NULL); + // Make sure edge is a boundary edge. + xaDebugAssert(pair->face == NULL); + // Add new vertex. + Vertex *vertex = addVertex(pos); + vertex->nor = lerp(edge->from()->nor, edge->to()->nor, t); + vertex->tex = lerp(edge->from()->tex, edge->to()->tex, t); + disconnect(edge); + disconnect(pair); + // Add edges. + Edge *e0 = addEdge(edge->from()->id, vertex->id); + Edge *p0 = addEdge(vertex->id, pair->to()->id); + Edge *e1 = addEdge(vertex->id, edge->to()->id); + Edge *p1 = addEdge(pair->from()->id, vertex->id); + // Link edges. + e0->setNext(e1); + p1->setNext(p0); + e0->setPrev(edge->prev); + e1->setNext(edge->next); + p1->setPrev(pair->prev); + p0->setNext(pair->next); + xaDebugAssert(e0->next == e1); + xaDebugAssert(e1->prev == e0); + xaDebugAssert(p1->next == p0); + xaDebugAssert(p0->prev == p1); + xaDebugAssert(p0->pair == e0); + xaDebugAssert(e0->pair == p0); + xaDebugAssert(p1->pair == e1); + xaDebugAssert(e1->pair == p1); + // Link faces. + e0->face = edge->face; + e1->face = edge->face; + // Link vertices. + edge->from()->setEdge(e0); + vertex->setEdge(e1); + delete edge; + delete pair; + return vertex; + } + +private: + std::vector<Vertex *> m_vertexArray; + std::vector<Edge *> m_edgeArray; + std::vector<Face *> m_faceArray; + + struct Key { + Key() {} + Key(const Key &k) : + p0(k.p0), + p1(k.p1) {} + Key(uint32_t v0, uint32_t v1) : + p0(v0), + p1(v1) {} + void operator=(const Key &k) { + p0 = k.p0; + p1 = k.p1; + } + bool operator==(const Key &k) const { + return p0 == k.p0 && p1 == k.p1; + } + + uint32_t p0; + uint32_t p1; + }; + + friend struct Hash<Mesh::Key>; + std::unordered_map<Key, Edge *, Hash<Key>, Equal<Key> > m_edgeMap; + uint32_t m_colocalVertexCount; +}; + +class MeshTopology { +public: + MeshTopology(const Mesh *mesh) { + buildTopologyInfo(mesh); + } + + /// Determine if the mesh is connected. + bool isConnected() const { + return m_connectedCount == 1; + } + + /// Determine if the mesh is closed. (Each edge is shared by two faces) + bool isClosed() const { + return m_boundaryCount == 0; + } + + /// Return true if the mesh has the topology of a disk. + bool isDisk() const { + return isConnected() && m_boundaryCount == 1 /* && m_eulerNumber == 1*/; + } + +private: + void buildTopologyInfo(const Mesh *mesh) { + const uint32_t vertexCount = mesh->colocalVertexCount(); + const uint32_t faceCount = mesh->faceCount(); + const uint32_t edgeCount = mesh->edgeCount(); + xaPrint("--- Building mesh topology:\n"); + std::vector<uint32_t> stack(faceCount); + BitArray bitFlags(faceCount); + bitFlags.clearAll(); + // Compute connectivity. + xaPrint("--- Computing connectivity.\n"); + m_connectedCount = 0; + for (uint32_t f = 0; f < faceCount; f++) { + if (bitFlags.bitAt(f) == false) { + m_connectedCount++; + stack.push_back(f); + while (!stack.empty()) { + const uint32_t top = stack.back(); + xaAssert(top != uint32_t(~0)); + stack.pop_back(); + if (bitFlags.bitAt(top) == false) { + bitFlags.setBitAt(top); + const Face *face = mesh->faceAt(top); + const Edge *firstEdge = face->edge; + const Edge *edge = firstEdge; + do { + const Face *neighborFace = edge->pair->face; + if (neighborFace != NULL) { + stack.push_back(neighborFace->id); + } + edge = edge->next; + } while (edge != firstEdge); + } + } + } + } + xaAssert(stack.empty()); + xaPrint("--- %d connected components.\n", m_connectedCount); + // Count boundary loops. + xaPrint("--- Counting boundary loops.\n"); + m_boundaryCount = 0; + bitFlags.resize(edgeCount); + bitFlags.clearAll(); + // Don't forget to link the boundary otherwise this won't work. + for (uint32_t e = 0; e < edgeCount; e++) { + const Edge *startEdge = mesh->edgeAt(e); + if (startEdge != NULL && startEdge->isBoundary() && bitFlags.bitAt(e) == false) { + xaDebugAssert(startEdge->face != NULL); + xaDebugAssert(startEdge->pair->face == NULL); + startEdge = startEdge->pair; + m_boundaryCount++; + const Edge *edge = startEdge; + do { + bitFlags.setBitAt(edge->id / 2); + edge = edge->next; + } while (startEdge != edge); + } + } + xaPrint("--- %d boundary loops found.\n", m_boundaryCount); + // Compute euler number. + m_eulerNumber = vertexCount - edgeCount + faceCount; + xaPrint("--- Euler number: %d.\n", m_eulerNumber); + // Compute genus. (only valid on closed connected surfaces) + m_genus = -1; + if (isClosed() && isConnected()) { + m_genus = (2 - m_eulerNumber) / 2; + xaPrint("--- Genus: %d.\n", m_genus); + } + } + +private: + ///< Number of boundary loops. + int m_boundaryCount; + + ///< Number of connected components. + int m_connectedCount; + + ///< Euler number. + int m_eulerNumber; + + /// Mesh genus. + int m_genus; +}; + +float computeSurfaceArea(const halfedge::Mesh *mesh) { + float area = 0; + for (halfedge::Mesh::ConstFaceIterator it(mesh->faces()); !it.isDone(); it.advance()) { + const halfedge::Face *face = it.current(); + area += face->area(); + } + xaDebugAssert(area >= 0); + return area; +} + +float computeParametricArea(const halfedge::Mesh *mesh) { + float area = 0; + for (halfedge::Mesh::ConstFaceIterator it(mesh->faces()); !it.isDone(); it.advance()) { + const halfedge::Face *face = it.current(); + area += face->parametricArea(); + } + return area; +} + +uint32_t countMeshTriangles(const Mesh *mesh) { + const uint32_t faceCount = mesh->faceCount(); + uint32_t triangleCount = 0; + for (uint32_t f = 0; f < faceCount; f++) { + const Face *face = mesh->faceAt(f); + uint32_t edgeCount = face->edgeCount(); + xaDebugAssert(edgeCount > 2); + triangleCount += edgeCount - 2; + } + return triangleCount; +} + +Mesh *unifyVertices(const Mesh *inputMesh) { + Mesh *mesh = new Mesh; + // Only add the first colocal. + const uint32_t vertexCount = inputMesh->vertexCount(); + for (uint32_t v = 0; v < vertexCount; v++) { + const Vertex *vertex = inputMesh->vertexAt(v); + if (vertex->isFirstColocal()) { + mesh->addVertex(vertex->pos); + } + } + std::vector<uint32_t> indexArray; + // Add new faces pointing to first colocals. + uint32_t faceCount = inputMesh->faceCount(); + for (uint32_t f = 0; f < faceCount; f++) { + const Face *face = inputMesh->faceAt(f); + indexArray.clear(); + for (Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const Edge *edge = it.current(); + const Vertex *vertex = edge->vertex->firstColocal(); + indexArray.push_back(vertex->id); + } + mesh->addFace(indexArray); + } + mesh->linkBoundary(); + return mesh; +} + +static bool pointInTriangle(const Vector2 &p, const Vector2 &a, const Vector2 &b, const Vector2 &c) { + return triangleArea(a, b, p) >= 0.00001f && + triangleArea(b, c, p) >= 0.00001f && + triangleArea(c, a, p) >= 0.00001f; +} + +// This is doing a simple ear-clipping algorithm that skips invalid triangles. Ideally, we should +// also sort the ears by angle, start with the ones that have the smallest angle and proceed in order. +Mesh *triangulate(const Mesh *inputMesh) { + Mesh *mesh = new Mesh; + // Add all vertices. + const uint32_t vertexCount = inputMesh->vertexCount(); + for (uint32_t v = 0; v < vertexCount; v++) { + const Vertex *vertex = inputMesh->vertexAt(v); + mesh->addVertex(vertex->pos); + } + std::vector<int> polygonVertices; + std::vector<float> polygonAngles; + std::vector<Vector2> polygonPoints; + const uint32_t faceCount = inputMesh->faceCount(); + for (uint32_t f = 0; f < faceCount; f++) { + const Face *face = inputMesh->faceAt(f); + xaDebugAssert(face != NULL); + const uint32_t edgeCount = face->edgeCount(); + xaDebugAssert(edgeCount >= 3); + polygonVertices.clear(); + polygonVertices.reserve(edgeCount); + if (edgeCount == 3) { + // Simple case for triangles. + for (Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const Edge *edge = it.current(); + const Vertex *vertex = edge->vertex; + polygonVertices.push_back(vertex->id); + } + int v0 = polygonVertices[0]; + int v1 = polygonVertices[1]; + int v2 = polygonVertices[2]; + mesh->addFace(v0, v1, v2); + } else { + // Build 2D polygon projecting vertices onto normal plane. + // Faces are not necesarily planar, this is for example the case, when the face comes from filling a hole. In such cases + // it's much better to use the best fit plane. + const Vector3 fn = face->normal(); + Basis basis; + basis.buildFrameForDirection(fn); + polygonPoints.clear(); + polygonPoints.reserve(edgeCount); + polygonAngles.clear(); + polygonAngles.reserve(edgeCount); + for (Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const Edge *edge = it.current(); + const Vertex *vertex = edge->vertex; + polygonVertices.push_back(vertex->id); + Vector2 p; + p.x = dot(basis.tangent, vertex->pos); + p.y = dot(basis.bitangent, vertex->pos); + polygonPoints.push_back(p); + } + polygonAngles.resize(edgeCount); + while (polygonVertices.size() > 2) { + uint32_t size = polygonVertices.size(); + // Update polygon angles. @@ Update only those that have changed. + float minAngle = 2 * PI; + uint32_t bestEar = 0; // Use first one if none of them is valid. + bool bestIsValid = false; + for (uint32_t i = 0; i < size; i++) { + uint32_t i0 = i; + uint32_t i1 = (i + 1) % size; // Use Sean's polygon interation trick. + uint32_t i2 = (i + 2) % size; + Vector2 p0 = polygonPoints[i0]; + Vector2 p1 = polygonPoints[i1]; + Vector2 p2 = polygonPoints[i2]; + + bool degenerate = distance(p0, p1) < NV_EPSILON || distance(p0, p2) < NV_EPSILON || distance(p1, p2) < NV_EPSILON; + if (degenerate) { + continue; + } + + float d = clamp(dot(p0 - p1, p2 - p1) / (length(p0 - p1) * length(p2 - p1)), -1.0f, 1.0f); + float angle = acosf(d); + float area = triangleArea(p0, p1, p2); + if (area < 0.0f) angle = 2.0f * PI - angle; + polygonAngles[i1] = angle; + if (angle < minAngle || !bestIsValid) { + // Make sure this is a valid ear, if not, skip this point. + bool valid = true; + for (uint32_t j = 0; j < size; j++) { + if (j == i0 || j == i1 || j == i2) continue; + Vector2 p = polygonPoints[j]; + if (pointInTriangle(p, p0, p1, p2)) { + valid = false; + break; + } + } + if (valid || !bestIsValid) { + minAngle = angle; + bestEar = i1; + bestIsValid = valid; + } + } + } + if (!bestIsValid) + break; + + xaDebugAssert(minAngle <= 2 * PI); + // Clip best ear: + uint32_t i0 = (bestEar + size - 1) % size; + uint32_t i1 = (bestEar + 0) % size; + uint32_t i2 = (bestEar + 1) % size; + int v0 = polygonVertices[i0]; + int v1 = polygonVertices[i1]; + int v2 = polygonVertices[i2]; + mesh->addFace(v0, v1, v2); + polygonVertices.erase(polygonVertices.begin() + i1); + polygonPoints.erase(polygonPoints.begin() + i1); + polygonAngles.erase(polygonAngles.begin() + i1); + } + } + } + mesh->linkBoundary(); + return mesh; +} + +} // namespace halfedge + +/// Mersenne twister random number generator. +class MTRand { +public: + enum time_e { Time }; + enum { N = 624 }; // length of state vector + enum { M = 397 }; + + /// Constructor that uses the current time as the seed. + MTRand(time_e) { + seed((uint32_t)time(NULL)); + } + + /// Constructor that uses the given seed. + MTRand(uint32_t s = 0) { + seed(s); + } + + /// Provide a new seed. + void seed(uint32_t s) { + initialize(s); + reload(); + } + + /// Get a random number between 0 - 65536. + uint32_t get() { + // Pull a 32-bit integer from the generator state + // Every other access function simply transforms the numbers extracted here + if (left == 0) { + reload(); + } + left--; + uint32_t s1; + s1 = *next++; + s1 ^= (s1 >> 11); + s1 ^= (s1 << 7) & 0x9d2c5680U; + s1 ^= (s1 << 15) & 0xefc60000U; + return (s1 ^ (s1 >> 18)); + }; + + /// Get a random number on [0, max] interval. + uint32_t getRange(uint32_t max) { + if (max == 0) return 0; + if (max == NV_UINT32_MAX) return get(); + const uint32_t np2 = nextPowerOfTwo(max + 1); // @@ This fails if max == NV_UINT32_MAX + const uint32_t mask = np2 - 1; + uint32_t n; + do { + n = get() & mask; + } while (n > max); + return n; + } + +private: + void initialize(uint32_t seed) { + // Initialize generator state with seed + // See Knuth TAOCP Vol 2, 3rd Ed, p.106 for multiplier. + // In previous versions, most significant bits (MSBs) of the seed affect + // only MSBs of the state array. Modified 9 Jan 2002 by Makoto Matsumoto. + uint32_t *s = state; + uint32_t *r = state; + int i = 1; + *s++ = seed & 0xffffffffUL; + for (; i < N; ++i) { + *s++ = (1812433253UL * (*r ^ (*r >> 30)) + i) & 0xffffffffUL; + r++; + } + } + + void reload() { + // Generate N new values in state + // Made clearer and faster by Matthew Bellew (matthew.bellew@home.com) + uint32_t *p = state; + int i; + for (i = N - M; i--; ++p) + *p = twist(p[M], p[0], p[1]); + for (i = M; --i; ++p) + *p = twist(p[M - N], p[0], p[1]); + *p = twist(p[M - N], p[0], state[0]); + left = N, next = state; + } + + uint32_t hiBit(uint32_t u) const { + return u & 0x80000000U; + } + uint32_t loBit(uint32_t u) const { + return u & 0x00000001U; + } + uint32_t loBits(uint32_t u) const { + return u & 0x7fffffffU; + } + uint32_t mixBits(uint32_t u, uint32_t v) const { + return hiBit(u) | loBits(v); + } + uint32_t twist(uint32_t m, uint32_t s0, uint32_t s1) const { + return m ^ (mixBits(s0, s1) >> 1) ^ ((~loBit(s1) + 1) & 0x9908b0dfU); + } + + uint32_t state[N]; // internal state + uint32_t *next; // next value to get from state + int left; // number of values left before reload needed +}; + +namespace morton { +// Code from ryg: +// http://fgiesen.wordpress.com/2009/12/13/decoding-morton-codes/ + +// Inverse of part1By1 - "delete" all odd-indexed bits +uint32_t compact1By1(uint32_t x) { + x &= 0x55555555; // x = -f-e -d-c -b-a -9-8 -7-6 -5-4 -3-2 -1-0 + x = (x ^ (x >> 1)) & 0x33333333; // x = --fe --dc --ba --98 --76 --54 --32 --10 + x = (x ^ (x >> 2)) & 0x0f0f0f0f; // x = ---- fedc ---- ba98 ---- 7654 ---- 3210 + x = (x ^ (x >> 4)) & 0x00ff00ff; // x = ---- ---- fedc ba98 ---- ---- 7654 3210 + x = (x ^ (x >> 8)) & 0x0000ffff; // x = ---- ---- ---- ---- fedc ba98 7654 3210 + return x; +} + +// Inverse of part1By2 - "delete" all bits not at positions divisible by 3 +uint32_t compact1By2(uint32_t x) { + x &= 0x09249249; // x = ---- 9--8 --7- -6-- 5--4 --3- -2-- 1--0 + x = (x ^ (x >> 2)) & 0x030c30c3; // x = ---- --98 ---- 76-- --54 ---- 32-- --10 + x = (x ^ (x >> 4)) & 0x0300f00f; // x = ---- --98 ---- ---- 7654 ---- ---- 3210 + x = (x ^ (x >> 8)) & 0xff0000ff; // x = ---- --98 ---- ---- ---- ---- 7654 3210 + x = (x ^ (x >> 16)) & 0x000003ff; // x = ---- ---- ---- ---- ---- --98 7654 3210 + return x; +} + +uint32_t decodeMorton2X(uint32_t code) { + return compact1By1(code >> 0); +} + +uint32_t decodeMorton2Y(uint32_t code) { + return compact1By1(code >> 1); +} + +uint32_t decodeMorton3X(uint32_t code) { + return compact1By2(code >> 0); +} + +uint32_t decodeMorton3Y(uint32_t code) { + return compact1By2(code >> 1); +} + +uint32_t decodeMorton3Z(uint32_t code) { + return compact1By2(code >> 2); +} +} // namespace morton + +// A simple, dynamic proximity grid based on Jon's code. +// Instead of storing pointers here I store indices. +struct ProximityGrid { + void init(const Box &box, uint32_t count) { + cellArray.clear(); + // Determine grid size. + float cellWidth; + Vector3 diagonal = box.extents() * 2.f; + float volume = box.volume(); + if (equal(volume, 0)) { + // Degenerate box, treat like a quad. + Vector2 quad; + if (diagonal.x < diagonal.y && diagonal.x < diagonal.z) { + quad.x = diagonal.y; + quad.y = diagonal.z; + } else if (diagonal.y < diagonal.x && diagonal.y < diagonal.z) { + quad.x = diagonal.x; + quad.y = diagonal.z; + } else { + quad.x = diagonal.x; + quad.y = diagonal.y; + } + float cellArea = quad.x * quad.y / count; + cellWidth = sqrtf(cellArea); // pow(cellArea, 1.0f / 2.0f); + } else { + // Ideally we want one cell per point. + float cellVolume = volume / count; + cellWidth = powf(cellVolume, 1.0f / 3.0f); + } + xaDebugAssert(cellWidth != 0); + sx = std::max(1, ftoi_ceil(diagonal.x / cellWidth)); + sy = std::max(1, ftoi_ceil(diagonal.y / cellWidth)); + sz = std::max(1, ftoi_ceil(diagonal.z / cellWidth)); + invCellSize.x = float(sx) / diagonal.x; + invCellSize.y = float(sy) / diagonal.y; + invCellSize.z = float(sz) / diagonal.z; + cellArray.resize(sx * sy * sz); + corner = box.minCorner; // @@ Align grid better? + } + + int index_x(float x) const { + return clamp(ftoi_floor((x - corner.x) * invCellSize.x), 0, sx - 1); + } + + int index_y(float y) const { + return clamp(ftoi_floor((y - corner.y) * invCellSize.y), 0, sy - 1); + } + + int index_z(float z) const { + return clamp(ftoi_floor((z - corner.z) * invCellSize.z), 0, sz - 1); + } + + int index(int x, int y, int z) const { + xaDebugAssert(x >= 0 && x < sx); + xaDebugAssert(y >= 0 && y < sy); + xaDebugAssert(z >= 0 && z < sz); + int idx = (z * sy + y) * sx + x; + xaDebugAssert(idx >= 0 && uint32_t(idx) < cellArray.size()); + return idx; + } + + uint32_t mortonCount() const { + uint64_t s = uint64_t(max3(sx, sy, sz)); + s = nextPowerOfTwo(s); + if (s > 1024) { + return uint32_t(s * s * min3(sx, sy, sz)); + } + return uint32_t(s * s * s); + } + + int mortonIndex(uint32_t code) const { + uint32_t x, y, z; + uint32_t s = uint32_t(max3(sx, sy, sz)); + if (s > 1024) { + // Use layered two-dimensional morton order. + s = nextPowerOfTwo(s); + uint32_t layer = code / (s * s); + code = code % (s * s); + uint32_t layer_count = uint32_t(min3(sx, sy, sz)); + if (sx == (int)layer_count) { + x = layer; + y = morton::decodeMorton2X(code); + z = morton::decodeMorton2Y(code); + } else if (sy == (int)layer_count) { + x = morton::decodeMorton2Y(code); + y = layer; + z = morton::decodeMorton2X(code); + } else { /*if (sz == layer_count)*/ + x = morton::decodeMorton2X(code); + y = morton::decodeMorton2Y(code); + z = layer; + } + } else { + x = morton::decodeMorton3X(code); + y = morton::decodeMorton3Y(code); + z = morton::decodeMorton3Z(code); + } + if (x >= uint32_t(sx) || y >= uint32_t(sy) || z >= uint32_t(sz)) { + return -1; + } + return index(x, y, z); + } + + void add(const Vector3 &pos, uint32_t key) { + int x = index_x(pos.x); + int y = index_y(pos.y); + int z = index_z(pos.z); + uint32_t idx = index(x, y, z); + cellArray[idx].indexArray.push_back(key); + } + + // Gather all points inside the given sphere. + // Radius is assumed to be small, so we don't bother culling the cells. + void gather(const Vector3 &position, float radius, std::vector<uint32_t> &indexArray) { + int x0 = index_x(position.x - radius); + int x1 = index_x(position.x + radius); + int y0 = index_y(position.y - radius); + int y1 = index_y(position.y + radius); + int z0 = index_z(position.z - radius); + int z1 = index_z(position.z + radius); + for (int z = z0; z <= z1; z++) { + for (int y = y0; y <= y1; y++) { + for (int x = x0; x <= x1; x++) { + int idx = index(x, y, z); + indexArray.insert(indexArray.begin(), cellArray[idx].indexArray.begin(), cellArray[idx].indexArray.end()); + } + } + } + } + + struct Cell { + std::vector<uint32_t> indexArray; + }; + + std::vector<Cell> cellArray; + + Vector3 corner; + Vector3 invCellSize; + int sx, sy, sz; +}; + +// Based on Pierre Terdiman's and Michael Herf's source code. +// http://www.codercorner.com/RadixSortRevisited.htm +// http://www.stereopsis.com/radix.html +class RadixSort { +public: + RadixSort() : + m_size(0), + m_ranks(NULL), + m_ranks2(NULL), + m_validRanks(false) {} + ~RadixSort() { + // Release everything + free(m_ranks2); + free(m_ranks); + } + + RadixSort &sort(const float *input, uint32_t count) { + if (input == NULL || count == 0) return *this; + // Resize lists if needed + if (count != m_size) { + if (count > m_size) { + m_ranks2 = (uint32_t *)realloc(m_ranks2, sizeof(uint32_t) * count); + m_ranks = (uint32_t *)realloc(m_ranks, sizeof(uint32_t) * count); + } + m_size = count; + m_validRanks = false; + } + if (count < 32) { + insertionSort(input, count); + } else { + // @@ Avoid touching the input multiple times. + for (uint32_t i = 0; i < count; i++) { + FloatFlip((uint32_t &)input[i]); + } + radixSort<uint32_t>((const uint32_t *)input, count); + for (uint32_t i = 0; i < count; i++) { + IFloatFlip((uint32_t &)input[i]); + } + } + return *this; + } + + RadixSort &sort(const std::vector<float> &input) { + return sort(input.data(), input.size()); + } + + // Access to results. m_ranks is a list of indices in sorted order, i.e. in the order you may further process your data + const uint32_t *ranks() const { + xaDebugAssert(m_validRanks); + return m_ranks; + } + uint32_t *ranks() { + xaDebugAssert(m_validRanks); + return m_ranks; + } + +private: + uint32_t m_size; + uint32_t *m_ranks; + uint32_t *m_ranks2; + bool m_validRanks; + + void FloatFlip(uint32_t &f) { + int32_t mask = (int32_t(f) >> 31) | 0x80000000; // Warren Hunt, Manchor Ko. + f ^= mask; + } + + void IFloatFlip(uint32_t &f) { + uint32_t mask = ((f >> 31) - 1) | 0x80000000; // Michael Herf. + f ^= mask; + } + + template <typename T> + void createHistograms(const T *buffer, uint32_t count, uint32_t *histogram) { + const uint32_t bucketCount = sizeof(T); // (8 * sizeof(T)) / log2(radix) + // Init bucket pointers. + uint32_t *h[bucketCount]; + for (uint32_t i = 0; i < bucketCount; i++) { + h[i] = histogram + 256 * i; + } + // Clear histograms. + memset(histogram, 0, 256 * bucketCount * sizeof(uint32_t)); + // @@ Add support for signed integers. + // Build histograms. + const uint8_t *p = (const uint8_t *)buffer; // @@ Does this break aliasing rules? + const uint8_t *pe = p + count * sizeof(T); + while (p != pe) { + h[0][*p++]++, h[1][*p++]++, h[2][*p++]++, h[3][*p++]++; +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable : 4127) +#endif + if (bucketCount == 8) h[4][*p++]++, h[5][*p++]++, h[6][*p++]++, h[7][*p++]++; +#ifdef _MSC_VER +#pragma warning(pop) +#endif + } + } + + template <typename T> + void insertionSort(const T *input, uint32_t count) { + if (!m_validRanks) { + m_ranks[0] = 0; + for (uint32_t i = 1; i != count; ++i) { + int rank = m_ranks[i] = i; + uint32_t j = i; + while (j != 0 && input[rank] < input[m_ranks[j - 1]]) { + m_ranks[j] = m_ranks[j - 1]; + --j; + } + if (i != j) { + m_ranks[j] = rank; + } + } + m_validRanks = true; + } else { + for (uint32_t i = 1; i != count; ++i) { + int rank = m_ranks[i]; + uint32_t j = i; + while (j != 0 && input[rank] < input[m_ranks[j - 1]]) { + m_ranks[j] = m_ranks[j - 1]; + --j; + } + if (i != j) { + m_ranks[j] = rank; + } + } + } + } + + template <typename T> + void radixSort(const T *input, uint32_t count) { + const uint32_t P = sizeof(T); // pass count + // Allocate histograms & offsets on the stack + uint32_t histogram[256 * P]; + uint32_t *link[256]; + createHistograms(input, count, histogram); + // Radix sort, j is the pass number (0=LSB, P=MSB) + for (uint32_t j = 0; j < P; j++) { + // Pointer to this bucket. + const uint32_t *h = &histogram[j * 256]; + const uint8_t *inputBytes = (const uint8_t *)input; // @@ Is this aliasing legal? + inputBytes += j; + if (h[inputBytes[0]] == count) { + // Skip this pass, all values are the same. + continue; + } + // Create offsets + link[0] = m_ranks2; + for (uint32_t i = 1; i < 256; i++) + link[i] = link[i - 1] + h[i - 1]; + // Perform Radix Sort + if (!m_validRanks) { + for (uint32_t i = 0; i < count; i++) { + *link[inputBytes[i * P]]++ = i; + } + m_validRanks = true; + } else { + for (uint32_t i = 0; i < count; i++) { + const uint32_t idx = m_ranks[i]; + *link[inputBytes[idx * P]]++ = idx; + } + } + // Swap pointers for next pass. Valid indices - the most recent ones - are in m_ranks after the swap. + std::swap(m_ranks, m_ranks2); + } + // All values were equal, generate linear ranks. + if (!m_validRanks) { + for (uint32_t i = 0; i < count; i++) { + m_ranks[i] = i; + } + m_validRanks = true; + } + } +}; + +namespace raster { +class ClippedTriangle { +public: + ClippedTriangle(Vector2::Arg a, Vector2::Arg b, Vector2::Arg c) { + m_numVertices = 3; + m_activeVertexBuffer = 0; + m_verticesA[0] = a; + m_verticesA[1] = b; + m_verticesA[2] = c; + m_vertexBuffers[0] = m_verticesA; + m_vertexBuffers[1] = m_verticesB; + } + + uint32_t vertexCount() { + return m_numVertices; + } + + const Vector2 *vertices() { + return m_vertexBuffers[m_activeVertexBuffer]; + } + + void clipHorizontalPlane(float offset, float clipdirection) { + Vector2 *v = m_vertexBuffers[m_activeVertexBuffer]; + m_activeVertexBuffer ^= 1; + Vector2 *v2 = m_vertexBuffers[m_activeVertexBuffer]; + v[m_numVertices] = v[0]; + float dy2, dy1 = offset - v[0].y; + int dy2in, dy1in = clipdirection * dy1 >= 0; + uint32_t p = 0; + for (uint32_t k = 0; k < m_numVertices; k++) { + dy2 = offset - v[k + 1].y; + dy2in = clipdirection * dy2 >= 0; + if (dy1in) v2[p++] = v[k]; + if (dy1in + dy2in == 1) { // not both in/out + float dx = v[k + 1].x - v[k].x; + float dy = v[k + 1].y - v[k].y; + v2[p++] = Vector2(v[k].x + dy1 * (dx / dy), offset); + } + dy1 = dy2; + dy1in = dy2in; + } + m_numVertices = p; + //for (uint32_t k=0; k<m_numVertices; k++) printf("(%f, %f)\n", v2[k].x, v2[k].y); printf("\n"); + } + + void clipVerticalPlane(float offset, float clipdirection) { + Vector2 *v = m_vertexBuffers[m_activeVertexBuffer]; + m_activeVertexBuffer ^= 1; + Vector2 *v2 = m_vertexBuffers[m_activeVertexBuffer]; + v[m_numVertices] = v[0]; + float dx2, dx1 = offset - v[0].x; + int dx2in, dx1in = clipdirection * dx1 >= 0; + uint32_t p = 0; + for (uint32_t k = 0; k < m_numVertices; k++) { + dx2 = offset - v[k + 1].x; + dx2in = clipdirection * dx2 >= 0; + if (dx1in) v2[p++] = v[k]; + if (dx1in + dx2in == 1) { // not both in/out + float dx = v[k + 1].x - v[k].x; + float dy = v[k + 1].y - v[k].y; + v2[p++] = Vector2(offset, v[k].y + dx1 * (dy / dx)); + } + dx1 = dx2; + dx1in = dx2in; + } + m_numVertices = p; + } + + void computeAreaCentroid() { + Vector2 *v = m_vertexBuffers[m_activeVertexBuffer]; + v[m_numVertices] = v[0]; + m_area = 0; + float centroidx = 0, centroidy = 0; + for (uint32_t k = 0; k < m_numVertices; k++) { + // http://local.wasp.uwa.edu.au/~pbourke/geometry/polyarea/ + float f = v[k].x * v[k + 1].y - v[k + 1].x * v[k].y; + m_area += f; + centroidx += f * (v[k].x + v[k + 1].x); + centroidy += f * (v[k].y + v[k + 1].y); + } + m_area = 0.5f * fabsf(m_area); + if (m_area == 0) { + m_centroid = Vector2(0.0f); + } else { + m_centroid = Vector2(centroidx / (6 * m_area), centroidy / (6 * m_area)); + } + } + + void clipAABox(float x0, float y0, float x1, float y1) { + clipVerticalPlane(x0, -1); + clipHorizontalPlane(y0, -1); + clipVerticalPlane(x1, 1); + clipHorizontalPlane(y1, 1); + computeAreaCentroid(); + } + + Vector2 centroid() { + return m_centroid; + } + + float area() { + return m_area; + } + +private: + Vector2 m_verticesA[7 + 1]; + Vector2 m_verticesB[7 + 1]; + Vector2 *m_vertexBuffers[2]; + uint32_t m_numVertices; + uint32_t m_activeVertexBuffer; + float m_area; + Vector2 m_centroid; +}; + +/// A callback to sample the environment. Return false to terminate rasterization. +typedef bool (*SamplingCallback)(void *param, int x, int y, Vector3::Arg bar, Vector3::Arg dx, Vector3::Arg dy, float coverage); + +/// A triangle for rasterization. +struct Triangle { + Triangle(Vector2::Arg v0, Vector2::Arg v1, Vector2::Arg v2, Vector3::Arg t0, Vector3::Arg t1, Vector3::Arg t2) { + // Init vertices. + this->v1 = v0; + this->v2 = v2; + this->v3 = v1; + // Set barycentric coordinates. + this->t1 = t0; + this->t2 = t2; + this->t3 = t1; + // make sure every triangle is front facing. + flipBackface(); + // Compute deltas. + valid = computeDeltas(); + computeUnitInwardNormals(); + } + + /// Compute texture space deltas. + /// This method takes two edge vectors that form a basis, determines the + /// coordinates of the canonic vectors in that basis, and computes the + /// texture gradient that corresponds to those vectors. + bool computeDeltas() { + Vector2 e0 = v3 - v1; + Vector2 e1 = v2 - v1; + Vector3 de0 = t3 - t1; + Vector3 de1 = t2 - t1; + float denom = 1.0f / (e0.y * e1.x - e1.y * e0.x); + if (!std::isfinite(denom)) { + return false; + } + float lambda1 = -e1.y * denom; + float lambda2 = e0.y * denom; + float lambda3 = e1.x * denom; + float lambda4 = -e0.x * denom; + dx = de0 * lambda1 + de1 * lambda2; + dy = de0 * lambda3 + de1 * lambda4; + return true; + } + + bool draw(const Vector2 &extents, bool enableScissors, SamplingCallback cb, void *param) { + // 28.4 fixed-point coordinates + const int Y1 = ftoi_round(16.0f * v1.y); + const int Y2 = ftoi_round(16.0f * v2.y); + const int Y3 = ftoi_round(16.0f * v3.y); + const int X1 = ftoi_round(16.0f * v1.x); + const int X2 = ftoi_round(16.0f * v2.x); + const int X3 = ftoi_round(16.0f * v3.x); + // Deltas + const int DX12 = X1 - X2; + const int DX23 = X2 - X3; + const int DX31 = X3 - X1; + const int DY12 = Y1 - Y2; + const int DY23 = Y2 - Y3; + const int DY31 = Y3 - Y1; + // Fixed-point deltas + const int FDX12 = DX12 << 4; + const int FDX23 = DX23 << 4; + const int FDX31 = DX31 << 4; + const int FDY12 = DY12 << 4; + const int FDY23 = DY23 << 4; + const int FDY31 = DY31 << 4; + int minx, miny, maxx, maxy; + if (enableScissors) { + int frustumX0 = 0 << 4; + int frustumY0 = 0 << 4; + int frustumX1 = (int)extents.x << 4; + int frustumY1 = (int)extents.y << 4; + // Bounding rectangle + minx = (std::max(min3(X1, X2, X3), frustumX0) + 0xF) >> 4; + miny = (std::max(min3(Y1, Y2, Y3), frustumY0) + 0xF) >> 4; + maxx = (std::min(max3(X1, X2, X3), frustumX1) + 0xF) >> 4; + maxy = (std::min(max3(Y1, Y2, Y3), frustumY1) + 0xF) >> 4; + } else { + // Bounding rectangle + minx = (min3(X1, X2, X3) + 0xF) >> 4; + miny = (min3(Y1, Y2, Y3) + 0xF) >> 4; + maxx = (max3(X1, X2, X3) + 0xF) >> 4; + maxy = (max3(Y1, Y2, Y3) + 0xF) >> 4; + } + // Block size, standard 8x8 (must be power of two) + const int q = 8; + // @@ This won't work when minx,miny are negative. This code path is not used. Leaving as is for now. + xaAssert(minx >= 0); + xaAssert(miny >= 0); + // Start in corner of 8x8 block + minx &= ~(q - 1); + miny &= ~(q - 1); + // Half-edge constants + int C1 = DY12 * X1 - DX12 * Y1; + int C2 = DY23 * X2 - DX23 * Y2; + int C3 = DY31 * X3 - DX31 * Y3; + // Correct for fill convention + if (DY12 < 0 || (DY12 == 0 && DX12 > 0)) C1++; + if (DY23 < 0 || (DY23 == 0 && DX23 > 0)) C2++; + if (DY31 < 0 || (DY31 == 0 && DX31 > 0)) C3++; + // Loop through blocks + for (int y = miny; y < maxy; y += q) { + for (int x = minx; x < maxx; x += q) { + // Corners of block + int x0 = x << 4; + int x1 = (x + q - 1) << 4; + int y0 = y << 4; + int y1 = (y + q - 1) << 4; + // Evaluate half-space functions + bool a00 = C1 + DX12 * y0 - DY12 * x0 > 0; + bool a10 = C1 + DX12 * y0 - DY12 * x1 > 0; + bool a01 = C1 + DX12 * y1 - DY12 * x0 > 0; + bool a11 = C1 + DX12 * y1 - DY12 * x1 > 0; + int a = (a00 << 0) | (a10 << 1) | (a01 << 2) | (a11 << 3); + bool b00 = C2 + DX23 * y0 - DY23 * x0 > 0; + bool b10 = C2 + DX23 * y0 - DY23 * x1 > 0; + bool b01 = C2 + DX23 * y1 - DY23 * x0 > 0; + bool b11 = C2 + DX23 * y1 - DY23 * x1 > 0; + int b = (b00 << 0) | (b10 << 1) | (b01 << 2) | (b11 << 3); + bool c00 = C3 + DX31 * y0 - DY31 * x0 > 0; + bool c10 = C3 + DX31 * y0 - DY31 * x1 > 0; + bool c01 = C3 + DX31 * y1 - DY31 * x0 > 0; + bool c11 = C3 + DX31 * y1 - DY31 * x1 > 0; + int c = (c00 << 0) | (c10 << 1) | (c01 << 2) | (c11 << 3); + // Skip block when outside an edge + if (a == 0x0 || b == 0x0 || c == 0x0) continue; + // Accept whole block when totally covered + if (a == 0xF && b == 0xF && c == 0xF) { + Vector3 texRow = t1 + dy * (y0 - v1.y) + dx * (x0 - v1.x); + for (int iy = y; iy < y + q; iy++) { + Vector3 tex = texRow; + for (int ix = x; ix < x + q; ix++) { + //Vector3 tex = t1 + dx * (ix - v1.x) + dy * (iy - v1.y); + if (!cb(param, ix, iy, tex, dx, dy, 1.0)) { + // early out. + return false; + } + tex += dx; + } + texRow += dy; + } + } else { // Partially covered block + int CY1 = C1 + DX12 * y0 - DY12 * x0; + int CY2 = C2 + DX23 * y0 - DY23 * x0; + int CY3 = C3 + DX31 * y0 - DY31 * x0; + Vector3 texRow = t1 + dy * (y0 - v1.y) + dx * (x0 - v1.x); + for (int iy = y; iy < y + q; iy++) { + int CX1 = CY1; + int CX2 = CY2; + int CX3 = CY3; + Vector3 tex = texRow; + for (int ix = x; ix < x + q; ix++) { + if (CX1 > 0 && CX2 > 0 && CX3 > 0) { + if (!cb(param, ix, iy, tex, dx, dy, 1.0)) { + // early out. + return false; + } + } + CX1 -= FDY12; + CX2 -= FDY23; + CX3 -= FDY31; + tex += dx; + } + CY1 += FDX12; + CY2 += FDX23; + CY3 += FDX31; + texRow += dy; + } + } + } + } + return true; + } + + // extents has to be multiple of BK_SIZE!! + bool drawAA(const Vector2 &extents, bool enableScissors, SamplingCallback cb, void *param) { + const float PX_INSIDE = 1.0f / sqrt(2.0f); + const float PX_OUTSIDE = -1.0f / sqrt(2.0f); + const float BK_SIZE = 8; + const float BK_INSIDE = sqrt(BK_SIZE * BK_SIZE / 2.0f); + const float BK_OUTSIDE = -sqrt(BK_SIZE * BK_SIZE / 2.0f); + + float minx, miny, maxx, maxy; + if (enableScissors) { + // Bounding rectangle + minx = floorf(std::max(min3(v1.x, v2.x, v3.x), 0.0f)); + miny = floorf(std::max(min3(v1.y, v2.y, v3.y), 0.0f)); + maxx = ceilf(std::min(max3(v1.x, v2.x, v3.x), extents.x - 1.0f)); + maxy = ceilf(std::min(max3(v1.y, v2.y, v3.y), extents.y - 1.0f)); + } else { + // Bounding rectangle + minx = floorf(min3(v1.x, v2.x, v3.x)); + miny = floorf(min3(v1.y, v2.y, v3.y)); + maxx = ceilf(max3(v1.x, v2.x, v3.x)); + maxy = ceilf(max3(v1.y, v2.y, v3.y)); + } + // There's no reason to align the blocks to the viewport, instead we align them to the origin of the triangle bounds. + minx = floorf(minx); + miny = floorf(miny); + //minx = (float)(((int)minx) & (~((int)BK_SIZE - 1))); // align to blocksize (we don't need to worry about blocks partially out of viewport) + //miny = (float)(((int)miny) & (~((int)BK_SIZE - 1))); + minx += 0.5; + miny += 0.5; // sampling at texel centers! + maxx += 0.5; + maxy += 0.5; + // Half-edge constants + float C1 = n1.x * (-v1.x) + n1.y * (-v1.y); + float C2 = n2.x * (-v2.x) + n2.y * (-v2.y); + float C3 = n3.x * (-v3.x) + n3.y * (-v3.y); + // Loop through blocks + for (float y0 = miny; y0 <= maxy; y0 += BK_SIZE) { + for (float x0 = minx; x0 <= maxx; x0 += BK_SIZE) { + // Corners of block + float xc = (x0 + (BK_SIZE - 1) / 2.0f); + float yc = (y0 + (BK_SIZE - 1) / 2.0f); + // Evaluate half-space functions + float aC = C1 + n1.x * xc + n1.y * yc; + float bC = C2 + n2.x * xc + n2.y * yc; + float cC = C3 + n3.x * xc + n3.y * yc; + // Skip block when outside an edge + if ((aC <= BK_OUTSIDE) || (bC <= BK_OUTSIDE) || (cC <= BK_OUTSIDE)) continue; + // Accept whole block when totally covered + if ((aC >= BK_INSIDE) && (bC >= BK_INSIDE) && (cC >= BK_INSIDE)) { + Vector3 texRow = t1 + dy * (y0 - v1.y) + dx * (x0 - v1.x); + for (float y = y0; y < y0 + BK_SIZE; y++) { + Vector3 tex = texRow; + for (float x = x0; x < x0 + BK_SIZE; x++) { + if (!cb(param, (int)x, (int)y, tex, dx, dy, 1.0f)) { + return false; + } + tex += dx; + } + texRow += dy; + } + } else { // Partially covered block + float CY1 = C1 + n1.x * x0 + n1.y * y0; + float CY2 = C2 + n2.x * x0 + n2.y * y0; + float CY3 = C3 + n3.x * x0 + n3.y * y0; + Vector3 texRow = t1 + dy * (y0 - v1.y) + dx * (x0 - v1.x); + for (float y = y0; y < y0 + BK_SIZE; y++) { // @@ This is not clipping to scissor rectangle correctly. + float CX1 = CY1; + float CX2 = CY2; + float CX3 = CY3; + Vector3 tex = texRow; + for (float x = x0; x < x0 + BK_SIZE; x++) { // @@ This is not clipping to scissor rectangle correctly. + if (CX1 >= PX_INSIDE && CX2 >= PX_INSIDE && CX3 >= PX_INSIDE) { + // pixel completely covered + Vector3 tex2 = t1 + dx * (x - v1.x) + dy * (y - v1.y); + if (!cb(param, (int)x, (int)y, tex2, dx, dy, 1.0f)) { + return false; + } + } else if ((CX1 >= PX_OUTSIDE) && (CX2 >= PX_OUTSIDE) && (CX3 >= PX_OUTSIDE)) { + // triangle partially covers pixel. do clipping. + ClippedTriangle ct(v1 - Vector2(x, y), v2 - Vector2(x, y), v3 - Vector2(x, y)); + ct.clipAABox(-0.5, -0.5, 0.5, 0.5); + Vector2 centroid = ct.centroid(); + float area = ct.area(); + if (area > 0.0f) { + Vector3 texCent = tex - dx * centroid.x - dy * centroid.y; + //xaAssert(texCent.x >= -0.1f && texCent.x <= 1.1f); // @@ Centroid is not very exact... + //xaAssert(texCent.y >= -0.1f && texCent.y <= 1.1f); + //xaAssert(texCent.z >= -0.1f && texCent.z <= 1.1f); + //Vector3 texCent2 = t1 + dx * (x - v1.x) + dy * (y - v1.y); + if (!cb(param, (int)x, (int)y, texCent, dx, dy, area)) { + return false; + } + } + } + CX1 += n1.x; + CX2 += n2.x; + CX3 += n3.x; + tex += dx; + } + CY1 += n1.y; + CY2 += n2.y; + CY3 += n3.y; + texRow += dy; + } + } + } + } + return true; + } + + void flipBackface() { + // check if triangle is backfacing, if so, swap two vertices + if (((v3.x - v1.x) * (v2.y - v1.y) - (v3.y - v1.y) * (v2.x - v1.x)) < 0) { + Vector2 hv = v1; + v1 = v2; + v2 = hv; // swap pos + Vector3 ht = t1; + t1 = t2; + t2 = ht; // swap tex + } + } + + // compute unit inward normals for each edge. + void computeUnitInwardNormals() { + n1 = v1 - v2; + n1 = Vector2(-n1.y, n1.x); + n1 = n1 * (1.0f / sqrtf(n1.x * n1.x + n1.y * n1.y)); + n2 = v2 - v3; + n2 = Vector2(-n2.y, n2.x); + n2 = n2 * (1.0f / sqrtf(n2.x * n2.x + n2.y * n2.y)); + n3 = v3 - v1; + n3 = Vector2(-n3.y, n3.x); + n3 = n3 * (1.0f / sqrtf(n3.x * n3.x + n3.y * n3.y)); + } + + // Vertices. + Vector2 v1, v2, v3; + Vector2 n1, n2, n3; // unit inward normals + Vector3 t1, t2, t3; + + // Deltas. + Vector3 dx, dy; + + float sign; + bool valid; +}; + +enum Mode { + Mode_Nearest, + Mode_Antialiased +}; + +// Process the given triangle. Returns false if rasterization was interrupted by the callback. +static bool drawTriangle(Mode mode, Vector2::Arg extents, bool enableScissors, const Vector2 v[3], SamplingCallback cb, void *param) { + Triangle tri(v[0], v[1], v[2], Vector3(1, 0, 0), Vector3(0, 1, 0), Vector3(0, 0, 1)); + // @@ It would be nice to have a conservative drawing mode that enlarges the triangle extents by one texel and is able to handle degenerate triangles. + // @@ Maybe the simplest thing to do would be raster triangle edges. + if (tri.valid) { + if (mode == Mode_Antialiased) { + return tri.drawAA(extents, enableScissors, cb, param); + } + if (mode == Mode_Nearest) { + return tri.draw(extents, enableScissors, cb, param); + } + } + return true; +} + +// Process the given quad. Returns false if rasterization was interrupted by the callback. +static bool drawQuad(Mode mode, Vector2::Arg extents, bool enableScissors, const Vector2 v[4], SamplingCallback cb, void *param) { + bool sign0 = triangleArea2(v[0], v[1], v[2]) > 0.0f; + bool sign1 = triangleArea2(v[0], v[2], v[3]) > 0.0f; + // Divide the quad into two non overlapping triangles. + if (sign0 == sign1) { + Triangle tri0(v[0], v[1], v[2], Vector3(0, 0, 0), Vector3(1, 0, 0), Vector3(1, 1, 0)); + Triangle tri1(v[0], v[2], v[3], Vector3(0, 0, 0), Vector3(1, 1, 0), Vector3(0, 1, 0)); + if (tri0.valid && tri1.valid) { + if (mode == Mode_Antialiased) { + return tri0.drawAA(extents, enableScissors, cb, param) && tri1.drawAA(extents, enableScissors, cb, param); + } else { + return tri0.draw(extents, enableScissors, cb, param) && tri1.draw(extents, enableScissors, cb, param); + } + } + } else { + Triangle tri0(v[0], v[1], v[3], Vector3(0, 0, 0), Vector3(1, 0, 0), Vector3(0, 1, 0)); + Triangle tri1(v[1], v[2], v[3], Vector3(1, 0, 0), Vector3(1, 1, 0), Vector3(0, 1, 0)); + if (tri0.valid && tri1.valid) { + if (mode == Mode_Antialiased) { + return tri0.drawAA(extents, enableScissors, cb, param) && tri1.drawAA(extents, enableScissors, cb, param); + } else { + return tri0.draw(extents, enableScissors, cb, param) && tri1.draw(extents, enableScissors, cb, param); + } + } + } + return true; +} +} // namespace raster + +// Full and sparse vector and matrix classes. BLAS subset. +// Pseudo-BLAS interface. +namespace sparse { +enum Transpose { + NoTransposed = 0, + Transposed = 1 +}; + +/** +* Sparse matrix class. The matrix is assumed to be sparse and to have +* very few non-zero elements, for this reason it's stored in indexed +* format. To multiply column vectors efficiently, the matrix stores +* the elements in indexed-column order, there is a list of indexed +* elements for each row of the matrix. As with the FullVector the +* dimension of the matrix is constant. +**/ +class Matrix { +public: + // An element of the sparse array. + struct Coefficient { + uint32_t x; // column + float v; // value + }; + + Matrix(uint32_t d) : + m_width(d) { m_array.resize(d); } + Matrix(uint32_t w, uint32_t h) : + m_width(w) { m_array.resize(h); } + Matrix(const Matrix &m) : + m_width(m.m_width) { m_array = m.m_array; } + + const Matrix &operator=(const Matrix &m) { + xaAssert(width() == m.width()); + xaAssert(height() == m.height()); + m_array = m.m_array; + return *this; + } + + uint32_t width() const { return m_width; } + uint32_t height() const { return m_array.size(); } + bool isSquare() const { return width() == height(); } + + // x is column, y is row + float getCoefficient(uint32_t x, uint32_t y) const { + xaDebugAssert(x < width()); + xaDebugAssert(y < height()); + const uint32_t count = m_array[y].size(); + for (uint32_t i = 0; i < count; i++) { + if (m_array[y][i].x == x) return m_array[y][i].v; + } + return 0.0f; + } + + void setCoefficient(uint32_t x, uint32_t y, float f) { + xaDebugAssert(x < width()); + xaDebugAssert(y < height()); + const uint32_t count = m_array[y].size(); + for (uint32_t i = 0; i < count; i++) { + if (m_array[y][i].x == x) { + m_array[y][i].v = f; + return; + } + } + if (f != 0.0f) { + Coefficient c = { x, f }; + m_array[y].push_back(c); + } + } + + float dotRow(uint32_t y, const FullVector &v) const { + xaDebugAssert(y < height()); + const uint32_t count = m_array[y].size(); + float sum = 0; + for (uint32_t i = 0; i < count; i++) { + sum += m_array[y][i].v * v[m_array[y][i].x]; + } + return sum; + } + + void madRow(uint32_t y, float alpha, FullVector &v) const { + xaDebugAssert(y < height()); + const uint32_t count = m_array[y].size(); + for (uint32_t i = 0; i < count; i++) { + v[m_array[y][i].x] += alpha * m_array[y][i].v; + } + } + + void clearRow(uint32_t y) { + xaDebugAssert(y < height()); + m_array[y].clear(); + } + + void scaleRow(uint32_t y, float f) { + xaDebugAssert(y < height()); + const uint32_t count = m_array[y].size(); + for (uint32_t i = 0; i < count; i++) { + m_array[y][i].v *= f; + } + } + + const std::vector<Coefficient> &getRow(uint32_t y) const { return m_array[y]; } + +private: + /// Number of columns. + const uint32_t m_width; + + /// Array of matrix elements. + std::vector<std::vector<Coefficient> > m_array; +}; + +// y = a * x + y +static void saxpy(float a, const FullVector &x, FullVector &y) { + xaDebugAssert(x.dimension() == y.dimension()); + const uint32_t dim = x.dimension(); + for (uint32_t i = 0; i < dim; i++) { + y[i] += a * x[i]; + } +} + +static void copy(const FullVector &x, FullVector &y) { + xaDebugAssert(x.dimension() == y.dimension()); + const uint32_t dim = x.dimension(); + for (uint32_t i = 0; i < dim; i++) { + y[i] = x[i]; + } +} + +static void scal(float a, FullVector &x) { + const uint32_t dim = x.dimension(); + for (uint32_t i = 0; i < dim; i++) { + x[i] *= a; + } +} + +static float dot(const FullVector &x, const FullVector &y) { + xaDebugAssert(x.dimension() == y.dimension()); + const uint32_t dim = x.dimension(); + float sum = 0; + for (uint32_t i = 0; i < dim; i++) { + sum += x[i] * y[i]; + } + return sum; +} + +static void mult(Transpose TM, const Matrix &M, const FullVector &x, FullVector &y) { + const uint32_t w = M.width(); + const uint32_t h = M.height(); + if (TM == Transposed) { + xaDebugAssert(h == x.dimension()); + xaDebugAssert(w == y.dimension()); + y.fill(0.0f); + for (uint32_t i = 0; i < h; i++) { + M.madRow(i, x[i], y); + } + } else { + xaDebugAssert(w == x.dimension()); + xaDebugAssert(h == y.dimension()); + for (uint32_t i = 0; i < h; i++) { + y[i] = M.dotRow(i, x); + } + } +} + +// y = M * x +static void mult(const Matrix &M, const FullVector &x, FullVector &y) { + mult(NoTransposed, M, x, y); +} + +static void sgemv(float alpha, Transpose TA, const Matrix &A, const FullVector &x, float beta, FullVector &y) { + const uint32_t w = A.width(); + const uint32_t h = A.height(); + if (TA == Transposed) { + xaDebugAssert(h == x.dimension()); + xaDebugAssert(w == y.dimension()); + for (uint32_t i = 0; i < h; i++) { + A.madRow(i, alpha * x[i], y); + } + } else { + xaDebugAssert(w == x.dimension()); + xaDebugAssert(h == y.dimension()); + for (uint32_t i = 0; i < h; i++) { + y[i] = alpha * A.dotRow(i, x) + beta * y[i]; + } + } +} + +// y = alpha*A*x + beta*y +static void sgemv(float alpha, const Matrix &A, const FullVector &x, float beta, FullVector &y) { + sgemv(alpha, NoTransposed, A, x, beta, y); +} + +// dot y-row of A by x-column of B +static float dotRowColumn(int y, const Matrix &A, int x, const Matrix &B) { + const std::vector<Matrix::Coefficient> &row = A.getRow(y); + const uint32_t count = row.size(); + float sum = 0.0f; + for (uint32_t i = 0; i < count; i++) { + const Matrix::Coefficient &c = row[i]; + sum += c.v * B.getCoefficient(x, c.x); + } + return sum; +} + +// dot y-row of A by x-row of B +static float dotRowRow(int y, const Matrix &A, int x, const Matrix &B) { + const std::vector<Matrix::Coefficient> &row = A.getRow(y); + const uint32_t count = row.size(); + float sum = 0.0f; + for (uint32_t i = 0; i < count; i++) { + const Matrix::Coefficient &c = row[i]; + sum += c.v * B.getCoefficient(c.x, x); + } + return sum; +} + +// dot y-column of A by x-column of B +static float dotColumnColumn(int y, const Matrix &A, int x, const Matrix &B) { + xaDebugAssert(A.height() == B.height()); + const uint32_t h = A.height(); + float sum = 0.0f; + for (uint32_t i = 0; i < h; i++) { + sum += A.getCoefficient(y, i) * B.getCoefficient(x, i); + } + return sum; +} + +static void transpose(const Matrix &A, Matrix &B) { + xaDebugAssert(A.width() == B.height()); + xaDebugAssert(B.width() == A.height()); + const uint32_t w = A.width(); + for (uint32_t x = 0; x < w; x++) { + B.clearRow(x); + } + const uint32_t h = A.height(); + for (uint32_t y = 0; y < h; y++) { + const std::vector<Matrix::Coefficient> &row = A.getRow(y); + const uint32_t count = row.size(); + for (uint32_t i = 0; i < count; i++) { + const Matrix::Coefficient &c = row[i]; + xaDebugAssert(c.x < w); + B.setCoefficient(y, c.x, c.v); + } + } +} + +static void sgemm(float alpha, Transpose TA, const Matrix &A, Transpose TB, const Matrix &B, float beta, Matrix &C) { + const uint32_t w = C.width(); + const uint32_t h = C.height(); + uint32_t aw = (TA == NoTransposed) ? A.width() : A.height(); + uint32_t ah = (TA == NoTransposed) ? A.height() : A.width(); + uint32_t bw = (TB == NoTransposed) ? B.width() : B.height(); + uint32_t bh = (TB == NoTransposed) ? B.height() : B.width(); + xaDebugAssert(aw == bh); + xaDebugAssert(bw == ah); + xaDebugAssert(w == bw); + xaDebugAssert(h == ah); +#ifdef NDEBUG + aw = ah = bw = bh = 0; // silence unused parameter warning +#endif + for (uint32_t y = 0; y < h; y++) { + for (uint32_t x = 0; x < w; x++) { + float c = beta * C.getCoefficient(x, y); + if (TA == NoTransposed && TB == NoTransposed) { + // dot y-row of A by x-column of B. + c += alpha * dotRowColumn(y, A, x, B); + } else if (TA == Transposed && TB == Transposed) { + // dot y-column of A by x-row of B. + c += alpha * dotRowColumn(x, B, y, A); + } else if (TA == Transposed && TB == NoTransposed) { + // dot y-column of A by x-column of B. + c += alpha * dotColumnColumn(y, A, x, B); + } else if (TA == NoTransposed && TB == Transposed) { + // dot y-row of A by x-row of B. + c += alpha * dotRowRow(y, A, x, B); + } + C.setCoefficient(x, y, c); + } + } +} + +static void mult(Transpose TA, const Matrix &A, Transpose TB, const Matrix &B, Matrix &C) { + sgemm(1.0f, TA, A, TB, B, 0.0f, C); +} + +// C = A * B +static void mult(const Matrix &A, const Matrix &B, Matrix &C) { + mult(NoTransposed, A, NoTransposed, B, C); +} + +} // namespace sparse + +class JacobiPreconditioner { +public: + JacobiPreconditioner(const sparse::Matrix &M, bool symmetric) : + m_inverseDiagonal(M.width()) { + xaAssert(M.isSquare()); + for (uint32_t x = 0; x < M.width(); x++) { + float elem = M.getCoefficient(x, x); + //xaDebugAssert( elem != 0.0f ); // This can be zero in the presence of zero area triangles. + if (symmetric) { + m_inverseDiagonal[x] = (elem != 0) ? 1.0f / sqrtf(fabsf(elem)) : 1.0f; + } else { + m_inverseDiagonal[x] = (elem != 0) ? 1.0f / elem : 1.0f; + } + } + } + + void apply(const FullVector &x, FullVector &y) const { + xaDebugAssert(x.dimension() == m_inverseDiagonal.dimension()); + xaDebugAssert(y.dimension() == m_inverseDiagonal.dimension()); + // @@ Wrap vector component-wise product into a separate function. + const uint32_t D = x.dimension(); + for (uint32_t i = 0; i < D; i++) { + y[i] = m_inverseDiagonal[i] * x[i]; + } + } + +private: + FullVector m_inverseDiagonal; +}; + +// Linear solvers. +class Solver { +public: + // Solve the symmetric system: At·A·x = At·b + static bool LeastSquaresSolver(const sparse::Matrix &A, const FullVector &b, FullVector &x, float epsilon = 1e-5f) { + xaDebugAssert(A.width() == x.dimension()); + xaDebugAssert(A.height() == b.dimension()); + xaDebugAssert(A.height() >= A.width()); // @@ If height == width we could solve it directly... + const uint32_t D = A.width(); + sparse::Matrix At(A.height(), A.width()); + sparse::transpose(A, At); + FullVector Atb(D); + sparse::mult(At, b, Atb); + sparse::Matrix AtA(D); + sparse::mult(At, A, AtA); + return SymmetricSolver(AtA, Atb, x, epsilon); + } + + // See section 10.4.3 in: Mesh Parameterization: Theory and Practice, Siggraph Course Notes, August 2007 + static bool LeastSquaresSolver(const sparse::Matrix &A, const FullVector &b, FullVector &x, const uint32_t *lockedParameters, uint32_t lockedCount, float epsilon = 1e-5f) { + xaDebugAssert(A.width() == x.dimension()); + xaDebugAssert(A.height() == b.dimension()); + xaDebugAssert(A.height() >= A.width() - lockedCount); + // @@ This is not the most efficient way of building a system with reduced degrees of freedom. It would be faster to do it on the fly. + const uint32_t D = A.width() - lockedCount; + xaDebugAssert(D > 0); + // Compute: b - Al * xl + FullVector b_Alxl(b); + for (uint32_t y = 0; y < A.height(); y++) { + const uint32_t count = A.getRow(y).size(); + for (uint32_t e = 0; e < count; e++) { + uint32_t column = A.getRow(y)[e].x; + bool isFree = true; + for (uint32_t i = 0; i < lockedCount; i++) { + isFree &= (lockedParameters[i] != column); + } + if (!isFree) { + b_Alxl[y] -= x[column] * A.getRow(y)[e].v; + } + } + } + // Remove locked columns from A. + sparse::Matrix Af(D, A.height()); + for (uint32_t y = 0; y < A.height(); y++) { + const uint32_t count = A.getRow(y).size(); + for (uint32_t e = 0; e < count; e++) { + uint32_t column = A.getRow(y)[e].x; + uint32_t ix = column; + bool isFree = true; + for (uint32_t i = 0; i < lockedCount; i++) { + isFree &= (lockedParameters[i] != column); + if (column > lockedParameters[i]) ix--; // shift columns + } + if (isFree) { + Af.setCoefficient(ix, y, A.getRow(y)[e].v); + } + } + } + // Remove elements from x + FullVector xf(D); + for (uint32_t i = 0, j = 0; i < A.width(); i++) { + bool isFree = true; + for (uint32_t l = 0; l < lockedCount; l++) { + isFree &= (lockedParameters[l] != i); + } + if (isFree) { + xf[j++] = x[i]; + } + } + // Solve reduced system. + bool result = LeastSquaresSolver(Af, b_Alxl, xf, epsilon); + // Copy results back to x. + for (uint32_t i = 0, j = 0; i < A.width(); i++) { + bool isFree = true; + for (uint32_t l = 0; l < lockedCount; l++) { + isFree &= (lockedParameters[l] != i); + } + if (isFree) { + x[i] = xf[j++]; + } + } + return result; + } + +private: + /** + * Compute the solution of the sparse linear system Ab=x using the Conjugate + * Gradient method. + * + * Solving sparse linear systems: + * (1) A·x = b + * + * The conjugate gradient algorithm solves (1) only in the case that A is + * symmetric and positive definite. It is based on the idea of minimizing the + * function + * + * (2) f(x) = 1/2·x·A·x - b·x + * + * This function is minimized when its gradient + * + * (3) df = A·x - b + * + * is zero, which is equivalent to (1). The minimization is carried out by + * generating a succession of search directions p.k and improved minimizers x.k. + * At each stage a quantity alfa.k is found that minimizes f(x.k + alfa.k·p.k), + * and x.k+1 is set equal to the new point x.k + alfa.k·p.k. The p.k and x.k are + * built up in such a way that x.k+1 is also the minimizer of f over the whole + * vector space of directions already taken, {p.1, p.2, . . . , p.k}. After N + * iterations you arrive at the minimizer over the entire vector space, i.e., the + * solution to (1). + * + * For a really good explanation of the method see: + * + * "An Introduction to the Conjugate Gradient Method Without the Agonizing Pain", + * Jonhathan Richard Shewchuk. + * + **/ + static bool ConjugateGradientSolver(const sparse::Matrix &A, const FullVector &b, FullVector &x, float epsilon) { + xaDebugAssert(A.isSquare()); + xaDebugAssert(A.width() == b.dimension()); + xaDebugAssert(A.width() == x.dimension()); + int i = 0; + const int D = A.width(); + const int i_max = 4 * D; // Convergence should be linear, but in some cases, it's not. + FullVector r(D); // residual + FullVector p(D); // search direction + FullVector q(D); // + float delta_0; + float delta_old; + float delta_new; + float alpha; + float beta; + // r = b - A·x; + sparse::copy(b, r); + sparse::sgemv(-1, A, x, 1, r); + // p = r; + sparse::copy(r, p); + delta_new = sparse::dot(r, r); + delta_0 = delta_new; + while (i < i_max && delta_new > epsilon * epsilon * delta_0) { + i++; + // q = A·p + mult(A, p, q); + // alpha = delta_new / p·q + alpha = delta_new / sparse::dot(p, q); + // x = alfa·p + x + sparse::saxpy(alpha, p, x); + if ((i & 31) == 0) { // recompute r after 32 steps + // r = b - A·x + sparse::copy(b, r); + sparse::sgemv(-1, A, x, 1, r); + } else { + // r = r - alpha·q + sparse::saxpy(-alpha, q, r); + } + delta_old = delta_new; + delta_new = sparse::dot(r, r); + beta = delta_new / delta_old; + // p = beta·p + r + sparse::scal(beta, p); + sparse::saxpy(1, r, p); + } + return delta_new <= epsilon * epsilon * delta_0; + } + + // Conjugate gradient with preconditioner. + static bool ConjugateGradientSolver(const JacobiPreconditioner &preconditioner, const sparse::Matrix &A, const FullVector &b, FullVector &x, float epsilon) { + xaDebugAssert(A.isSquare()); + xaDebugAssert(A.width() == b.dimension()); + xaDebugAssert(A.width() == x.dimension()); + int i = 0; + const int D = A.width(); + const int i_max = 4 * D; // Convergence should be linear, but in some cases, it's not. + FullVector r(D); // residual + FullVector p(D); // search direction + FullVector q(D); // + FullVector s(D); // preconditioned + float delta_0; + float delta_old; + float delta_new; + float alpha; + float beta; + // r = b - A·x + sparse::copy(b, r); + sparse::sgemv(-1, A, x, 1, r); + // p = M^-1 · r + preconditioner.apply(r, p); + delta_new = sparse::dot(r, p); + delta_0 = delta_new; + while (i < i_max && delta_new > epsilon * epsilon * delta_0) { + i++; + // q = A·p + mult(A, p, q); + // alpha = delta_new / p·q + alpha = delta_new / sparse::dot(p, q); + // x = alfa·p + x + sparse::saxpy(alpha, p, x); + if ((i & 31) == 0) { // recompute r after 32 steps + // r = b - A·x + sparse::copy(b, r); + sparse::sgemv(-1, A, x, 1, r); + } else { + // r = r - alfa·q + sparse::saxpy(-alpha, q, r); + } + // s = M^-1 · r + preconditioner.apply(r, s); + delta_old = delta_new; + delta_new = sparse::dot(r, s); + beta = delta_new / delta_old; + // p = s + beta·p + sparse::scal(beta, p); + sparse::saxpy(1, s, p); + } + return delta_new <= epsilon * epsilon * delta_0; + } + + static bool SymmetricSolver(const sparse::Matrix &A, const FullVector &b, FullVector &x, float epsilon = 1e-5f) { + xaDebugAssert(A.height() == A.width()); + xaDebugAssert(A.height() == b.dimension()); + xaDebugAssert(b.dimension() == x.dimension()); + JacobiPreconditioner jacobi(A, true); + return ConjugateGradientSolver(jacobi, A, b, x, epsilon); + } +}; + +namespace param { +class Atlas; +class Chart; + +// Fast sweep in 3 directions +static bool findApproximateDiameterVertices(halfedge::Mesh *mesh, halfedge::Vertex **a, halfedge::Vertex **b) { + xaDebugAssert(mesh != NULL); + xaDebugAssert(a != NULL); + xaDebugAssert(b != NULL); + const uint32_t vertexCount = mesh->vertexCount(); + halfedge::Vertex *minVertex[3]; + halfedge::Vertex *maxVertex[3]; + minVertex[0] = minVertex[1] = minVertex[2] = NULL; + maxVertex[0] = maxVertex[1] = maxVertex[2] = NULL; + for (uint32_t v = 1; v < vertexCount; v++) { + halfedge::Vertex *vertex = mesh->vertexAt(v); + xaDebugAssert(vertex != NULL); + if (vertex->isBoundary()) { + minVertex[0] = minVertex[1] = minVertex[2] = vertex; + maxVertex[0] = maxVertex[1] = maxVertex[2] = vertex; + break; + } + } + if (minVertex[0] == NULL) { + // Input mesh has not boundaries. + return false; + } + for (uint32_t v = 1; v < vertexCount; v++) { + halfedge::Vertex *vertex = mesh->vertexAt(v); + xaDebugAssert(vertex != NULL); + if (!vertex->isBoundary()) { + // Skip interior vertices. + continue; + } + if (vertex->pos.x < minVertex[0]->pos.x) + minVertex[0] = vertex; + else if (vertex->pos.x > maxVertex[0]->pos.x) + maxVertex[0] = vertex; + if (vertex->pos.y < minVertex[1]->pos.y) + minVertex[1] = vertex; + else if (vertex->pos.y > maxVertex[1]->pos.y) + maxVertex[1] = vertex; + if (vertex->pos.z < minVertex[2]->pos.z) + minVertex[2] = vertex; + else if (vertex->pos.z > maxVertex[2]->pos.z) + maxVertex[2] = vertex; + } + float lengths[3]; + for (int i = 0; i < 3; i++) { + lengths[i] = length(minVertex[i]->pos - maxVertex[i]->pos); + } + if (lengths[0] > lengths[1] && lengths[0] > lengths[2]) { + *a = minVertex[0]; + *b = maxVertex[0]; + } else if (lengths[1] > lengths[2]) { + *a = minVertex[1]; + *b = maxVertex[1]; + } else { + *a = minVertex[2]; + *b = maxVertex[2]; + } + return true; +} + +// Conformal relations from Brecht Van Lommel (based on ABF): + +static float vec_angle_cos(Vector3::Arg v1, Vector3::Arg v2, Vector3::Arg v3) { + Vector3 d1 = v1 - v2; + Vector3 d2 = v3 - v2; + return clamp(dot(d1, d2) / (length(d1) * length(d2)), -1.0f, 1.0f); +} + +static float vec_angle(Vector3::Arg v1, Vector3::Arg v2, Vector3::Arg v3) { + float dot = vec_angle_cos(v1, v2, v3); + return acosf(dot); +} + +static void triangle_angles(Vector3::Arg v1, Vector3::Arg v2, Vector3::Arg v3, float *a1, float *a2, float *a3) { + *a1 = vec_angle(v3, v1, v2); + *a2 = vec_angle(v1, v2, v3); + *a3 = PI - *a2 - *a1; +} + +static void setup_abf_relations(sparse::Matrix &A, int row, const halfedge::Vertex *v0, const halfedge::Vertex *v1, const halfedge::Vertex *v2) { + int id0 = v0->id; + int id1 = v1->id; + int id2 = v2->id; + Vector3 p0 = v0->pos; + Vector3 p1 = v1->pos; + Vector3 p2 = v2->pos; + // @@ IC: Wouldn't it be more accurate to return cos and compute 1-cos^2? + // It does indeed seem to be a little bit more robust. + // @@ Need to revisit this more carefully! + float a0, a1, a2; + triangle_angles(p0, p1, p2, &a0, &a1, &a2); + float s0 = sinf(a0); + float s1 = sinf(a1); + float s2 = sinf(a2); + if (s1 > s0 && s1 > s2) { + std::swap(s1, s2); + std::swap(s0, s1); + std::swap(a1, a2); + std::swap(a0, a1); + std::swap(id1, id2); + std::swap(id0, id1); + } else if (s0 > s1 && s0 > s2) { + std::swap(s0, s2); + std::swap(s0, s1); + std::swap(a0, a2); + std::swap(a0, a1); + std::swap(id0, id2); + std::swap(id0, id1); + } + float c0 = cosf(a0); + float ratio = (s2 == 0.0f) ? 1.0f : s1 / s2; + float cosine = c0 * ratio; + float sine = s0 * ratio; + // Note : 2*id + 0 --> u + // 2*id + 1 --> v + int u0_id = 2 * id0 + 0; + int v0_id = 2 * id0 + 1; + int u1_id = 2 * id1 + 0; + int v1_id = 2 * id1 + 1; + int u2_id = 2 * id2 + 0; + int v2_id = 2 * id2 + 1; + // Real part + A.setCoefficient(u0_id, 2 * row + 0, cosine - 1.0f); + A.setCoefficient(v0_id, 2 * row + 0, -sine); + A.setCoefficient(u1_id, 2 * row + 0, -cosine); + A.setCoefficient(v1_id, 2 * row + 0, sine); + A.setCoefficient(u2_id, 2 * row + 0, 1); + // Imaginary part + A.setCoefficient(u0_id, 2 * row + 1, sine); + A.setCoefficient(v0_id, 2 * row + 1, cosine - 1.0f); + A.setCoefficient(u1_id, 2 * row + 1, -sine); + A.setCoefficient(v1_id, 2 * row + 1, -cosine); + A.setCoefficient(v2_id, 2 * row + 1, 1); +} + +bool computeLeastSquaresConformalMap(halfedge::Mesh *mesh) { + xaDebugAssert(mesh != NULL); + // For this to work properly, mesh should not have colocals that have the same + // attributes, unless you want the vertices to actually have different texcoords. + const uint32_t vertexCount = mesh->vertexCount(); + const uint32_t D = 2 * vertexCount; + const uint32_t N = 2 * halfedge::countMeshTriangles(mesh); + // N is the number of equations (one per triangle) + // D is the number of variables (one per vertex; there are 2 pinned vertices). + if (N < D - 4) { + return false; + } + sparse::Matrix A(D, N); + FullVector b(N); + FullVector x(D); + // Fill b: + b.fill(0.0f); + // Fill x: + halfedge::Vertex *v0; + halfedge::Vertex *v1; + if (!findApproximateDiameterVertices(mesh, &v0, &v1)) { + // Mesh has no boundaries. + return false; + } + if (v0->tex == v1->tex) { + // LSCM expects an existing parameterization. + return false; + } + for (uint32_t v = 0; v < vertexCount; v++) { + halfedge::Vertex *vertex = mesh->vertexAt(v); + xaDebugAssert(vertex != NULL); + // Initial solution. + x[2 * v + 0] = vertex->tex.x; + x[2 * v + 1] = vertex->tex.y; + } + // Fill A: + const uint32_t faceCount = mesh->faceCount(); + for (uint32_t f = 0, t = 0; f < faceCount; f++) { + const halfedge::Face *face = mesh->faceAt(f); + xaDebugAssert(face != NULL); + xaDebugAssert(face->edgeCount() == 3); + const halfedge::Vertex *vertex0 = NULL; + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + xaAssert(edge != NULL); + if (vertex0 == NULL) { + vertex0 = edge->vertex; + } else if (edge->next->vertex != vertex0) { + const halfedge::Vertex *vertex1 = edge->from(); + const halfedge::Vertex *vertex2 = edge->to(); + setup_abf_relations(A, t, vertex0, vertex1, vertex2); + //setup_conformal_map_relations(A, t, vertex0, vertex1, vertex2); + t++; + } + } + } + const uint32_t lockedParameters[] = { + 2 * v0->id + 0, + 2 * v0->id + 1, + 2 * v1->id + 0, + 2 * v1->id + 1 + }; + // Solve + Solver::LeastSquaresSolver(A, b, x, lockedParameters, 4, 0.000001f); + // Map x back to texcoords: + for (uint32_t v = 0; v < vertexCount; v++) { + halfedge::Vertex *vertex = mesh->vertexAt(v); + xaDebugAssert(vertex != NULL); + vertex->tex = Vector2(x[2 * v + 0], x[2 * v + 1]); + } + return true; +} + +bool computeOrthogonalProjectionMap(halfedge::Mesh *mesh) { + Vector3 axis[2]; + uint32_t vertexCount = mesh->vertexCount(); + std::vector<Vector3> points(vertexCount); + points.resize(vertexCount); + for (uint32_t i = 0; i < vertexCount; i++) { + points[i] = mesh->vertexAt(i)->pos; + } + // Avoid redundant computations. + float matrix[6]; + Fit::computeCovariance(vertexCount, points.data(), matrix); + if (matrix[0] == 0 && matrix[3] == 0 && matrix[5] == 0) { + return false; + } + float eigenValues[3]; + Vector3 eigenVectors[3]; + if (!Fit::eigenSolveSymmetric3(matrix, eigenValues, eigenVectors)) { + return false; + } + axis[0] = normalize(eigenVectors[0]); + axis[1] = normalize(eigenVectors[1]); + // Project vertices to plane. + for (halfedge::Mesh::VertexIterator it(mesh->vertices()); !it.isDone(); it.advance()) { + halfedge::Vertex *vertex = it.current(); + vertex->tex.x = dot(axis[0], vertex->pos); + vertex->tex.y = dot(axis[1], vertex->pos); + } + return true; +} + +void computeSingleFaceMap(halfedge::Mesh *mesh) { + xaDebugAssert(mesh != NULL); + xaDebugAssert(mesh->faceCount() == 1); + halfedge::Face *face = mesh->faceAt(0); + xaAssert(face != NULL); + Vector3 p0 = face->edge->from()->pos; + Vector3 p1 = face->edge->to()->pos; + Vector3 X = normalizeSafe(p1 - p0, Vector3(0.0f), 0.0f); + Vector3 Z = face->normal(); + Vector3 Y = normalizeSafe(cross(Z, X), Vector3(0.0f), 0.0f); + uint32_t i = 0; + for (halfedge::Face::EdgeIterator it(face->edges()); !it.isDone(); it.advance(), i++) { + halfedge::Vertex *vertex = it.vertex(); + xaAssert(vertex != NULL); + if (i == 0) { + vertex->tex = Vector2(0); + } else { + Vector3 pn = vertex->pos; + float xn = dot((pn - p0), X); + float yn = dot((pn - p0), Y); + vertex->tex = Vector2(xn, yn); + } + } +} + +// Dummy implementation of a priority queue using sort at insertion. +// - Insertion is o(n) +// - Smallest element goes at the end, so that popping it is o(1). +// - Resorting is n*log(n) +// @@ Number of elements in the queue is usually small, and we'd have to rebalance often. I'm not sure it's worth implementing a heap. +// @@ Searcing at removal would remove the need for sorting when priorities change. +struct PriorityQueue { + PriorityQueue(uint32_t size = UINT_MAX) : + maxSize(size) {} + + void push(float priority, uint32_t face) { + uint32_t i = 0; + const uint32_t count = pairs.size(); + for (; i < count; i++) { + if (pairs[i].priority > priority) break; + } + Pair p = { priority, face }; + pairs.insert(pairs.begin() + i, p); + if (pairs.size() > maxSize) { + pairs.erase(pairs.begin()); + } + } + + // push face out of order, to be sorted later. + void push(uint32_t face) { + Pair p = { 0.0f, face }; + pairs.push_back(p); + } + + uint32_t pop() { + uint32_t f = pairs.back().face; + pairs.pop_back(); + return f; + } + + void sort() { + //sort(pairs); // @@ My intro sort appears to be much slower than it should! + std::sort(pairs.begin(), pairs.end()); + } + + void clear() { + pairs.clear(); + } + + uint32_t count() const { + return pairs.size(); + } + + float firstPriority() const { + return pairs.back().priority; + } + + const uint32_t maxSize; + + struct Pair { + bool operator<(const Pair &p) const { + return priority > p.priority; // !! Sort in inverse priority order! + } + + float priority; + uint32_t face; + }; + + std::vector<Pair> pairs; +}; + +struct ChartBuildData { + ChartBuildData(int p_id) : + id(p_id) { + planeNormal = Vector3(0); + centroid = Vector3(0); + coneAxis = Vector3(0); + coneAngle = 0; + area = 0; + boundaryLength = 0; + normalSum = Vector3(0); + centroidSum = Vector3(0); + } + + int id; + + // Proxy info: + Vector3 planeNormal; + Vector3 centroid; + Vector3 coneAxis; + float coneAngle; + + float area; + float boundaryLength; + Vector3 normalSum; + Vector3 centroidSum; + + std::vector<uint32_t> seeds; // @@ These could be a pointers to the halfedge faces directly. + std::vector<uint32_t> faces; + PriorityQueue candidates; +}; + +struct AtlasBuilder { + AtlasBuilder(const halfedge::Mesh *m) : + mesh(m), + facesLeft(m->faceCount()) { + const uint32_t faceCount = m->faceCount(); + faceChartArray.resize(faceCount, -1); + faceCandidateArray.resize(faceCount, (uint32_t)-1); + // @@ Floyd for the whole mesh is too slow. We could compute floyd progressively per patch as the patch grows. We need a better solution to compute most central faces. + //computeShortestPaths(); + // Precompute edge lengths and face areas. + uint32_t edgeCount = m->edgeCount(); + edgeLengths.resize(edgeCount); + for (uint32_t i = 0; i < edgeCount; i++) { + uint32_t id = m->edgeAt(i)->id; + xaDebugAssert(id / 2 == i); +#ifdef NDEBUG + id = 0; // silence unused parameter warning +#endif + edgeLengths[i] = m->edgeAt(i)->length(); + } + faceAreas.resize(faceCount); + for (uint32_t i = 0; i < faceCount; i++) { + faceAreas[i] = m->faceAt(i)->area(); + } + } + + ~AtlasBuilder() { + const uint32_t chartCount = chartArray.size(); + for (uint32_t i = 0; i < chartCount; i++) { + delete chartArray[i]; + } + } + + void markUnchartedFaces(const std::vector<uint32_t> &unchartedFaces) { + const uint32_t unchartedFaceCount = unchartedFaces.size(); + for (uint32_t i = 0; i < unchartedFaceCount; i++) { + uint32_t f = unchartedFaces[i]; + faceChartArray[f] = -2; + //faceCandidateArray[f] = -2; // @@ ? + removeCandidate(f); + } + xaDebugAssert(facesLeft >= unchartedFaceCount); + facesLeft -= unchartedFaceCount; + } + + void computeShortestPaths() { + const uint32_t faceCount = mesh->faceCount(); + shortestPaths.resize(faceCount * faceCount, FLT_MAX); + // Fill edges: + for (uint32_t i = 0; i < faceCount; i++) { + shortestPaths[i * faceCount + i] = 0.0f; + const halfedge::Face *face_i = mesh->faceAt(i); + Vector3 centroid_i = face_i->centroid(); + for (halfedge::Face::ConstEdgeIterator it(face_i->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + if (!edge->isBoundary()) { + const halfedge::Face *face_j = edge->pair->face; + uint32_t j = face_j->id; + Vector3 centroid_j = face_j->centroid(); + shortestPaths[i * faceCount + j] = shortestPaths[j * faceCount + i] = length(centroid_i - centroid_j); + } + } + } + // Use Floyd-Warshall algorithm to compute all paths: + for (uint32_t k = 0; k < faceCount; k++) { + for (uint32_t i = 0; i < faceCount; i++) { + for (uint32_t j = 0; j < faceCount; j++) { + shortestPaths[i * faceCount + j] = std::min(shortestPaths[i * faceCount + j], shortestPaths[i * faceCount + k] + shortestPaths[k * faceCount + j]); + } + } + } + } + + void placeSeeds(float threshold, uint32_t maxSeedCount) { + // Instead of using a predefiened number of seeds: + // - Add seeds one by one, growing chart until a certain treshold. + // - Undo charts and restart growing process. + // @@ How can we give preference to faces far from sharp features as in the LSCM paper? + // - those points can be found using a simple flood filling algorithm. + // - how do we weight the probabilities? + for (uint32_t i = 0; i < maxSeedCount; i++) { + if (facesLeft == 0) { + // No faces left, stop creating seeds. + break; + } + createRandomChart(threshold); + } + } + + void createRandomChart(float threshold) { + ChartBuildData *chart = new ChartBuildData(chartArray.size()); + chartArray.push_back(chart); + // Pick random face that is not used by any chart yet. + uint32_t randomFaceIdx = rand.getRange(facesLeft - 1); + uint32_t i = 0; + for (uint32_t f = 0; f != randomFaceIdx; f++, i++) { + while (faceChartArray[i] != -1) + i++; + } + while (faceChartArray[i] != -1) + i++; + chart->seeds.push_back(i); + addFaceToChart(chart, i, true); + // Grow the chart as much as possible within the given threshold. + growChart(chart, threshold * 0.5f, facesLeft); + //growCharts(threshold - threshold * 0.75f / chartCount(), facesLeft); + } + + void addFaceToChart(ChartBuildData *chart, uint32_t f, bool recomputeProxy = false) { + // Add face to chart. + chart->faces.push_back(f); + xaDebugAssert(faceChartArray[f] == -1); + faceChartArray[f] = chart->id; + facesLeft--; + // Update area and boundary length. + chart->area = evaluateChartArea(chart, f); + chart->boundaryLength = evaluateBoundaryLength(chart, f); + chart->normalSum = evaluateChartNormalSum(chart, f); + chart->centroidSum = evaluateChartCentroidSum(chart, f); + if (recomputeProxy) { + // Update proxy and candidate's priorities. + updateProxy(chart); + } + // Update candidates. + removeCandidate(f); + updateCandidates(chart, f); + updatePriorities(chart); + } + + // Returns true if any of the charts can grow more. + bool growCharts(float threshold, uint32_t faceCount) { + // Using one global list. + faceCount = std::min(faceCount, facesLeft); + for (uint32_t i = 0; i < faceCount; i++) { + const Candidate &candidate = getBestCandidate(); + if (candidate.metric > threshold) { + return false; // Can't grow more. + } + addFaceToChart(candidate.chart, candidate.face); + } + return facesLeft != 0; // Can continue growing. + } + + bool growChart(ChartBuildData *chart, float threshold, uint32_t faceCount) { + // Try to add faceCount faces within threshold to chart. + for (uint32_t i = 0; i < faceCount;) { + if (chart->candidates.count() == 0 || chart->candidates.firstPriority() > threshold) { + return false; + } + uint32_t f = chart->candidates.pop(); + if (faceChartArray[f] == -1) { + addFaceToChart(chart, f); + i++; + } + } + if (chart->candidates.count() == 0 || chart->candidates.firstPriority() > threshold) { + return false; + } + return true; + } + + void resetCharts() { + const uint32_t faceCount = mesh->faceCount(); + for (uint32_t i = 0; i < faceCount; i++) { + faceChartArray[i] = -1; + faceCandidateArray[i] = (uint32_t)-1; + } + facesLeft = faceCount; + candidateArray.clear(); + const uint32_t chartCount = chartArray.size(); + for (uint32_t i = 0; i < chartCount; i++) { + ChartBuildData *chart = chartArray[i]; + const uint32_t seed = chart->seeds.back(); + chart->area = 0.0f; + chart->boundaryLength = 0.0f; + chart->normalSum = Vector3(0); + chart->centroidSum = Vector3(0); + chart->faces.clear(); + chart->candidates.clear(); + addFaceToChart(chart, seed); + } + } + + void updateCandidates(ChartBuildData *chart, uint32_t f) { + const halfedge::Face *face = mesh->faceAt(f); + // Traverse neighboring faces, add the ones that do not belong to any chart yet. + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current()->pair; + if (!edge->isBoundary()) { + uint32_t faceId = edge->face->id; + if (faceChartArray[faceId] == -1) { + chart->candidates.push(faceId); + } + } + } + } + + void updateProxies() { + const uint32_t chartCount = chartArray.size(); + for (uint32_t i = 0; i < chartCount; i++) { + updateProxy(chartArray[i]); + } + } + + void updateProxy(ChartBuildData *chart) { + //#pragma message(NV_FILE_LINE "TODO: Use best fit plane instead of average normal.") + chart->planeNormal = normalizeSafe(chart->normalSum, Vector3(0), 0.0f); + chart->centroid = chart->centroidSum / float(chart->faces.size()); + } + + bool relocateSeeds() { + bool anySeedChanged = false; + const uint32_t chartCount = chartArray.size(); + for (uint32_t i = 0; i < chartCount; i++) { + if (relocateSeed(chartArray[i])) { + anySeedChanged = true; + } + } + return anySeedChanged; + } + + bool relocateSeed(ChartBuildData *chart) { + Vector3 centroid = computeChartCentroid(chart); + const uint32_t N = 10; // @@ Hardcoded to 10? + PriorityQueue bestTriangles(N); + // Find the first N triangles that fit the proxy best. + const uint32_t faceCount = chart->faces.size(); + for (uint32_t i = 0; i < faceCount; i++) { + float priority = evaluateProxyFitMetric(chart, chart->faces[i]); + bestTriangles.push(priority, chart->faces[i]); + } + // Of those, choose the most central triangle. + uint32_t mostCentral; + float maxDistance = -1; + const uint32_t bestCount = bestTriangles.count(); + for (uint32_t i = 0; i < bestCount; i++) { + const halfedge::Face *face = mesh->faceAt(bestTriangles.pairs[i].face); + Vector3 faceCentroid = face->triangleCenter(); + float distance = length(centroid - faceCentroid); + if (distance > maxDistance) { + maxDistance = distance; + mostCentral = bestTriangles.pairs[i].face; + } + } + xaDebugAssert(maxDistance >= 0); + // In order to prevent k-means cyles we record all the previously chosen seeds. + uint32_t index = std::find(chart->seeds.begin(), chart->seeds.end(), mostCentral) - chart->seeds.begin(); + if (index < chart->seeds.size()) { + // Move new seed to the end of the seed array. + uint32_t last = chart->seeds.size() - 1; + std::swap(chart->seeds[index], chart->seeds[last]); + return false; + } else { + // Append new seed. + chart->seeds.push_back(mostCentral); + return true; + } + } + + void updatePriorities(ChartBuildData *chart) { + // Re-evaluate candidate priorities. + uint32_t candidateCount = chart->candidates.count(); + for (uint32_t i = 0; i < candidateCount; i++) { + chart->candidates.pairs[i].priority = evaluatePriority(chart, chart->candidates.pairs[i].face); + if (faceChartArray[chart->candidates.pairs[i].face] == -1) { + updateCandidate(chart, chart->candidates.pairs[i].face, chart->candidates.pairs[i].priority); + } + } + // Sort candidates. + chart->candidates.sort(); + } + + // Evaluate combined metric. + float evaluatePriority(ChartBuildData *chart, uint32_t face) { + // Estimate boundary length and area: + float newBoundaryLength = evaluateBoundaryLength(chart, face); + float newChartArea = evaluateChartArea(chart, face); + float F = evaluateProxyFitMetric(chart, face); + float C = evaluateRoundnessMetric(chart, face, newBoundaryLength, newChartArea); + float P = evaluateStraightnessMetric(chart, face); + // Penalize faces that cross seams, reward faces that close seams or reach boundaries. + float N = evaluateNormalSeamMetric(chart, face); + float T = evaluateTextureSeamMetric(chart, face); + //float R = evaluateCompletenessMetric(chart, face); + //float D = evaluateDihedralAngleMetric(chart, face); + // @@ Add a metric based on local dihedral angle. + // @@ Tweaking the normal and texture seam metrics. + // - Cause more impedance. Never cross 90 degree edges. + // - + float cost = float( + options.proxyFitMetricWeight * F + + options.roundnessMetricWeight * C + + options.straightnessMetricWeight * P + + options.normalSeamMetricWeight * N + + options.textureSeamMetricWeight * T); + // Enforce limits strictly: + if (newChartArea > options.maxChartArea) cost = FLT_MAX; + if (newBoundaryLength > options.maxBoundaryLength) cost = FLT_MAX; + // Make sure normal seams are fully respected: + if (options.normalSeamMetricWeight >= 1000 && N != 0) cost = FLT_MAX; + xaAssert(std::isfinite(cost)); + return cost; + } + + // Returns a value in [0-1]. + float evaluateProxyFitMetric(ChartBuildData *chart, uint32_t f) { + const halfedge::Face *face = mesh->faceAt(f); + Vector3 faceNormal = face->triangleNormal(); + // Use plane fitting metric for now: + return 1 - dot(faceNormal, chart->planeNormal); // @@ normal deviations should be weighted by face area + } + + float evaluateRoundnessMetric(ChartBuildData *chart, uint32_t /*face*/, float newBoundaryLength, float newChartArea) { + float roundness = square(chart->boundaryLength) / chart->area; + float newRoundness = square(newBoundaryLength) / newChartArea; + if (newRoundness > roundness) { + return square(newBoundaryLength) / (newChartArea * 4 * PI); + } else { + // Offer no impedance to faces that improve roundness. + return 0; + } + } + + float evaluateStraightnessMetric(ChartBuildData *chart, uint32_t f) { + float l_out = 0.0f; + float l_in = 0.0f; + const halfedge::Face *face = mesh->faceAt(f); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + float l = edgeLengths[edge->id / 2]; + if (edge->isBoundary()) { + l_out += l; + } else { + uint32_t neighborFaceId = edge->pair->face->id; + if (faceChartArray[neighborFaceId] != chart->id) { + l_out += l; + } else { + l_in += l; + } + } + } + xaDebugAssert(l_in != 0.0f); // Candidate face must be adjacent to chart. @@ This is not true if the input mesh has zero-length edges. + float ratio = (l_out - l_in) / (l_out + l_in); + return std::min(ratio, 0.0f); // Only use the straightness metric to close gaps. + } + + float evaluateNormalSeamMetric(ChartBuildData *chart, uint32_t f) { + float seamFactor = 0.0f; + float totalLength = 0.0f; + const halfedge::Face *face = mesh->faceAt(f); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + if (edge->isBoundary()) { + continue; + } + const uint32_t neighborFaceId = edge->pair->face->id; + if (faceChartArray[neighborFaceId] != chart->id) { + continue; + } + //float l = edge->length(); + float l = edgeLengths[edge->id / 2]; + totalLength += l; + if (!edge->isSeam()) { + continue; + } + // Make sure it's a normal seam. + if (edge->isNormalSeam()) { + float d0 = clamp(dot(edge->vertex->nor, edge->pair->next->vertex->nor), 0.0f, 1.0f); + float d1 = clamp(dot(edge->next->vertex->nor, edge->pair->vertex->nor), 0.0f, 1.0f); + l *= 1 - (d0 + d1) * 0.5f; + seamFactor += l; + } + } + if (seamFactor == 0) return 0.0f; + return seamFactor / totalLength; + } + + float evaluateTextureSeamMetric(ChartBuildData *chart, uint32_t f) { + float seamLength = 0.0f; + float totalLength = 0.0f; + const halfedge::Face *face = mesh->faceAt(f); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + if (edge->isBoundary()) { + continue; + } + const uint32_t neighborFaceId = edge->pair->face->id; + if (faceChartArray[neighborFaceId] != chart->id) { + continue; + } + //float l = edge->length(); + float l = edgeLengths[edge->id / 2]; + totalLength += l; + if (!edge->isSeam()) { + continue; + } + // Make sure it's a texture seam. + if (edge->isTextureSeam()) { + seamLength += l; + } + } + if (seamLength == 0.0f) { + return 0.0f; // Avoid division by zero. + } + return seamLength / totalLength; + } + + float evaluateChartArea(ChartBuildData *chart, uint32_t f) { + const halfedge::Face *face = mesh->faceAt(f); + return chart->area + faceAreas[face->id]; + } + + float evaluateBoundaryLength(ChartBuildData *chart, uint32_t f) { + float boundaryLength = chart->boundaryLength; + // Add new edges, subtract edges shared with the chart. + const halfedge::Face *face = mesh->faceAt(f); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + //float edgeLength = edge->length(); + float edgeLength = edgeLengths[edge->id / 2]; + if (edge->isBoundary()) { + boundaryLength += edgeLength; + } else { + uint32_t neighborFaceId = edge->pair->face->id; + if (faceChartArray[neighborFaceId] != chart->id) { + boundaryLength += edgeLength; + } else { + boundaryLength -= edgeLength; + } + } + } + return std::max(0.0f, boundaryLength); // @@ Hack! + } + + Vector3 evaluateChartNormalSum(ChartBuildData *chart, uint32_t f) { + const halfedge::Face *face = mesh->faceAt(f); + return chart->normalSum + face->triangleNormalAreaScaled(); + } + + Vector3 evaluateChartCentroidSum(ChartBuildData *chart, uint32_t f) { + const halfedge::Face *face = mesh->faceAt(f); + return chart->centroidSum + face->centroid(); + } + + Vector3 computeChartCentroid(const ChartBuildData *chart) { + Vector3 centroid(0); + const uint32_t faceCount = chart->faces.size(); + for (uint32_t i = 0; i < faceCount; i++) { + const halfedge::Face *face = mesh->faceAt(chart->faces[i]); + centroid += face->triangleCenter(); + } + return centroid / float(faceCount); + } + + void fillHoles(float threshold) { + while (facesLeft > 0) + createRandomChart(threshold); + } + + void mergeCharts() { + std::vector<float> sharedBoundaryLengths; + const uint32_t chartCount = chartArray.size(); + for (int c = chartCount - 1; c >= 0; c--) { + sharedBoundaryLengths.clear(); + sharedBoundaryLengths.resize(chartCount, 0.0f); + ChartBuildData *chart = chartArray[c]; + float externalBoundary = 0.0f; + const uint32_t faceCount = chart->faces.size(); + for (uint32_t i = 0; i < faceCount; i++) { + uint32_t f = chart->faces[i]; + const halfedge::Face *face = mesh->faceAt(f); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + //float l = edge->length(); + float l = edgeLengths[edge->id / 2]; + if (edge->isBoundary()) { + externalBoundary += l; + } else { + uint32_t neighborFace = edge->pair->face->id; + uint32_t neighborChart = faceChartArray[neighborFace]; + if (neighborChart != (uint32_t)c) { + if ((edge->isSeam() && (edge->isNormalSeam() || edge->isTextureSeam())) || neighborChart == -2) { + externalBoundary += l; + } else { + sharedBoundaryLengths[neighborChart] += l; + } + } + } + } + } + for (int cc = chartCount - 1; cc >= 0; cc--) { + if (cc == c) + continue; + ChartBuildData *chart2 = chartArray[cc]; + if (chart2 == NULL) + continue; + if (sharedBoundaryLengths[cc] > 0.8 * std::max(0.0f, chart->boundaryLength - externalBoundary)) { + // Try to avoid degenerate configurations. + if (chart2->boundaryLength > sharedBoundaryLengths[cc]) { + if (dot(chart2->planeNormal, chart->planeNormal) > -0.25) { + mergeChart(chart2, chart, sharedBoundaryLengths[cc]); + delete chart; + chartArray[c] = NULL; + break; + } + } + } + if (sharedBoundaryLengths[cc] > 0.20 * std::max(0.0f, chart->boundaryLength - externalBoundary)) { + // Compare proxies. + if (dot(chart2->planeNormal, chart->planeNormal) > 0) { + mergeChart(chart2, chart, sharedBoundaryLengths[cc]); + delete chart; + chartArray[c] = NULL; + break; + } + } + } + } + // Remove deleted charts. + for (int c = 0; c < int32_t(chartArray.size()); /*do not increment if removed*/) { + if (chartArray[c] == NULL) { + chartArray.erase(chartArray.begin() + c); + // Update faceChartArray. + const uint32_t faceCount = faceChartArray.size(); + for (uint32_t i = 0; i < faceCount; i++) { + xaDebugAssert(faceChartArray[i] != -1); + xaDebugAssert(faceChartArray[i] != c); + xaDebugAssert(faceChartArray[i] <= int32_t(chartArray.size())); + if (faceChartArray[i] > c) { + faceChartArray[i]--; + } + } + } else { + chartArray[c]->id = c; + c++; + } + } + } + + // @@ Cleanup. + struct Candidate { + uint32_t face; + ChartBuildData *chart; + float metric; + }; + + // @@ Get N best candidates in one pass. + const Candidate &getBestCandidate() const { + uint32_t best = 0; + float bestCandidateMetric = FLT_MAX; + const uint32_t candidateCount = candidateArray.size(); + xaAssert(candidateCount > 0); + for (uint32_t i = 0; i < candidateCount; i++) { + const Candidate &candidate = candidateArray[i]; + if (candidate.metric < bestCandidateMetric) { + bestCandidateMetric = candidate.metric; + best = i; + } + } + return candidateArray[best]; + } + + void removeCandidate(uint32_t f) { + int c = faceCandidateArray[f]; + if (c != -1) { + faceCandidateArray[f] = (uint32_t)-1; + if (c == int(candidateArray.size() - 1)) { + candidateArray.pop_back(); + } else { + // Replace with last. + candidateArray[c] = candidateArray[candidateArray.size() - 1]; + candidateArray.pop_back(); + faceCandidateArray[candidateArray[c].face] = c; + } + } + } + + void updateCandidate(ChartBuildData *chart, uint32_t f, float metric) { + if (faceCandidateArray[f] == -1) { + const uint32_t index = candidateArray.size(); + faceCandidateArray[f] = index; + candidateArray.resize(index + 1); + candidateArray[index].face = f; + candidateArray[index].chart = chart; + candidateArray[index].metric = metric; + } else { + int c = faceCandidateArray[f]; + xaDebugAssert(c != -1); + Candidate &candidate = candidateArray[c]; + xaDebugAssert(candidate.face == f); + if (metric < candidate.metric || chart == candidate.chart) { + candidate.metric = metric; + candidate.chart = chart; + } + } + } + + void mergeChart(ChartBuildData *owner, ChartBuildData *chart, float sharedBoundaryLength) { + const uint32_t faceCount = chart->faces.size(); + for (uint32_t i = 0; i < faceCount; i++) { + uint32_t f = chart->faces[i]; + xaDebugAssert(faceChartArray[f] == chart->id); + faceChartArray[f] = owner->id; + owner->faces.push_back(f); + } + // Update adjacencies? + owner->area += chart->area; + owner->boundaryLength += chart->boundaryLength - sharedBoundaryLength; + owner->normalSum += chart->normalSum; + owner->centroidSum += chart->centroidSum; + updateProxy(owner); + } + + uint32_t chartCount() const { return chartArray.size(); } + const std::vector<uint32_t> &chartFaces(uint32_t i) const { return chartArray[i]->faces; } + + const halfedge::Mesh *mesh; + uint32_t facesLeft; + std::vector<int> faceChartArray; + std::vector<ChartBuildData *> chartArray; + std::vector<float> shortestPaths; + std::vector<float> edgeLengths; + std::vector<float> faceAreas; + std::vector<Candidate> candidateArray; // + std::vector<uint32_t> faceCandidateArray; // Map face index to candidate index. + MTRand rand; + CharterOptions options; +}; + +/// A chart is a connected set of faces with a certain topology (usually a disk). +class Chart { +public: + Chart() : + m_isDisk(false), + m_isVertexMapped(false) {} + + void build(const halfedge::Mesh *originalMesh, const std::vector<uint32_t> &faceArray) { + // Copy face indices. + m_faceArray = faceArray; + const uint32_t meshVertexCount = originalMesh->vertexCount(); + m_chartMesh.reset(new halfedge::Mesh()); + m_unifiedMesh.reset(new halfedge::Mesh()); + std::vector<uint32_t> chartMeshIndices(meshVertexCount, (uint32_t)~0); + std::vector<uint32_t> unifiedMeshIndices(meshVertexCount, (uint32_t)~0); + // Add vertices. + const uint32_t faceCount = faceArray.size(); + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = originalMesh->faceAt(faceArray[f]); + xaDebugAssert(face != NULL); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Vertex *vertex = it.current()->vertex; + const halfedge::Vertex *unifiedVertex = vertex->firstColocal(); + if (unifiedMeshIndices[unifiedVertex->id] == ~0) { + unifiedMeshIndices[unifiedVertex->id] = m_unifiedMesh->vertexCount(); + xaDebugAssert(vertex->pos == unifiedVertex->pos); + m_unifiedMesh->addVertex(vertex->pos); + } + if (chartMeshIndices[vertex->id] == ~0) { + chartMeshIndices[vertex->id] = m_chartMesh->vertexCount(); + m_chartToOriginalMap.push_back(vertex->original_id); + m_chartToUnifiedMap.push_back(unifiedMeshIndices[unifiedVertex->id]); + halfedge::Vertex *v = m_chartMesh->addVertex(vertex->pos); + v->nor = vertex->nor; + v->tex = vertex->tex; + } + } + } + // This is ignoring the canonical map: + // - Is it really necessary to link colocals? + m_chartMesh->linkColocals(); + //m_unifiedMesh->linkColocals(); // Not strictly necessary, no colocals in the unified mesh. # Wrong. + // This check is not valid anymore, if the original mesh vertices were linked with a canonical map, then it might have + // some colocal vertices that were unlinked. So, the unified mesh might have some duplicate vertices, because firstColocal() + // is not guaranteed to return the same vertex for two colocal vertices. + //xaAssert(m_chartMesh->colocalVertexCount() == m_unifiedMesh->vertexCount()); + // Is that OK? What happens in meshes were that happens? Does anything break? Apparently not... + std::vector<uint32_t> faceIndices; + faceIndices.reserve(7); + // Add faces. + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = originalMesh->faceAt(faceArray[f]); + xaDebugAssert(face != NULL); + faceIndices.clear(); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Vertex *vertex = it.current()->vertex; + xaDebugAssert(vertex != NULL); + faceIndices.push_back(chartMeshIndices[vertex->id]); + } + m_chartMesh->addFace(faceIndices); + faceIndices.clear(); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Vertex *vertex = it.current()->vertex; + xaDebugAssert(vertex != NULL); + vertex = vertex->firstColocal(); + faceIndices.push_back(unifiedMeshIndices[vertex->id]); + } + m_unifiedMesh->addFace(faceIndices); + } + m_chartMesh->linkBoundary(); + m_unifiedMesh->linkBoundary(); + //exportMesh(m_unifiedMesh.ptr(), "debug_input.obj"); + if (m_unifiedMesh->splitBoundaryEdges()) { + m_unifiedMesh.reset(halfedge::unifyVertices(m_unifiedMesh.get())); + } + //exportMesh(m_unifiedMesh.ptr(), "debug_split.obj"); + // Closing the holes is not always the best solution and does not fix all the problems. + // We need to do some analysis of the holes and the genus to: + // - Find cuts that reduce genus. + // - Find cuts to connect holes. + // - Use minimal spanning trees or seamster. + if (!closeHoles()) { + /*static int pieceCount = 0; + StringBuilder fileName; + fileName.format("debug_hole_%d.obj", pieceCount++); + exportMesh(m_unifiedMesh.ptr(), fileName.str());*/ + } + m_unifiedMesh.reset(halfedge::triangulate(m_unifiedMesh.get())); + //exportMesh(m_unifiedMesh.ptr(), "debug_triangulated.obj"); + // Analyze chart topology. + halfedge::MeshTopology topology(m_unifiedMesh.get()); + m_isDisk = topology.isDisk(); + } + + void buildVertexMap(const halfedge::Mesh *originalMesh, const std::vector<uint32_t> &unchartedMaterialArray) { + xaAssert(m_chartMesh.get() == NULL && m_unifiedMesh.get() == NULL); + m_isVertexMapped = true; + // Build face indices. + m_faceArray.clear(); + const uint32_t meshFaceCount = originalMesh->faceCount(); + for (uint32_t f = 0; f < meshFaceCount; f++) { + const halfedge::Face *face = originalMesh->faceAt(f); + if (std::find(unchartedMaterialArray.begin(), unchartedMaterialArray.end(), face->material) != unchartedMaterialArray.end()) { + m_faceArray.push_back(f); + } + } + const uint32_t faceCount = m_faceArray.size(); + if (faceCount == 0) { + return; + } + // @@ The chartMesh construction is basically the same as with regular charts, don't duplicate! + const uint32_t meshVertexCount = originalMesh->vertexCount(); + m_chartMesh.reset(new halfedge::Mesh()); + std::vector<uint32_t> chartMeshIndices(meshVertexCount, (uint32_t)~0); + // Vertex map mesh only has disconnected vertices. + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = originalMesh->faceAt(m_faceArray[f]); + xaDebugAssert(face != NULL); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Vertex *vertex = it.current()->vertex; + if (chartMeshIndices[vertex->id] == ~0) { + chartMeshIndices[vertex->id] = m_chartMesh->vertexCount(); + m_chartToOriginalMap.push_back(vertex->original_id); + halfedge::Vertex *v = m_chartMesh->addVertex(vertex->pos); + v->nor = vertex->nor; + v->tex = vertex->tex; // @@ Not necessary. + } + } + } + // @@ Link colocals using the original mesh canonical map? Build canonical map on the fly? Do we need to link colocals at all for this? + //m_chartMesh->linkColocals(); + std::vector<uint32_t> faceIndices; + faceIndices.reserve(7); + // Add faces. + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = originalMesh->faceAt(m_faceArray[f]); + xaDebugAssert(face != NULL); + faceIndices.clear(); + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Vertex *vertex = it.current()->vertex; + xaDebugAssert(vertex != NULL); + xaDebugAssert(chartMeshIndices[vertex->id] != ~0); + faceIndices.push_back(chartMeshIndices[vertex->id]); + } + halfedge::Face *new_face = m_chartMesh->addFace(faceIndices); + xaDebugAssert(new_face != NULL); +#ifdef NDEBUG + new_face = NULL; // silence unused parameter warning +#endif + } + m_chartMesh->linkBoundary(); + const uint32_t chartVertexCount = m_chartMesh->vertexCount(); + Box bounds; + bounds.clearBounds(); + for (uint32_t i = 0; i < chartVertexCount; i++) { + halfedge::Vertex *vertex = m_chartMesh->vertexAt(i); + bounds.addPointToBounds(vertex->pos); + } + ProximityGrid grid; + grid.init(bounds, chartVertexCount); + for (uint32_t i = 0; i < chartVertexCount; i++) { + halfedge::Vertex *vertex = m_chartMesh->vertexAt(i); + grid.add(vertex->pos, i); + } + uint32_t texelCount = 0; + const float positionThreshold = 0.01f; + const float normalThreshold = 0.01f; + uint32_t verticesVisited = 0; + uint32_t cellsVisited = 0; + std::vector<int> vertexIndexArray(chartVertexCount, -1); // Init all indices to -1. + // Traverse vertices in morton order. @@ It may be more interesting to sort them based on orientation. + const uint32_t cellCodeCount = grid.mortonCount(); + for (uint32_t cellCode = 0; cellCode < cellCodeCount; cellCode++) { + int cell = grid.mortonIndex(cellCode); + if (cell < 0) continue; + cellsVisited++; + const std::vector<uint32_t> &indexArray = grid.cellArray[cell].indexArray; + for (uint32_t i = 0; i < indexArray.size(); i++) { + uint32_t idx = indexArray[i]; + halfedge::Vertex *vertex = m_chartMesh->vertexAt(idx); + xaDebugAssert(vertexIndexArray[idx] == -1); + std::vector<uint32_t> neighbors; + grid.gather(vertex->pos, positionThreshold, /*ref*/ neighbors); + // Compare against all nearby vertices, cluster greedily. + for (uint32_t j = 0; j < neighbors.size(); j++) { + uint32_t otherIdx = neighbors[j]; + if (vertexIndexArray[otherIdx] != -1) { + halfedge::Vertex *otherVertex = m_chartMesh->vertexAt(otherIdx); + if (distance(vertex->pos, otherVertex->pos) < positionThreshold && + distance(vertex->nor, otherVertex->nor) < normalThreshold) { + vertexIndexArray[idx] = vertexIndexArray[otherIdx]; + break; + } + } + } + // If index not assigned, assign new one. + if (vertexIndexArray[idx] == -1) { + vertexIndexArray[idx] = texelCount++; + } + verticesVisited++; + } + } + xaDebugAssert(cellsVisited == grid.cellArray.size()); + xaDebugAssert(verticesVisited == chartVertexCount); + vertexMapWidth = ftoi_ceil(sqrtf(float(texelCount))); + vertexMapWidth = (vertexMapWidth + 3) & ~3; // Width aligned to 4. + vertexMapHeight = vertexMapWidth == 0 ? 0 : (texelCount + vertexMapWidth - 1) / vertexMapWidth; + //vertexMapHeight = (vertexMapHeight + 3) & ~3; // Height aligned to 4. + xaDebugAssert(vertexMapWidth >= vertexMapHeight); + xaPrint("Reduced vertex count from %d to %d.\n", chartVertexCount, texelCount); + // Lay down the clustered vertices in morton order. + std::vector<uint32_t> texelCodes(texelCount); + // For each texel, assign one morton code. + uint32_t texelCode = 0; + for (uint32_t i = 0; i < texelCount; i++) { + uint32_t x, y; + do { + x = morton::decodeMorton2X(texelCode); + y = morton::decodeMorton2Y(texelCode); + texelCode++; + } while (x >= uint32_t(vertexMapWidth) || y >= uint32_t(vertexMapHeight)); + texelCodes[i] = texelCode - 1; + } + for (uint32_t i = 0; i < chartVertexCount; i++) { + halfedge::Vertex *vertex = m_chartMesh->vertexAt(i); + int idx = vertexIndexArray[i]; + if (idx != -1) { + uint32_t tc = texelCodes[idx]; + uint32_t x = morton::decodeMorton2X(tc); + uint32_t y = morton::decodeMorton2Y(tc); + vertex->tex.x = float(x); + vertex->tex.y = float(y); + } + } + } + + bool closeHoles() { + xaDebugAssert(!m_isVertexMapped); + std::vector<halfedge::Edge *> boundaryEdges; + getBoundaryEdges(m_unifiedMesh.get(), boundaryEdges); + uint32_t boundaryCount = boundaryEdges.size(); + if (boundaryCount <= 1) { + // Nothing to close. + return true; + } + // Compute lengths and areas. + std::vector<float> boundaryLengths; + for (uint32_t i = 0; i < boundaryCount; i++) { + const halfedge::Edge *startEdge = boundaryEdges[i]; + xaAssert(startEdge->face == NULL); + //float boundaryEdgeCount = 0; + float boundaryLength = 0.0f; + //Vector3 boundaryCentroid(zero); + const halfedge::Edge *edge = startEdge; + do { + Vector3 t0 = edge->from()->pos; + Vector3 t1 = edge->to()->pos; + //boundaryEdgeCount++; + boundaryLength += length(t1 - t0); + //boundaryCentroid += edge->vertex()->pos; + edge = edge->next; + } while (edge != startEdge); + boundaryLengths.push_back(boundaryLength); + //boundaryCentroids.append(boundaryCentroid / boundaryEdgeCount); + } + // Find disk boundary. + uint32_t diskBoundary = 0; + float maxLength = boundaryLengths[0]; + for (uint32_t i = 1; i < boundaryCount; i++) { + if (boundaryLengths[i] > maxLength) { + maxLength = boundaryLengths[i]; + diskBoundary = i; + } + } + // Close holes. + for (uint32_t i = 0; i < boundaryCount; i++) { + if (diskBoundary == i) { + // Skip disk boundary. + continue; + } + halfedge::Edge *startEdge = boundaryEdges[i]; + xaDebugAssert(startEdge != NULL); + xaDebugAssert(startEdge->face == NULL); + std::vector<halfedge::Vertex *> vertexLoop; + std::vector<halfedge::Edge *> edgeLoop; + halfedge::Edge *edge = startEdge; + do { + halfedge::Vertex *vertex = edge->next->vertex; // edge->to() + uint32_t j; + for (j = 0; j < vertexLoop.size(); j++) { + if (vertex->isColocal(vertexLoop[j])) { + break; + } + } + bool isCrossing = (j != vertexLoop.size()); + if (isCrossing) { + halfedge::Edge *prev = edgeLoop[j]; // Previous edge before the loop. + halfedge::Edge *next = edge->next; // Next edge after the loop. + xaDebugAssert(prev->to()->isColocal(next->from())); + // Close loop. + edgeLoop.push_back(edge); + closeLoop(j + 1, edgeLoop); + // Link boundary loop. + prev->setNext(next); + vertex->setEdge(next); + // Start over again. + vertexLoop.clear(); + edgeLoop.clear(); + edge = startEdge; + vertex = edge->to(); + } + vertexLoop.push_back(vertex); + edgeLoop.push_back(edge); + edge = edge->next; + } while (edge != startEdge); + closeLoop(0, edgeLoop); + } + getBoundaryEdges(m_unifiedMesh.get(), boundaryEdges); + boundaryCount = boundaryEdges.size(); + xaDebugAssert(boundaryCount == 1); + return boundaryCount == 1; + } + + bool isDisk() const { + return m_isDisk; + } + bool isVertexMapped() const { + return m_isVertexMapped; + } + + uint32_t vertexCount() const { + return m_chartMesh->vertexCount(); + } + uint32_t colocalVertexCount() const { + return m_unifiedMesh->vertexCount(); + } + + uint32_t faceCount() const { + return m_faceArray.size(); + } + uint32_t faceAt(uint32_t i) const { + return m_faceArray[i]; + } + + const halfedge::Mesh *chartMesh() const { + return m_chartMesh.get(); + } + halfedge::Mesh *chartMesh() { + return m_chartMesh.get(); + } + const halfedge::Mesh *unifiedMesh() const { + return m_unifiedMesh.get(); + } + halfedge::Mesh *unifiedMesh() { + return m_unifiedMesh.get(); + } + + //uint32_t vertexIndex(uint32_t i) const { return m_vertexIndexArray[i]; } + + uint32_t mapChartVertexToOriginalVertex(uint32_t i) const { + return m_chartToOriginalMap[i]; + } + uint32_t mapChartVertexToUnifiedVertex(uint32_t i) const { + return m_chartToUnifiedMap[i]; + } + + const std::vector<uint32_t> &faceArray() const { + return m_faceArray; + } + + // Transfer parameterization from unified mesh to chart mesh. + void transferParameterization() { + xaDebugAssert(!m_isVertexMapped); + uint32_t vertexCount = m_chartMesh->vertexCount(); + for (uint32_t v = 0; v < vertexCount; v++) { + halfedge::Vertex *vertex = m_chartMesh->vertexAt(v); + halfedge::Vertex *unifiedVertex = m_unifiedMesh->vertexAt(mapChartVertexToUnifiedVertex(v)); + vertex->tex = unifiedVertex->tex; + } + } + + float computeSurfaceArea() const { + return halfedge::computeSurfaceArea(m_chartMesh.get()) * scale; + } + + float computeParametricArea() const { + // This only makes sense in parameterized meshes. + xaDebugAssert(m_isDisk); + xaDebugAssert(!m_isVertexMapped); + return halfedge::computeParametricArea(m_chartMesh.get()); + } + + Vector2 computeParametricBounds() const { + // This only makes sense in parameterized meshes. + xaDebugAssert(m_isDisk); + xaDebugAssert(!m_isVertexMapped); + Box bounds; + bounds.clearBounds(); + uint32_t vertexCount = m_chartMesh->vertexCount(); + for (uint32_t v = 0; v < vertexCount; v++) { + halfedge::Vertex *vertex = m_chartMesh->vertexAt(v); + bounds.addPointToBounds(Vector3(vertex->tex, 0)); + } + return bounds.extents().xy(); + } + + float scale = 1.0f; + uint32_t vertexMapWidth; + uint32_t vertexMapHeight; + bool blockAligned = true; + +private: + bool closeLoop(uint32_t start, const std::vector<halfedge::Edge *> &loop) { + const uint32_t vertexCount = loop.size() - start; + xaDebugAssert(vertexCount >= 3); + if (vertexCount < 3) return false; + xaDebugAssert(loop[start]->vertex->isColocal(loop[start + vertexCount - 1]->to())); + // If the hole is planar, then we add a single face that will be properly triangulated later. + // If the hole is not planar, we add a triangle fan with a vertex at the hole centroid. + // This is still a bit of a hack. There surely are better hole filling algorithms out there. + std::vector<Vector3> points(vertexCount); + for (uint32_t i = 0; i < vertexCount; i++) { + points[i] = loop[start + i]->vertex->pos; + } + bool isPlanar = Fit::isPlanar(vertexCount, points.data()); + if (isPlanar) { + // Add face and connect edges. + halfedge::Face *face = m_unifiedMesh->addFace(); + for (uint32_t i = 0; i < vertexCount; i++) { + halfedge::Edge *edge = loop[start + i]; + edge->face = face; + edge->setNext(loop[start + (i + 1) % vertexCount]); + } + face->edge = loop[start]; + xaDebugAssert(face->isValid()); + } else { + // If the polygon is not planar, we just cross our fingers, and hope this will work: + // Compute boundary centroid: + Vector3 centroidPos(0); + for (uint32_t i = 0; i < vertexCount; i++) { + centroidPos += points[i]; + } + centroidPos *= (1.0f / vertexCount); + halfedge::Vertex *centroid = m_unifiedMesh->addVertex(centroidPos); + // Add one pair of edges for each boundary vertex. + for (uint32_t j = vertexCount - 1, i = 0; i < vertexCount; j = i++) { + halfedge::Face *face = m_unifiedMesh->addFace(centroid->id, loop[start + j]->vertex->id, loop[start + i]->vertex->id); + xaDebugAssert(face != NULL); +#ifdef NDEBUG + face = NULL; // silence unused parameter warning +#endif + } + } + return true; + } + + static void getBoundaryEdges(halfedge::Mesh *mesh, std::vector<halfedge::Edge *> &boundaryEdges) { + xaDebugAssert(mesh != NULL); + const uint32_t edgeCount = mesh->edgeCount(); + BitArray bitFlags(edgeCount); + bitFlags.clearAll(); + boundaryEdges.clear(); + // Search for boundary edges. Mark all the edges that belong to the same boundary. + for (uint32_t e = 0; e < edgeCount; e++) { + halfedge::Edge *startEdge = mesh->edgeAt(e); + if (startEdge != NULL && startEdge->isBoundary() && bitFlags.bitAt(e) == false) { + xaDebugAssert(startEdge->face != NULL); + xaDebugAssert(startEdge->pair->face == NULL); + startEdge = startEdge->pair; + const halfedge::Edge *edge = startEdge; + do { + xaDebugAssert(edge->face == NULL); + xaDebugAssert(bitFlags.bitAt(edge->id / 2) == false); + bitFlags.setBitAt(edge->id / 2); + edge = edge->next; + } while (startEdge != edge); + boundaryEdges.push_back(startEdge); + } + } + } + + // Chart mesh. + std::auto_ptr<halfedge::Mesh> m_chartMesh; + + std::auto_ptr<halfedge::Mesh> m_unifiedMesh; + bool m_isDisk; + bool m_isVertexMapped; + + // List of faces of the original mesh that belong to this chart. + std::vector<uint32_t> m_faceArray; + + // Map vertices of the chart mesh to vertices of the original mesh. + std::vector<uint32_t> m_chartToOriginalMap; + + std::vector<uint32_t> m_chartToUnifiedMap; +}; + +// Estimate quality of existing parameterization. +class ParameterizationQuality { +public: + ParameterizationQuality() { + m_totalTriangleCount = 0; + m_flippedTriangleCount = 0; + m_zeroAreaTriangleCount = 0; + m_parametricArea = 0.0f; + m_geometricArea = 0.0f; + m_stretchMetric = 0.0f; + m_maxStretchMetric = 0.0f; + m_conformalMetric = 0.0f; + m_authalicMetric = 0.0f; + } + + ParameterizationQuality(const halfedge::Mesh *mesh) { + xaDebugAssert(mesh != NULL); + m_totalTriangleCount = 0; + m_flippedTriangleCount = 0; + m_zeroAreaTriangleCount = 0; + m_parametricArea = 0.0f; + m_geometricArea = 0.0f; + m_stretchMetric = 0.0f; + m_maxStretchMetric = 0.0f; + m_conformalMetric = 0.0f; + m_authalicMetric = 0.0f; + const uint32_t faceCount = mesh->faceCount(); + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = mesh->faceAt(f); + const halfedge::Vertex *vertex0 = NULL; + Vector3 p[3]; + Vector2 t[3]; + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + if (vertex0 == NULL) { + vertex0 = edge->vertex; + p[0] = vertex0->pos; + t[0] = vertex0->tex; + } else if (edge->to() != vertex0) { + p[1] = edge->from()->pos; + p[2] = edge->to()->pos; + t[1] = edge->from()->tex; + t[2] = edge->to()->tex; + processTriangle(p, t); + } + } + } + if (m_flippedTriangleCount + m_zeroAreaTriangleCount == faceCount) { + // If all triangles are flipped, then none is. + m_flippedTriangleCount = 0; + } + xaDebugAssert(std::isfinite(m_parametricArea) && m_parametricArea >= 0); + xaDebugAssert(std::isfinite(m_geometricArea) && m_geometricArea >= 0); + xaDebugAssert(std::isfinite(m_stretchMetric)); + xaDebugAssert(std::isfinite(m_maxStretchMetric)); + xaDebugAssert(std::isfinite(m_conformalMetric)); + xaDebugAssert(std::isfinite(m_authalicMetric)); + } + + bool isValid() const { + return m_flippedTriangleCount == 0; // @@ Does not test for self-overlaps. + } + + float rmsStretchMetric() const { + if (m_geometricArea == 0) return 0.0f; + float normFactor = sqrtf(m_parametricArea / m_geometricArea); + return sqrtf(m_stretchMetric / m_geometricArea) * normFactor; + } + + float maxStretchMetric() const { + if (m_geometricArea == 0) return 0.0f; + float normFactor = sqrtf(m_parametricArea / m_geometricArea); + return m_maxStretchMetric * normFactor; + } + + float rmsConformalMetric() const { + if (m_geometricArea == 0) return 0.0f; + return sqrtf(m_conformalMetric / m_geometricArea); + } + + float maxAuthalicMetric() const { + if (m_geometricArea == 0) return 0.0f; + return sqrtf(m_authalicMetric / m_geometricArea); + } + + void operator+=(const ParameterizationQuality &pq) { + m_totalTriangleCount += pq.m_totalTriangleCount; + m_flippedTriangleCount += pq.m_flippedTriangleCount; + m_zeroAreaTriangleCount += pq.m_zeroAreaTriangleCount; + m_parametricArea += pq.m_parametricArea; + m_geometricArea += pq.m_geometricArea; + m_stretchMetric += pq.m_stretchMetric; + m_maxStretchMetric = std::max(m_maxStretchMetric, pq.m_maxStretchMetric); + m_conformalMetric += pq.m_conformalMetric; + m_authalicMetric += pq.m_authalicMetric; + } + +private: + void processTriangle(Vector3 q[3], Vector2 p[3]) { + m_totalTriangleCount++; + // Evaluate texture stretch metric. See: + // - "Texture Mapping Progressive Meshes", Sander, Snyder, Gortler & Hoppe + // - "Mesh Parameterization: Theory and Practice", Siggraph'07 Course Notes, Hormann, Levy & Sheffer. + float t1 = p[0].x; + float s1 = p[0].y; + float t2 = p[1].x; + float s2 = p[1].y; + float t3 = p[2].x; + float s3 = p[2].y; + float geometricArea = length(cross(q[1] - q[0], q[2] - q[0])) / 2; + float parametricArea = ((s2 - s1) * (t3 - t1) - (s3 - s1) * (t2 - t1)) / 2; + if (isZero(parametricArea)) { + m_zeroAreaTriangleCount++; + return; + } + Vector3 Ss = (q[0] * (t2 - t3) + q[1] * (t3 - t1) + q[2] * (t1 - t2)) / (2 * parametricArea); + Vector3 St = (q[0] * (s3 - s2) + q[1] * (s1 - s3) + q[2] * (s2 - s1)) / (2 * parametricArea); + float a = dot(Ss, Ss); // E + float b = dot(Ss, St); // F + float c = dot(St, St); // G + // Compute eigen-values of the first fundamental form: + float sigma1 = sqrtf(0.5f * std::max(0.0f, a + c - sqrtf(square(a - c) + 4 * square(b)))); // gamma uppercase, min eigenvalue. + float sigma2 = sqrtf(0.5f * std::max(0.0f, a + c + sqrtf(square(a - c) + 4 * square(b)))); // gamma lowercase, max eigenvalue. + xaAssert(sigma2 >= sigma1); + // isometric: sigma1 = sigma2 = 1 + // conformal: sigma1 / sigma2 = 1 + // authalic: sigma1 * sigma2 = 1 + float rmsStretch = sqrtf((a + c) * 0.5f); + float rmsStretch2 = sqrtf((square(sigma1) + square(sigma2)) * 0.5f); + xaDebugAssert(equal(rmsStretch, rmsStretch2, 0.01f)); +#ifdef NDEBUG + rmsStretch2 = 0; // silence unused parameter warning +#endif + if (parametricArea < 0.0f) { + // Count flipped triangles. + m_flippedTriangleCount++; + parametricArea = fabsf(parametricArea); + } + m_stretchMetric += square(rmsStretch) * geometricArea; + m_maxStretchMetric = std::max(m_maxStretchMetric, sigma2); + if (!isZero(sigma1, 0.000001f)) { + // sigma1 is zero when geometricArea is zero. + m_conformalMetric += (sigma2 / sigma1) * geometricArea; + } + m_authalicMetric += (sigma1 * sigma2) * geometricArea; + // Accumulate total areas. + m_geometricArea += geometricArea; + m_parametricArea += parametricArea; + //triangleConformalEnergy(q, p); + } + + uint32_t m_totalTriangleCount; + uint32_t m_flippedTriangleCount; + uint32_t m_zeroAreaTriangleCount; + float m_parametricArea; + float m_geometricArea; + float m_stretchMetric; + float m_maxStretchMetric; + float m_conformalMetric; + float m_authalicMetric; +}; + +// Set of charts corresponding to a single mesh. +class MeshCharts { +public: + MeshCharts(const halfedge::Mesh *mesh) : + m_mesh(mesh) {} + + ~MeshCharts() { + for (size_t i = 0; i < m_chartArray.size(); i++) + delete m_chartArray[i]; + } + + uint32_t chartCount() const { + return m_chartArray.size(); + } + uint32_t vertexCount() const { + return m_totalVertexCount; + } + + const Chart *chartAt(uint32_t i) const { + return m_chartArray[i]; + } + Chart *chartAt(uint32_t i) { + return m_chartArray[i]; + } + + // Extract the charts of the input mesh. + void extractCharts() { + const uint32_t faceCount = m_mesh->faceCount(); + int first = 0; + std::vector<uint32_t> queue; + queue.reserve(faceCount); + BitArray bitFlags(faceCount); + bitFlags.clearAll(); + for (uint32_t f = 0; f < faceCount; f++) { + if (bitFlags.bitAt(f) == false) { + // Start new patch. Reset queue. + first = 0; + queue.clear(); + queue.push_back(f); + bitFlags.setBitAt(f); + while (first != (int)queue.size()) { + const halfedge::Face *face = m_mesh->faceAt(queue[first]); + // Visit face neighbors of queue[first] + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + const halfedge::Edge *edge = it.current(); + xaDebugAssert(edge->pair != NULL); + if (!edge->isBoundary() && /*!edge->isSeam()*/ + //!(edge->from()->tex() != edge->pair()->to()->tex() || edge->to()->tex() != edge->pair()->from()->tex())) + !(edge->from() != edge->pair->to() || edge->to() != edge->pair->from())) { // Preserve existing seams (not just texture seams). + const halfedge::Face *neighborFace = edge->pair->face; + xaDebugAssert(neighborFace != NULL); + if (bitFlags.bitAt(neighborFace->id) == false) { + queue.push_back(neighborFace->id); + bitFlags.setBitAt(neighborFace->id); + } + } + } + first++; + } + Chart *chart = new Chart(); + chart->build(m_mesh, queue); + m_chartArray.push_back(chart); + } + } + } + + /* + Compute charts using a simple segmentation algorithm. + + LSCM: + - identify sharp features using local dihedral angles. + - identify seed faces farthest from sharp features. + - grow charts from these seeds. + + MCGIM: + - phase 1: chart growth + - grow all charts simultaneously using dijkstra search on the dual graph of the mesh. + - graph edges are weighted based on planarity metric. + - metric uses distance to global chart normal. + - terminate when all faces have been assigned. + - phase 2: seed computation: + - place new seed of the chart at the most interior face. + - most interior is evaluated using distance metric only. + + - method repeates the two phases, until the location of the seeds does not change. + - cycles are detected by recording all the previous seeds and chartification terminates. + + D-Charts: + + - Uniaxial conic metric: + - N_c = axis of the generalized cone that best fits the chart. (cone can a be cylinder or a plane). + - omega_c = angle between the face normals and the axis. + - Fitting error between chart C and tringle t: F(c,t) = (N_c*n_t - cos(omega_c))^2 + + - Compactness metrics: + - Roundness: + - C(c,t) = pi * D(S_c,t)^2 / A_c + - S_c = chart seed. + - D(S_c,t) = length of the shortest path inside the chart betwen S_c and t. + - A_c = chart area. + - Straightness: + - P(c,t) = l_out(c,t) / l_in(c,t) + - l_out(c,t) = lenght of the edges not shared between C and t. + - l_in(c,t) = lenght of the edges shared between C and t. + + - Combined metric: + - Cost(c,t) = F(c,t)^alpha + C(c,t)^beta + P(c,t)^gamma + - alpha = 1, beta = 0.7, gamma = 0.5 + + Our basic approach: + - Just one iteration of k-means? + - Avoid dijkstra by greedily growing charts until a threshold is met. Increase threshold and repeat until no faces left. + - If distortion metric is too high, split chart, add two seeds. + - If chart size is low, try removing chart. + + Postprocess: + - If topology is not disk: + - Fill holes, if new faces fit proxy. + - Find best cut, otherwise. + - After parameterization: + - If boundary self-intersects: + - cut chart along the closest two diametral boundary vertices, repeat parametrization. + - what if the overlap is on an appendix? How do we find that out and cut appropiately? + - emphasize roundness metrics to prevent those cases. + - If interior self-overlaps: preserve boundary parameterization and use mean-value map. + */ + void computeCharts(const CharterOptions &options, const std::vector<uint32_t> &unchartedMaterialArray) { + Chart *vertexMap = NULL; + if (unchartedMaterialArray.size() != 0) { + vertexMap = new Chart(); + vertexMap->buildVertexMap(m_mesh, unchartedMaterialArray); + if (vertexMap->faceCount() == 0) { + delete vertexMap; + vertexMap = NULL; + } + } + AtlasBuilder builder(m_mesh); + if (vertexMap != NULL) { + // Mark faces that do not need to be charted. + builder.markUnchartedFaces(vertexMap->faceArray()); + m_chartArray.push_back(vertexMap); + } + if (builder.facesLeft != 0) { + // Tweak these values: + const float maxThreshold = 2; + const uint32_t growFaceCount = 32; + const uint32_t maxIterations = 4; + builder.options = options; + //builder.options.proxyFitMetricWeight *= 0.75; // relax proxy fit weight during initial seed placement. + //builder.options.roundnessMetricWeight = 0; + //builder.options.straightnessMetricWeight = 0; + // This seems a reasonable estimate. + uint32_t maxSeedCount = std::max(6U, builder.facesLeft); + // Create initial charts greedely. + xaPrint("### Placing seeds\n"); + builder.placeSeeds(maxThreshold, maxSeedCount); + xaPrint("### Placed %d seeds (max = %d)\n", builder.chartCount(), maxSeedCount); + builder.updateProxies(); + builder.mergeCharts(); +#if 1 + xaPrint("### Relocating seeds\n"); + builder.relocateSeeds(); + xaPrint("### Reset charts\n"); + builder.resetCharts(); + if (vertexMap != NULL) { + builder.markUnchartedFaces(vertexMap->faceArray()); + } + builder.options = options; + xaPrint("### Growing charts\n"); + // Restart process growing charts in parallel. + uint32_t iteration = 0; + while (true) { + if (!builder.growCharts(maxThreshold, growFaceCount)) { + xaPrint("### Can't grow anymore\n"); + // If charts cannot grow more: fill holes, merge charts, relocate seeds and start new iteration. + xaPrint("### Filling holes\n"); + builder.fillHoles(maxThreshold); + xaPrint("### Using %d charts now\n", builder.chartCount()); + builder.updateProxies(); + xaPrint("### Merging charts\n"); + builder.mergeCharts(); + xaPrint("### Using %d charts now\n", builder.chartCount()); + xaPrint("### Reseeding\n"); + if (!builder.relocateSeeds()) { + xaPrint("### Cannot relocate seeds anymore\n"); + // Done! + break; + } + if (iteration == maxIterations) { + xaPrint("### Reached iteration limit\n"); + break; + } + iteration++; + xaPrint("### Reset charts\n"); + builder.resetCharts(); + if (vertexMap != NULL) { + builder.markUnchartedFaces(vertexMap->faceArray()); + } + xaPrint("### Growing charts\n"); + } + }; +#endif + // Make sure no holes are left! + xaDebugAssert(builder.facesLeft == 0); + const uint32_t chartCount = builder.chartArray.size(); + for (uint32_t i = 0; i < chartCount; i++) { + Chart *chart = new Chart(); + m_chartArray.push_back(chart); + chart->build(m_mesh, builder.chartFaces(i)); + } + } + const uint32_t chartCount = m_chartArray.size(); + // Build face indices. + m_faceChart.resize(m_mesh->faceCount()); + m_faceIndex.resize(m_mesh->faceCount()); + for (uint32_t i = 0; i < chartCount; i++) { + const Chart *chart = m_chartArray[i]; + const uint32_t faceCount = chart->faceCount(); + for (uint32_t f = 0; f < faceCount; f++) { + uint32_t idx = chart->faceAt(f); + m_faceChart[idx] = i; + m_faceIndex[idx] = f; + } + } + // Build an exclusive prefix sum of the chart vertex counts. + m_chartVertexCountPrefixSum.resize(chartCount); + if (chartCount > 0) { + m_chartVertexCountPrefixSum[0] = 0; + for (uint32_t i = 1; i < chartCount; i++) { + const Chart *chart = m_chartArray[i - 1]; + m_chartVertexCountPrefixSum[i] = m_chartVertexCountPrefixSum[i - 1] + chart->vertexCount(); + } + m_totalVertexCount = m_chartVertexCountPrefixSum[chartCount - 1] + m_chartArray[chartCount - 1]->vertexCount(); + } else { + m_totalVertexCount = 0; + } + } + + void parameterizeCharts() { + ParameterizationQuality globalParameterizationQuality; + // Parameterize the charts. + uint32_t diskCount = 0; + const uint32_t chartCount = m_chartArray.size(); + for (uint32_t i = 0; i < chartCount; i++) { + Chart *chart = m_chartArray[i]; + + bool isValid = false; + + if (chart->isVertexMapped()) { + continue; + } + + if (chart->isDisk()) { + diskCount++; + ParameterizationQuality chartParameterizationQuality; + if (chart->faceCount() == 1) { + computeSingleFaceMap(chart->unifiedMesh()); + chartParameterizationQuality = ParameterizationQuality(chart->unifiedMesh()); + } else { + computeOrthogonalProjectionMap(chart->unifiedMesh()); + ParameterizationQuality orthogonalQuality(chart->unifiedMesh()); + computeLeastSquaresConformalMap(chart->unifiedMesh()); + ParameterizationQuality lscmQuality(chart->unifiedMesh()); + chartParameterizationQuality = lscmQuality; + } + isValid = chartParameterizationQuality.isValid(); + if (!isValid) { + xaPrint("*** Invalid parameterization.\n"); + } + // @@ Check that parameterization quality is above a certain threshold. + // @@ Detect boundary self-intersections. + globalParameterizationQuality += chartParameterizationQuality; + } + + // Transfer parameterization from unified mesh to chart mesh. + chart->transferParameterization(); + } + xaPrint(" Parameterized %d/%d charts.\n", diskCount, chartCount); + xaPrint(" RMS stretch metric: %f\n", globalParameterizationQuality.rmsStretchMetric()); + xaPrint(" MAX stretch metric: %f\n", globalParameterizationQuality.maxStretchMetric()); + xaPrint(" RMS conformal metric: %f\n", globalParameterizationQuality.rmsConformalMetric()); + xaPrint(" RMS authalic metric: %f\n", globalParameterizationQuality.maxAuthalicMetric()); + } + + uint32_t faceChartAt(uint32_t i) const { + return m_faceChart[i]; + } + uint32_t faceIndexWithinChartAt(uint32_t i) const { + return m_faceIndex[i]; + } + + uint32_t vertexCountBeforeChartAt(uint32_t i) const { + return m_chartVertexCountPrefixSum[i]; + } + +private: + const halfedge::Mesh *m_mesh; + + std::vector<Chart *> m_chartArray; + + std::vector<uint32_t> m_chartVertexCountPrefixSum; + uint32_t m_totalVertexCount; + + std::vector<uint32_t> m_faceChart; // the chart of every face of the input mesh. + std::vector<uint32_t> m_faceIndex; // the index within the chart for every face of the input mesh. +}; + +/// An atlas is a set of charts. +class Atlas { +public: + ~Atlas() { + for (size_t i = 0; i < m_meshChartsArray.size(); i++) + delete m_meshChartsArray[i]; + } + + uint32_t meshCount() const { + return m_meshChartsArray.size(); + } + + const MeshCharts *meshAt(uint32_t i) const { + return m_meshChartsArray[i]; + } + + MeshCharts *meshAt(uint32_t i) { + return m_meshChartsArray[i]; + } + + uint32_t chartCount() const { + uint32_t count = 0; + for (uint32_t c = 0; c < m_meshChartsArray.size(); c++) { + count += m_meshChartsArray[c]->chartCount(); + } + return count; + } + + const Chart *chartAt(uint32_t i) const { + for (uint32_t c = 0; c < m_meshChartsArray.size(); c++) { + uint32_t count = m_meshChartsArray[c]->chartCount(); + if (i < count) { + return m_meshChartsArray[c]->chartAt(i); + } + i -= count; + } + return NULL; + } + + Chart *chartAt(uint32_t i) { + for (uint32_t c = 0; c < m_meshChartsArray.size(); c++) { + uint32_t count = m_meshChartsArray[c]->chartCount(); + if (i < count) { + return m_meshChartsArray[c]->chartAt(i); + } + i -= count; + } + return NULL; + } + + // Add mesh charts and takes ownership. + // Extract the charts and add to this atlas. + void addMeshCharts(MeshCharts *meshCharts) { + m_meshChartsArray.push_back(meshCharts); + } + + void extractCharts(const halfedge::Mesh *mesh) { + MeshCharts *meshCharts = new MeshCharts(mesh); + meshCharts->extractCharts(); + addMeshCharts(meshCharts); + } + + void computeCharts(const halfedge::Mesh *mesh, const CharterOptions &options, const std::vector<uint32_t> &unchartedMaterialArray) { + MeshCharts *meshCharts = new MeshCharts(mesh); + meshCharts->computeCharts(options, unchartedMaterialArray); + addMeshCharts(meshCharts); + } + + void parameterizeCharts() { + for (uint32_t i = 0; i < m_meshChartsArray.size(); i++) { + m_meshChartsArray[i]->parameterizeCharts(); + } + } + +private: + std::vector<MeshCharts *> m_meshChartsArray; +}; + +struct AtlasPacker { + AtlasPacker(Atlas *atlas) : + m_atlas(atlas), + m_width(0), + m_height(0) { + // Save the original uvs. + m_originalChartUvs.resize(m_atlas->chartCount()); + for (uint32_t i = 0; i < m_atlas->chartCount(); i++) { + const halfedge::Mesh *mesh = atlas->chartAt(i)->chartMesh(); + m_originalChartUvs[i].resize(mesh->vertexCount()); + for (uint32_t j = 0; j < mesh->vertexCount(); j++) + m_originalChartUvs[i][j] = mesh->vertexAt(j)->tex; + } + } + + uint32_t getWidth() const { return m_width; } + uint32_t getHeight() const { return m_height; } + + // Pack charts in the smallest possible rectangle. + void packCharts(const PackerOptions &options) { + const uint32_t chartCount = m_atlas->chartCount(); + if (chartCount == 0) return; + float texelsPerUnit = 1; + if (options.method == PackMethod::TexelArea) + texelsPerUnit = options.texelArea; + for (int iteration = 0;; iteration++) { + m_rand = MTRand(); + std::vector<float> chartOrderArray(chartCount); + std::vector<Vector2> chartExtents(chartCount); + float meshArea = 0; + for (uint32_t c = 0; c < chartCount; c++) { + Chart *chart = m_atlas->chartAt(c); + if (!chart->isVertexMapped() && !chart->isDisk()) { + chartOrderArray[c] = 0; + // Skip non-disks. + continue; + } + Vector2 extents(0.0f); + if (chart->isVertexMapped()) { + // Arrange vertices in a rectangle. + extents.x = float(chart->vertexMapWidth); + extents.y = float(chart->vertexMapHeight); + } else { + // Compute surface area to sort charts. + float chartArea = chart->computeSurfaceArea(); + meshArea += chartArea; + //chartOrderArray[c] = chartArea; + // Compute chart scale + float parametricArea = fabsf(chart->computeParametricArea()); // @@ There doesn't seem to be anything preventing parametric area to be negative. + if (parametricArea < NV_EPSILON) { + // When the parametric area is too small we use a rough approximation to prevent divisions by very small numbers. + Vector2 bounds = chart->computeParametricBounds(); + parametricArea = bounds.x * bounds.y; + } + float scale = (chartArea / parametricArea) * texelsPerUnit; + if (parametricArea == 0) { // < NV_EPSILON) + scale = 0; + } + xaAssert(std::isfinite(scale)); + // Compute bounding box of chart. + Vector2 majorAxis, minorAxis, origin, end; + computeBoundingBox(chart, &majorAxis, &minorAxis, &origin, &end); + xaAssert(isFinite(majorAxis) && isFinite(minorAxis) && isFinite(origin)); + // Sort charts by perimeter. @@ This is sometimes producing somewhat unexpected results. Is this right? + //chartOrderArray[c] = ((end.x - origin.x) + (end.y - origin.y)) * scale; + // Translate, rotate and scale vertices. Compute extents. + halfedge::Mesh *mesh = chart->chartMesh(); + const uint32_t vertexCount = mesh->vertexCount(); + for (uint32_t i = 0; i < vertexCount; i++) { + halfedge::Vertex *vertex = mesh->vertexAt(i); + //Vector2 t = vertex->tex - origin; + Vector2 tmp; + tmp.x = dot(vertex->tex, majorAxis); + tmp.y = dot(vertex->tex, minorAxis); + tmp -= origin; + tmp *= scale; + if (tmp.x < 0 || tmp.y < 0) { + xaPrint("tmp: %f %f\n", tmp.x, tmp.y); + xaPrint("scale: %f\n", scale); + xaPrint("origin: %f %f\n", origin.x, origin.y); + xaPrint("majorAxis: %f %f\n", majorAxis.x, majorAxis.y); + xaPrint("minorAxis: %f %f\n", minorAxis.x, minorAxis.y); + xaDebugAssert(false); + } + //xaAssert(tmp.x >= 0 && tmp.y >= 0); + vertex->tex = tmp; + xaAssert(std::isfinite(vertex->tex.x) && std::isfinite(vertex->tex.y)); + extents = max(extents, tmp); + } + xaDebugAssert(extents.x >= 0 && extents.y >= 0); + // Limit chart size. + if (extents.x > 1024 || extents.y > 1024) { + float limit = std::max(extents.x, extents.y); + scale = 1024 / (limit + 1); + for (uint32_t i = 0; i < vertexCount; i++) { + halfedge::Vertex *vertex = mesh->vertexAt(i); + vertex->tex *= scale; + } + extents *= scale; + xaDebugAssert(extents.x <= 1024 && extents.y <= 1024); + } + // Scale the charts to use the entire texel area available. So, if the width is 0.1 we could scale it to 1 without increasing the lightmap usage and making a better + // use of it. In many cases this also improves the look of the seams, since vertices on the chart boundaries have more chances of being aligned with the texel centers. + float scale_x = 1.0f; + float scale_y = 1.0f; + float divide_x = 1.0f; + float divide_y = 1.0f; + if (extents.x > 0) { + int cw = ftoi_ceil(extents.x); + if (options.blockAlign && chart->blockAligned) { + // Align all chart extents to 4x4 blocks, but taking padding into account. + if (options.conservative) { + cw = align(cw + 2, 4) - 2; + } else { + cw = align(cw + 1, 4) - 1; + } + } + scale_x = (float(cw) - NV_EPSILON); + divide_x = extents.x; + extents.x = float(cw); + } + if (extents.y > 0) { + int ch = ftoi_ceil(extents.y); + if (options.blockAlign && chart->blockAligned) { + // Align all chart extents to 4x4 blocks, but taking padding into account. + if (options.conservative) { + ch = align(ch + 2, 4) - 2; + } else { + ch = align(ch + 1, 4) - 1; + } + } + scale_y = (float(ch) - NV_EPSILON); + divide_y = extents.y; + extents.y = float(ch); + } + for (uint32_t v = 0; v < vertexCount; v++) { + halfedge::Vertex *vertex = mesh->vertexAt(v); + vertex->tex.x /= divide_x; + vertex->tex.y /= divide_y; + vertex->tex.x *= scale_x; + vertex->tex.y *= scale_y; + xaAssert(std::isfinite(vertex->tex.x) && std::isfinite(vertex->tex.y)); + } + } + chartExtents[c] = extents; + // Sort charts by perimeter. + chartOrderArray[c] = extents.x + extents.y; + } + // @@ We can try to improve compression of small charts by sorting them by proximity like we do with vertex samples. + // @@ How to do that? One idea: compute chart centroid, insert into grid, compute morton index of the cell, sort based on morton index. + // @@ We would sort by morton index, first, then quantize the chart sizes, so that all small charts have the same size, and sort by size preserving the morton order. + //xaPrint("Sorting charts.\n"); + // Sort charts by area. + m_radix = RadixSort(); + m_radix.sort(chartOrderArray); + const uint32_t *ranks = m_radix.ranks(); + // First iteration - guess texelsPerUnit. + if (options.method != PackMethod::TexelArea && iteration == 0) { + // Estimate size of the map based on the mesh surface area and given texel scale. + const float texelCount = std::max(1.0f, meshArea * square(texelsPerUnit) / 0.75f); // Assume 75% utilization. + texelsPerUnit = sqrt((options.resolution * options.resolution) / texelCount); + resetUvs(); + continue; + } + // Init bit map. + m_bitmap.clearAll(); + m_bitmap.resize(options.resolution, options.resolution, false); + int w = 0; + int h = 0; + // Add sorted charts to bitmap. + for (uint32_t i = 0; i < chartCount; i++) { + uint32_t c = ranks[chartCount - i - 1]; // largest chart first + Chart *chart = m_atlas->chartAt(c); + if (!chart->isVertexMapped() && !chart->isDisk()) continue; + //float scale_x = 1; + //float scale_y = 1; + BitMap chart_bitmap; + if (chart->isVertexMapped()) { + chart->blockAligned = false; + // Init all bits to 1. + chart_bitmap.resize(ftoi_ceil(chartExtents[c].x), ftoi_ceil(chartExtents[c].y), /*initValue=*/true); + // @@ Another alternative would be to try to map each vertex to a different texel trying to fill all the available unused texels. + } else { + // @@ Add special cases for dot and line charts. @@ Lightmap rasterizer also needs to handle these special cases. + // @@ We could also have a special case for chart quads. If the quad surface <= 4 texels, align vertices with texel centers and do not add padding. May be very useful for foliage. + // @@ In general we could reduce the padding of all charts by one texel by using a rasterizer that takes into account the 2-texel footprint of the tent bilinear filter. For example, + // if we have a chart that is less than 1 texel wide currently we add one texel to the left and one texel to the right creating a 3-texel-wide bitmap. However, if we know that the + // chart is only 1 texel wide we could align it so that it only touches the footprint of two texels: + // | | <- Touches texels 0, 1 and 2. + // | | <- Only touches texels 0 and 1. + // \ \ / \ / / + // \ X X / + // \ / \ / \ / + // V V V + // 0 1 2 + if (options.conservative) { + // Init all bits to 0. + chart_bitmap.resize(ftoi_ceil(chartExtents[c].x) + 1 + options.padding, ftoi_ceil(chartExtents[c].y) + 1 + options.padding, /*initValue=*/false); // + 2 to add padding on both sides. + // Rasterize chart and dilate. + drawChartBitmapDilate(chart, &chart_bitmap, options.padding); + } else { + // Init all bits to 0. + chart_bitmap.resize(ftoi_ceil(chartExtents[c].x) + 1, ftoi_ceil(chartExtents[c].y) + 1, /*initValue=*/false); // Add half a texels on each side. + // Rasterize chart and dilate. + drawChartBitmap(chart, &chart_bitmap, Vector2(1), Vector2(0.5)); + } + } + int best_x, best_y; + int best_cw, best_ch; // Includes padding now. + int best_r; + findChartLocation(options.quality, &chart_bitmap, chartExtents[c], w, h, &best_x, &best_y, &best_cw, &best_ch, &best_r, chart->blockAligned); + /*if (w < best_x + best_cw || h < best_y + best_ch) + { + xaPrint("Resize extents to (%d, %d).\n", best_x + best_cw, best_y + best_ch); + }*/ + // Update parametric extents. + w = std::max(w, best_x + best_cw); + h = std::max(h, best_y + best_ch); + w = align(w, 4); + h = align(h, 4); + // Resize bitmap if necessary. + if (uint32_t(w) > m_bitmap.width() || uint32_t(h) > m_bitmap.height()) { + //xaPrint("Resize bitmap (%d, %d).\n", nextPowerOfTwo(w), nextPowerOfTwo(h)); + m_bitmap.resize(nextPowerOfTwo(uint32_t(w)), nextPowerOfTwo(uint32_t(h)), false); + } + //xaPrint("Add chart at (%d, %d).\n", best_x, best_y); + addChart(&chart_bitmap, w, h, best_x, best_y, best_r); + //float best_angle = 2 * PI * best_r; + // Translate and rotate chart texture coordinates. + halfedge::Mesh *mesh = chart->chartMesh(); + const uint32_t vertexCount = mesh->vertexCount(); + for (uint32_t v = 0; v < vertexCount; v++) { + halfedge::Vertex *vertex = mesh->vertexAt(v); + Vector2 t = vertex->tex; + if (best_r) std::swap(t.x, t.y); + //vertex->tex.x = best_x + t.x * cosf(best_angle) - t.y * sinf(best_angle); + //vertex->tex.y = best_y + t.x * sinf(best_angle) + t.y * cosf(best_angle); + vertex->tex.x = best_x + t.x + 0.5f; + vertex->tex.y = best_y + t.y + 0.5f; + xaAssert(vertex->tex.x >= 0 && vertex->tex.y >= 0); + xaAssert(std::isfinite(vertex->tex.x) && std::isfinite(vertex->tex.y)); + } + } + //w -= padding - 1; // Leave one pixel border! + //h -= padding - 1; + m_width = std::max(0, w); + m_height = std::max(0, h); + xaAssert(isAligned(m_width, 4)); + xaAssert(isAligned(m_height, 4)); + if (options.method == PackMethod::ExactResolution) { + texelsPerUnit *= sqrt((options.resolution * options.resolution) / (float)(m_width * m_height)); + if (iteration > 1 && m_width <= options.resolution && m_height <= options.resolution) { + m_width = m_height = options.resolution; + return; + } + resetUvs(); + } else { + return; + } + } + } + + float computeAtlasUtilization() const { + const uint32_t w = m_width; + const uint32_t h = m_height; + xaDebugAssert(w <= m_bitmap.width()); + xaDebugAssert(h <= m_bitmap.height()); + uint32_t count = 0; + for (uint32_t y = 0; y < h; y++) { + for (uint32_t x = 0; x < w; x++) { + count += m_bitmap.bitAt(x, y); + } + } + return float(count) / (w * h); + } + +private: + void resetUvs() { + for (uint32_t i = 0; i < m_atlas->chartCount(); i++) { + halfedge::Mesh *mesh = m_atlas->chartAt(i)->chartMesh(); + for (uint32_t j = 0; j < mesh->vertexCount(); j++) + mesh->vertexAt(j)->tex = m_originalChartUvs[i][j]; + } + } + + // IC: Brute force is slow, and random may take too much time to converge. We start inserting large charts in a small atlas. Using brute force is lame, because most of the space + // is occupied at this point. At the end we have many small charts and a large atlas with sparse holes. Finding those holes randomly is slow. A better approach would be to + // start stacking large charts as if they were tetris pieces. Once charts get small try to place them randomly. It may be interesting to try a intermediate strategy, first try + // along one axis and then try exhaustively along that axis. + void findChartLocation(int quality, const BitMap *bitmap, Vector2::Arg extents, int w, int h, int *best_x, int *best_y, int *best_w, int *best_h, int *best_r, bool blockAligned) { + int attempts = 256; + if (quality == 1) attempts = 4096; + if (quality == 2) attempts = 2048; + if (quality == 3) attempts = 1024; + if (quality == 4) attempts = 512; + if (quality == 0 || w * h < attempts) { + findChartLocation_bruteForce(bitmap, extents, w, h, best_x, best_y, best_w, best_h, best_r, blockAligned); + } else { + findChartLocation_random(bitmap, extents, w, h, best_x, best_y, best_w, best_h, best_r, attempts, blockAligned); + } + } + + void findChartLocation_bruteForce(const BitMap *bitmap, Vector2::Arg /*extents*/, int w, int h, int *best_x, int *best_y, int *best_w, int *best_h, int *best_r, bool blockAligned) { + const int BLOCK_SIZE = 4; + int best_metric = INT_MAX; + int step_size = blockAligned ? BLOCK_SIZE : 1; + // Try two different orientations. + for (int r = 0; r < 2; r++) { + int cw = bitmap->width(); + int ch = bitmap->height(); + if (r & 1) std::swap(cw, ch); + for (int y = 0; y <= h + 1; y += step_size) { // + 1 to extend atlas in case atlas full. + for (int x = 0; x <= w + 1; x += step_size) { // + 1 not really necessary here. + // Early out. + int area = std::max(w, x + cw) * std::max(h, y + ch); + //int perimeter = max(w, x+cw) + max(h, y+ch); + int extents = std::max(std::max(w, x + cw), std::max(h, y + ch)); + int metric = extents * extents + area; + if (metric > best_metric) { + continue; + } + if (metric == best_metric && std::max(x, y) >= std::max(*best_x, *best_y)) { + // If metric is the same, pick the one closest to the origin. + continue; + } + if (canAddChart(bitmap, w, h, x, y, r)) { + best_metric = metric; + *best_x = x; + *best_y = y; + *best_w = cw; + *best_h = ch; + *best_r = r; + if (area == w * h) { + // Chart is completely inside, do not look at any other location. + goto done; + } + } + } + } + } + done: + xaDebugAssert(best_metric != INT_MAX); + } + + void findChartLocation_random(const BitMap *bitmap, Vector2::Arg /*extents*/, int w, int h, int *best_x, int *best_y, int *best_w, int *best_h, int *best_r, int minTrialCount, bool blockAligned) { + const int BLOCK_SIZE = 4; + int best_metric = INT_MAX; + for (int i = 0; i < minTrialCount || best_metric == INT_MAX; i++) { + int r = m_rand.getRange(1); + int x = m_rand.getRange(w + 1); // + 1 to extend atlas in case atlas full. We may want to use a higher number to increase probability of extending atlas. + int y = m_rand.getRange(h + 1); // + 1 to extend atlas in case atlas full. + if (blockAligned) { + x = align(x, BLOCK_SIZE); + y = align(y, BLOCK_SIZE); + } + int cw = bitmap->width(); + int ch = bitmap->height(); + if (r & 1) std::swap(cw, ch); + // Early out. + int area = std::max(w, x + cw) * std::max(h, y + ch); + //int perimeter = max(w, x+cw) + max(h, y+ch); + int extents = std::max(std::max(w, x + cw), std::max(h, y + ch)); + int metric = extents * extents + area; + if (metric > best_metric) { + continue; + } + if (metric == best_metric && std::min(x, y) > std::min(*best_x, *best_y)) { + // If metric is the same, pick the one closest to the origin. + continue; + } + if (canAddChart(bitmap, w, h, x, y, r)) { + best_metric = metric; + *best_x = x; + *best_y = y; + *best_w = cw; + *best_h = ch; + *best_r = r; + if (area == w * h) { + // Chart is completely inside, do not look at any other location. + break; + } + } + } + } + + void drawChartBitmapDilate(const Chart *chart, BitMap *bitmap, int padding) { + const int w = bitmap->width(); + const int h = bitmap->height(); + const Vector2 extents = Vector2(float(w), float(h)); + // Rasterize chart faces, check that all bits are not set. + const uint32_t faceCount = chart->faceCount(); + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = chart->chartMesh()->faceAt(f); + Vector2 vertices[4]; + uint32_t edgeCount = 0; + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + if (edgeCount < 4) { + vertices[edgeCount] = it.vertex()->tex + Vector2(0.5) + Vector2(float(padding), float(padding)); + } + edgeCount++; + } + if (edgeCount == 3) { + raster::drawTriangle(raster::Mode_Antialiased, extents, true, vertices, AtlasPacker::setBitsCallback, bitmap); + } else { + raster::drawQuad(raster::Mode_Antialiased, extents, true, vertices, AtlasPacker::setBitsCallback, bitmap); + } + } + // Expand chart by padding pixels. (dilation) + BitMap tmp(w, h); + for (int i = 0; i < padding; i++) { + tmp.clearAll(); + for (int y = 0; y < h; y++) { + for (int x = 0; x < w; x++) { + bool b = bitmap->bitAt(x, y); + if (!b) { + if (x > 0) { + b |= bitmap->bitAt(x - 1, y); + if (y > 0) b |= bitmap->bitAt(x - 1, y - 1); + if (y < h - 1) b |= bitmap->bitAt(x - 1, y + 1); + } + if (y > 0) b |= bitmap->bitAt(x, y - 1); + if (y < h - 1) b |= bitmap->bitAt(x, y + 1); + if (x < w - 1) { + b |= bitmap->bitAt(x + 1, y); + if (y > 0) b |= bitmap->bitAt(x + 1, y - 1); + if (y < h - 1) b |= bitmap->bitAt(x + 1, y + 1); + } + } + if (b) tmp.setBitAt(x, y); + } + } + std::swap(tmp, *bitmap); + } + } + + void drawChartBitmap(const Chart *chart, BitMap *bitmap, const Vector2 &scale, const Vector2 &offset) { + const int w = bitmap->width(); + const int h = bitmap->height(); + const Vector2 extents = Vector2(float(w), float(h)); + static const Vector2 pad[4] = { + Vector2(-0.5, -0.5), + Vector2(0.5, -0.5), + Vector2(-0.5, 0.5), + Vector2(0.5, 0.5) + }; + // Rasterize 4 times to add proper padding. + for (int i = 0; i < 4; i++) { + // Rasterize chart faces, check that all bits are not set. + const uint32_t faceCount = chart->chartMesh()->faceCount(); + for (uint32_t f = 0; f < faceCount; f++) { + const halfedge::Face *face = chart->chartMesh()->faceAt(f); + Vector2 vertices[4]; + uint32_t edgeCount = 0; + for (halfedge::Face::ConstEdgeIterator it(face->edges()); !it.isDone(); it.advance()) { + if (edgeCount < 4) { + vertices[edgeCount] = it.vertex()->tex * scale + offset + pad[i]; + xaAssert(ftoi_ceil(vertices[edgeCount].x) >= 0); + xaAssert(ftoi_ceil(vertices[edgeCount].y) >= 0); + xaAssert(ftoi_ceil(vertices[edgeCount].x) <= w); + xaAssert(ftoi_ceil(vertices[edgeCount].y) <= h); + } + edgeCount++; + } + if (edgeCount == 3) { + raster::drawTriangle(raster::Mode_Antialiased, extents, /*enableScissors=*/true, vertices, AtlasPacker::setBitsCallback, bitmap); + } else { + raster::drawQuad(raster::Mode_Antialiased, extents, /*enableScissors=*/true, vertices, AtlasPacker::setBitsCallback, bitmap); + } + } + } + // Expand chart by padding pixels. (dilation) + BitMap tmp(w, h); + tmp.clearAll(); + for (int y = 0; y < h; y++) { + for (int x = 0; x < w; x++) { + bool b = bitmap->bitAt(x, y); + if (!b) { + if (x > 0) { + b |= bitmap->bitAt(x - 1, y); + if (y > 0) b |= bitmap->bitAt(x - 1, y - 1); + if (y < h - 1) b |= bitmap->bitAt(x - 1, y + 1); + } + if (y > 0) b |= bitmap->bitAt(x, y - 1); + if (y < h - 1) b |= bitmap->bitAt(x, y + 1); + if (x < w - 1) { + b |= bitmap->bitAt(x + 1, y); + if (y > 0) b |= bitmap->bitAt(x + 1, y - 1); + if (y < h - 1) b |= bitmap->bitAt(x + 1, y + 1); + } + } + if (b) tmp.setBitAt(x, y); + } + } + std::swap(tmp, *bitmap); + } + + bool canAddChart(const BitMap *bitmap, int atlas_w, int atlas_h, int offset_x, int offset_y, int r) { + xaDebugAssert(r == 0 || r == 1); + // Check whether the two bitmaps overlap. + const int w = bitmap->width(); + const int h = bitmap->height(); + if (r == 0) { + for (int y = 0; y < h; y++) { + int yy = y + offset_y; + if (yy >= 0) { + for (int x = 0; x < w; x++) { + int xx = x + offset_x; + if (xx >= 0) { + if (bitmap->bitAt(x, y)) { + if (xx < atlas_w && yy < atlas_h) { + if (m_bitmap.bitAt(xx, yy)) return false; + } + } + } + } + } + } + } else if (r == 1) { + for (int y = 0; y < h; y++) { + int xx = y + offset_x; + if (xx >= 0) { + for (int x = 0; x < w; x++) { + int yy = x + offset_y; + if (yy >= 0) { + if (bitmap->bitAt(x, y)) { + if (xx < atlas_w && yy < atlas_h) { + if (m_bitmap.bitAt(xx, yy)) return false; + } + } + } + } + } + } + } + return true; + } + + void addChart(const BitMap *bitmap, int atlas_w, int atlas_h, int offset_x, int offset_y, int r) { + xaDebugAssert(r == 0 || r == 1); + // Check whether the two bitmaps overlap. + const int w = bitmap->width(); + const int h = bitmap->height(); + if (r == 0) { + for (int y = 0; y < h; y++) { + int yy = y + offset_y; + if (yy >= 0) { + for (int x = 0; x < w; x++) { + int xx = x + offset_x; + if (xx >= 0) { + if (bitmap->bitAt(x, y)) { + if (xx < atlas_w && yy < atlas_h) { + xaDebugAssert(m_bitmap.bitAt(xx, yy) == false); + m_bitmap.setBitAt(xx, yy); + } + } + } + } + } + } + } else if (r == 1) { + for (int y = 0; y < h; y++) { + int xx = y + offset_x; + if (xx >= 0) { + for (int x = 0; x < w; x++) { + int yy = x + offset_y; + if (yy >= 0) { + if (bitmap->bitAt(x, y)) { + if (xx < atlas_w && yy < atlas_h) { + xaDebugAssert(m_bitmap.bitAt(xx, yy) == false); + m_bitmap.setBitAt(xx, yy); + } + } + } + } + } + } + } + } + + static bool setBitsCallback(void *param, int x, int y, Vector3::Arg, Vector3::Arg, Vector3::Arg, float area) { + BitMap *bitmap = (BitMap *)param; + if (area > 0.0) { + bitmap->setBitAt(x, y); + } + return true; + } + + // Compute the convex hull using Graham Scan. + static void convexHull(const std::vector<Vector2> &input, std::vector<Vector2> &output, float epsilon) { + const uint32_t inputCount = input.size(); + std::vector<float> coords(inputCount); + for (uint32_t i = 0; i < inputCount; i++) { + coords[i] = input[i].x; + } + RadixSort radix; + radix.sort(coords); + const uint32_t *ranks = radix.ranks(); + std::vector<Vector2> top; + top.reserve(inputCount); + std::vector<Vector2> bottom; + bottom.reserve(inputCount); + Vector2 P = input[ranks[0]]; + Vector2 Q = input[ranks[inputCount - 1]]; + float topy = std::max(P.y, Q.y); + float boty = std::min(P.y, Q.y); + for (uint32_t i = 0; i < inputCount; i++) { + Vector2 p = input[ranks[i]]; + if (p.y >= boty) top.push_back(p); + } + for (uint32_t i = 0; i < inputCount; i++) { + Vector2 p = input[ranks[inputCount - 1 - i]]; + if (p.y <= topy) bottom.push_back(p); + } + // Filter top list. + output.clear(); + output.push_back(top[0]); + output.push_back(top[1]); + for (uint32_t i = 2; i < top.size();) { + Vector2 a = output[output.size() - 2]; + Vector2 b = output[output.size() - 1]; + Vector2 c = top[i]; + float area = triangleArea(a, b, c); + if (area >= -epsilon) { + output.pop_back(); + } + if (area < -epsilon || output.size() == 1) { + output.push_back(c); + i++; + } + } + uint32_t top_count = output.size(); + output.push_back(bottom[1]); + // Filter bottom list. + for (uint32_t i = 2; i < bottom.size();) { + Vector2 a = output[output.size() - 2]; + Vector2 b = output[output.size() - 1]; + Vector2 c = bottom[i]; + float area = triangleArea(a, b, c); + if (area >= -epsilon) { + output.pop_back(); + } + if (area < -epsilon || output.size() == top_count) { + output.push_back(c); + i++; + } + } + // Remove duplicate element. + xaDebugAssert(output.front() == output.back()); + output.pop_back(); + } + + // This should compute convex hull and use rotating calipers to find the best box. Currently it uses a brute force method. + static void computeBoundingBox(Chart *chart, Vector2 *majorAxis, Vector2 *minorAxis, Vector2 *minCorner, Vector2 *maxCorner) { + // Compute list of boundary points. + std::vector<Vector2> points; + points.reserve(16); + halfedge::Mesh *mesh = chart->chartMesh(); + const uint32_t vertexCount = mesh->vertexCount(); + for (uint32_t i = 0; i < vertexCount; i++) { + halfedge::Vertex *vertex = mesh->vertexAt(i); + if (vertex->isBoundary()) { + points.push_back(vertex->tex); + } + } + xaDebugAssert(points.size() > 0); + std::vector<Vector2> hull; + convexHull(points, hull, 0.00001f); + // @@ Ideally I should use rotating calipers to find the best box. Using brute force for now. + float best_area = FLT_MAX; + Vector2 best_min; + Vector2 best_max; + Vector2 best_axis; + const uint32_t hullCount = hull.size(); + for (uint32_t i = 0, j = hullCount - 1; i < hullCount; j = i, i++) { + if (equal(hull[i], hull[j])) { + continue; + } + Vector2 axis = normalize(hull[i] - hull[j], 0.0f); + xaDebugAssert(isFinite(axis)); + // Compute bounding box. + Vector2 box_min(FLT_MAX, FLT_MAX); + Vector2 box_max(-FLT_MAX, -FLT_MAX); + for (uint32_t v = 0; v < hullCount; v++) { + Vector2 point = hull[v]; + float x = dot(axis, point); + if (x < box_min.x) box_min.x = x; + if (x > box_max.x) box_max.x = x; + float y = dot(Vector2(-axis.y, axis.x), point); + if (y < box_min.y) box_min.y = y; + if (y > box_max.y) box_max.y = y; + } + // Compute box area. + float area = (box_max.x - box_min.x) * (box_max.y - box_min.y); + if (area < best_area) { + best_area = area; + best_min = box_min; + best_max = box_max; + best_axis = axis; + } + } + // Consider all points, not only boundary points, in case the input chart is malformed. + for (uint32_t i = 0; i < vertexCount; i++) { + halfedge::Vertex *vertex = mesh->vertexAt(i); + Vector2 point = vertex->tex; + float x = dot(best_axis, point); + if (x < best_min.x) best_min.x = x; + if (x > best_max.x) best_max.x = x; + float y = dot(Vector2(-best_axis.y, best_axis.x), point); + if (y < best_min.y) best_min.y = y; + if (y > best_max.y) best_max.y = y; + } + *majorAxis = best_axis; + *minorAxis = Vector2(-best_axis.y, best_axis.x); + *minCorner = best_min; + *maxCorner = best_max; + } + + Atlas *m_atlas; + BitMap m_bitmap; + RadixSort m_radix; + uint32_t m_width; + uint32_t m_height; + MTRand m_rand; + std::vector<std::vector<Vector2> > m_originalChartUvs; +}; + +} // namespace param +} // namespace internal + +struct Atlas { + internal::param::Atlas atlas; + std::vector<internal::halfedge::Mesh *> heMeshes; + uint32_t width = 0; + uint32_t height = 0; + OutputMesh **outputMeshes = NULL; +}; + +void SetPrint(PrintFunc print) { + internal::s_print = print; +} + +Atlas *Create() { + Atlas *atlas = new Atlas(); + return atlas; +} + +void Destroy(Atlas *atlas) { + xaAssert(atlas); + for (int i = 0; i < (int)atlas->heMeshes.size(); i++) { + delete atlas->heMeshes[i]; + if (atlas->outputMeshes) { + OutputMesh *outputMesh = atlas->outputMeshes[i]; + for (uint32_t j = 0; j < outputMesh->chartCount; j++) + delete[] outputMesh->chartArray[j].indexArray; + delete[] outputMesh->chartArray; + delete[] outputMesh->vertexArray; + delete[] outputMesh->indexArray; + delete outputMesh; + } + } + delete[] atlas->outputMeshes; + delete atlas; +} + +static internal::Vector3 DecodePosition(const InputMesh &mesh, uint32_t index) { + xaAssert(mesh.vertexPositionData); + return *((const internal::Vector3 *)&((const uint8_t *)mesh.vertexPositionData)[mesh.vertexPositionStride * index]); +} + +static internal::Vector3 DecodeNormal(const InputMesh &mesh, uint32_t index) { + xaAssert(mesh.vertexNormalData); + return *((const internal::Vector3 *)&((const uint8_t *)mesh.vertexNormalData)[mesh.vertexNormalStride * index]); +} + +static internal::Vector2 DecodeUv(const InputMesh &mesh, uint32_t index) { + xaAssert(mesh.vertexUvData); + return *((const internal::Vector2 *)&((const uint8_t *)mesh.vertexUvData)[mesh.vertexUvStride * index]); +} + +static uint32_t DecodeIndex(IndexFormat::Enum format, const void *indexData, uint32_t i) { + if (format == IndexFormat::HalfFloat) + return (uint32_t)((const uint16_t *)indexData)[i]; + return ((const uint32_t *)indexData)[i]; +} + +static float EdgeLength(internal::Vector3 pos1, internal::Vector3 pos2) { + return internal::length(pos2 - pos1); +} + +AddMeshError AddMesh(Atlas *atlas, const InputMesh &mesh, bool useColocalVertices) { + xaAssert(atlas); + AddMeshError error; + error.code = AddMeshErrorCode::Success; + error.face = error.index0 = error.index1 = UINT32_MAX; + // Expecting triangle faces. + if ((mesh.indexCount % 3) != 0) { + error.code = AddMeshErrorCode::InvalidIndexCount; + return error; + } + // Check if any index is out of range. + for (uint32_t j = 0; j < mesh.indexCount; j++) { + const uint32_t index = DecodeIndex(mesh.indexFormat, mesh.indexData, j); + if (index < 0 || index >= mesh.vertexCount) { + error.code = AddMeshErrorCode::IndexOutOfRange; + error.index0 = index; + return error; + } + } + // Build half edge mesh. + internal::halfedge::Mesh *heMesh = new internal::halfedge::Mesh; + std::vector<uint32_t> canonicalMap; + canonicalMap.reserve(mesh.vertexCount); + for (uint32_t i = 0; i < mesh.vertexCount; i++) { + internal::halfedge::Vertex *vertex = heMesh->addVertex(DecodePosition(mesh, i)); + if (mesh.vertexNormalData) + vertex->nor = DecodeNormal(mesh, i); + if (mesh.vertexUvData) + vertex->tex = DecodeUv(mesh, i); + // Link colocals. You probably want to do this more efficiently! Sort by one axis or use a hash or grid. + uint32_t firstColocal = i; + if (useColocalVertices) { + for (uint32_t j = 0; j < i; j++) { + if (vertex->pos != DecodePosition(mesh, j)) + continue; +#if 0 + if (mesh.vertexNormalData && vertex->nor != DecodeNormal(mesh, j)) + continue; +#endif + if (mesh.vertexUvData && vertex->tex != DecodeUv(mesh, j)) + continue; + firstColocal = j; + break; + } + } + canonicalMap.push_back(firstColocal); + } + heMesh->linkColocalsWithCanonicalMap(canonicalMap); + for (uint32_t i = 0; i < mesh.indexCount / 3; i++) { + uint32_t tri[3]; + for (int j = 0; j < 3; j++) + tri[j] = DecodeIndex(mesh.indexFormat, mesh.indexData, i * 3 + j); + // Check for zero length edges. + for (int j = 0; j < 3; j++) { + const uint32_t edges[6] = { 0, 1, 1, 2, 2, 0 }; + const uint32_t index1 = tri[edges[j * 2 + 0]]; + const uint32_t index2 = tri[edges[j * 2 + 1]]; + const internal::Vector3 pos1 = DecodePosition(mesh, index1); + const internal::Vector3 pos2 = DecodePosition(mesh, index2); + if (EdgeLength(pos1, pos2) <= 0.0f) { + delete heMesh; + error.code = AddMeshErrorCode::ZeroLengthEdge; + error.face = i; + error.index0 = index1; + error.index1 = index2; + return error; + } + } + // Check for zero area faces. + { + const internal::Vector3 a = DecodePosition(mesh, tri[0]); + const internal::Vector3 b = DecodePosition(mesh, tri[1]); + const internal::Vector3 c = DecodePosition(mesh, tri[2]); + const float area = internal::length(internal::cross(b - a, c - a)) * 0.5f; + if (area <= 0.0f) { + delete heMesh; + error.code = AddMeshErrorCode::ZeroAreaFace; + error.face = i; + return error; + } + } + internal::halfedge::Face *face = heMesh->addFace(tri[0], tri[1], tri[2]); + + if (!face && heMesh->errorCode == internal::halfedge::Mesh::ErrorCode::AlreadyAddedEdge) { + //there is still hope for this, no reason to not add, at least add as separate + face = heMesh->addUniqueFace(tri[0], tri[1], tri[2]); + } + + if (!face) { + //continue; + + if (heMesh->errorCode == internal::halfedge::Mesh::ErrorCode::AlreadyAddedEdge) { + error.code = AddMeshErrorCode::AlreadyAddedEdge; + } else if (heMesh->errorCode == internal::halfedge::Mesh::ErrorCode::DegenerateColocalEdge) { + error.code = AddMeshErrorCode::DegenerateColocalEdge; + } else if (heMesh->errorCode == internal::halfedge::Mesh::ErrorCode::DegenerateEdge) { + error.code = AddMeshErrorCode::DegenerateEdge; + } else if (heMesh->errorCode == internal::halfedge::Mesh::ErrorCode::DuplicateEdge) { + error.code = AddMeshErrorCode::DuplicateEdge; + } + error.face = i; + error.index0 = heMesh->errorIndex0; + error.index1 = heMesh->errorIndex1; + delete heMesh; + return error; + } + if (mesh.faceMaterialData) + face->material = mesh.faceMaterialData[i]; + } + heMesh->linkBoundary(); + atlas->heMeshes.push_back(heMesh); + return error; +} + +void Generate(Atlas *atlas, CharterOptions charterOptions, PackerOptions packerOptions) { + xaAssert(atlas); + xaAssert(packerOptions.texelArea > 0); + // Chart meshes. + for (int i = 0; i < (int)atlas->heMeshes.size(); i++) { + std::vector<uint32_t> uncharted_materials; + atlas->atlas.computeCharts(atlas->heMeshes[i], charterOptions, uncharted_materials); + } + atlas->atlas.parameterizeCharts(); + internal::param::AtlasPacker packer(&atlas->atlas); + packer.packCharts(packerOptions); + //float utilization = return packer.computeAtlasUtilization(); + atlas->width = packer.getWidth(); + atlas->height = packer.getHeight(); + // Build output meshes. + atlas->outputMeshes = new OutputMesh *[atlas->heMeshes.size()]; + for (int i = 0; i < (int)atlas->heMeshes.size(); i++) { + const internal::halfedge::Mesh *heMesh = atlas->heMeshes[i]; + OutputMesh *outputMesh = atlas->outputMeshes[i] = new OutputMesh; + const internal::param::MeshCharts *charts = atlas->atlas.meshAt(i); + // Vertices. + outputMesh->vertexCount = charts->vertexCount(); + outputMesh->vertexArray = new OutputVertex[outputMesh->vertexCount]; + for (uint32_t i = 0; i < charts->chartCount(); i++) { + const internal::param::Chart *chart = charts->chartAt(i); + const uint32_t vertexOffset = charts->vertexCountBeforeChartAt(i); + for (uint32_t v = 0; v < chart->vertexCount(); v++) { + OutputVertex &output_vertex = outputMesh->vertexArray[vertexOffset + v]; + output_vertex.xref = chart->mapChartVertexToOriginalVertex(v); + internal::Vector2 uv = chart->chartMesh()->vertexAt(v)->tex; + output_vertex.uv[0] = uv.x; + output_vertex.uv[1] = uv.y; + } + } + // Indices. + outputMesh->indexCount = heMesh->faceCount() * 3; + outputMesh->indexArray = new uint32_t[outputMesh->indexCount]; + for (uint32_t f = 0; f < heMesh->faceCount(); f++) { + const uint32_t c = charts->faceChartAt(f); + const uint32_t i = charts->faceIndexWithinChartAt(f); + const uint32_t vertexOffset = charts->vertexCountBeforeChartAt(c); + const internal::param::Chart *chart = charts->chartAt(c); + xaDebugAssert(i < chart->chartMesh()->faceCount()); + xaDebugAssert(chart->faceAt(i) == f); + const internal::halfedge::Face *face = chart->chartMesh()->faceAt(i); + const internal::halfedge::Edge *edge = face->edge; + outputMesh->indexArray[3 * f + 0] = vertexOffset + edge->vertex->id; + outputMesh->indexArray[3 * f + 1] = vertexOffset + edge->next->vertex->id; + outputMesh->indexArray[3 * f + 2] = vertexOffset + edge->next->next->vertex->id; + } + // Charts. + outputMesh->chartCount = charts->chartCount(); + outputMesh->chartArray = new OutputChart[outputMesh->chartCount]; + for (uint32_t i = 0; i < charts->chartCount(); i++) { + OutputChart *outputChart = &outputMesh->chartArray[i]; + const internal::param::Chart *chart = charts->chartAt(i); + const uint32_t vertexOffset = charts->vertexCountBeforeChartAt(i); + const internal::halfedge::Mesh *mesh = chart->chartMesh(); + outputChart->indexCount = mesh->faceCount() * 3; + outputChart->indexArray = new uint32_t[outputChart->indexCount]; + for (uint32_t j = 0; j < mesh->faceCount(); j++) { + const internal::halfedge::Face *face = mesh->faceAt(j); + const internal::halfedge::Edge *edge = face->edge; + outputChart->indexArray[3 * j + 0] = vertexOffset + edge->vertex->id; + outputChart->indexArray[3 * j + 1] = vertexOffset + edge->next->vertex->id; + outputChart->indexArray[3 * j + 2] = vertexOffset + edge->next->next->vertex->id; + } + } + } +} + +uint32_t GetWidth(const Atlas *atlas) { + xaAssert(atlas); + return atlas->width; +} + +uint32_t GetHeight(const Atlas *atlas) { + xaAssert(atlas); + return atlas->height; +} + +uint32_t GetNumCharts(const Atlas *atlas) { + xaAssert(atlas); + return atlas->atlas.chartCount(); +} + +const OutputMesh *const *GetOutputMeshes(const Atlas *atlas) { + xaAssert(atlas); + return atlas->outputMeshes; +} + +const char *StringForEnum(AddMeshErrorCode::Enum error) { + if (error == AddMeshErrorCode::AlreadyAddedEdge) + return "already added edge"; + if (error == AddMeshErrorCode::DegenerateColocalEdge) + return "degenerate colocal edge"; + if (error == AddMeshErrorCode::DegenerateEdge) + return "degenerate edge"; + if (error == AddMeshErrorCode::DuplicateEdge) + return "duplicate edge"; + if (error == AddMeshErrorCode::IndexOutOfRange) + return "index out of range"; + if (error == AddMeshErrorCode::InvalidIndexCount) + return "invalid index count"; + if (error == AddMeshErrorCode::ZeroAreaFace) + return "zero area face"; + if (error == AddMeshErrorCode::ZeroLengthEdge) + return "zero length edge"; + return "success"; +} + +} // namespace xatlas diff --git a/thirdparty/xatlas/xatlas.h b/thirdparty/xatlas/xatlas.h new file mode 100644 index 0000000000..4140429fee --- /dev/null +++ b/thirdparty/xatlas/xatlas.h @@ -0,0 +1,160 @@ +// This code is in the public domain -- castanyo@yahoo.es +#pragma once +#ifndef XATLAS_H +#define XATLAS_H +#include <float.h> // FLT_MAX +#include <limits.h> +#include <stdint.h> +namespace xatlas { + +typedef void (*PrintFunc)(const char *, ...); + +struct Atlas; + +struct CharterOptions { + float proxyFitMetricWeight; + float roundnessMetricWeight; + float straightnessMetricWeight; + float normalSeamMetricWeight; + float textureSeamMetricWeight; + float maxChartArea; + float maxBoundaryLength; + + CharterOptions() { + // These are the default values we use on The Witness. + proxyFitMetricWeight = 2.0f; + roundnessMetricWeight = 0.01f; + straightnessMetricWeight = 6.0f; + normalSeamMetricWeight = 4.0f; + textureSeamMetricWeight = 0.5f; + /* + proxyFitMetricWeight = 1.0f; + roundnessMetricWeight = 0.1f; + straightnessMetricWeight = 0.25f; + normalSeamMetricWeight = 1.0f; + textureSeamMetricWeight = 0.1f; + */ + maxChartArea = FLT_MAX; + maxBoundaryLength = FLT_MAX; + } +}; + +struct PackMethod { + enum Enum { + TexelArea, // texel_area determines resolution + ApproximateResolution, // guess texel_area to approximately match desired resolution + ExactResolution // run the packer multiple times to exactly match the desired resolution (slow) + }; +}; + +struct PackerOptions { + PackMethod::Enum method; + + // 0 - brute force + // 1 - 4096 attempts + // 2 - 2048 + // 3 - 1024 + // 4 - 512 + // other - 256 + // Avoid brute force packing, since it can be unusably slow in some situations. + int quality; + + float texelArea; // This is not really texel area, but 1 / texel width? + uint32_t resolution; + bool blockAlign; // Align charts to 4x4 blocks. + bool conservative; // Pack charts with extra padding. + int padding; + + PackerOptions() { + method = PackMethod::ApproximateResolution; + quality = 1; + texelArea = 8; + resolution = 512; + blockAlign = false; + conservative = false; + padding = 0; + } +}; + +struct AddMeshErrorCode { + enum Enum { + Success, + AlreadyAddedEdge, // index0 and index1 are the edge indices + DegenerateColocalEdge, // index0 and index1 are the edge indices + DegenerateEdge, // index0 and index1 are the edge indices + DuplicateEdge, // index0 and index1 are the edge indices + IndexOutOfRange, // index0 is the index + InvalidIndexCount, // not evenly divisible by 3 - expecting triangles + ZeroAreaFace, + ZeroLengthEdge // index0 and index1 are the edge indices + }; +}; + +struct AddMeshError { + AddMeshErrorCode::Enum code; + uint32_t face; + uint32_t index0, index1; +}; + +struct IndexFormat { + enum Enum { + HalfFloat, + Float + }; +}; + +struct InputMesh { + uint32_t vertexCount; + const void *vertexPositionData; + uint32_t vertexPositionStride; + const void *vertexNormalData; // optional + uint32_t vertexNormalStride; // optional + + // optional + // The input UVs are provided as a hint to the chart generator. + const void *vertexUvData; + uint32_t vertexUvStride; + + uint32_t indexCount; + const void *indexData; + IndexFormat::Enum indexFormat; + + // optional. indexCount / 3 in length. + // Charter also uses material boundaries as a hint to cut charts. + const uint16_t *faceMaterialData; +}; + +struct OutputChart { + uint32_t *indexArray; + uint32_t indexCount; +}; + +struct OutputVertex { + float uv[2]; + uint32_t xref; // Index of input vertex from which this output vertex originated. +}; + +struct OutputMesh { + OutputChart *chartArray; + uint32_t chartCount; + uint32_t *indexArray; + uint32_t indexCount; + OutputVertex *vertexArray; + uint32_t vertexCount; +}; + +void SetPrint(PrintFunc print); +Atlas *Create(); +void Destroy(Atlas *atlas); +// useColocalVertices - generates fewer charts (good), but is more sensitive to bad geometry. +AddMeshError AddMesh(Atlas *atlas, const InputMesh &mesh, bool useColocalVertices = true); +void Generate(Atlas *atlas, CharterOptions charterOptions = CharterOptions(), PackerOptions packerOptions = PackerOptions()); +uint32_t GetWidth(const Atlas *atlas); +uint32_t GetHeight(const Atlas *atlas); +uint32_t GetNumCharts(const Atlas *atlas); +const OutputMesh *const *GetOutputMeshes(const Atlas *atlas); +const char *StringForEnum(AddMeshErrorCode::Enum error); + +} // namespace xatlas + +#endif // XATLAS_H diff --git a/thirdparty/zstd/SCsub b/thirdparty/zstd/SCsub deleted file mode 100644 index 899a18e1cf..0000000000 --- a/thirdparty/zstd/SCsub +++ /dev/null @@ -1,29 +0,0 @@ -#!/usr/bin/env python - -Import('env') - -thirdparty_zstd_dir = "#thirdparty/zstd/" -thirdparty_zstd_sources = [ - "common/entropy_common.c", - "common/error_private.c", - "common/fse_decompress.c", - "common/pool.c", - "common/threading.c", - "common/xxhash.c", - "common/zstd_common.c", - "compress/fse_compress.c", - "compress/huf_compress.c", - "compress/zstd_compress.c", - "compress/zstd_double_fast.c", - "compress/zstd_fast.c", - "compress/zstd_lazy.c", - "compress/zstd_ldm.c", - "compress/zstdmt_compress.c", - "compress/zstd_opt.c", - "decompress/huf_decompress.c", - "decompress/zstd_decompress.c", -] -thirdparty_zstd_sources = [thirdparty_zstd_dir + file for file in thirdparty_zstd_sources] -env.add_source_files(env.core_sources, thirdparty_zstd_sources) -env.Append(CPPPATH=["#thirdparty/zstd", "#thirdparty/zstd/common"]) -env.Append(CCFLAGS="-DZSTD_STATIC_LINKING_ONLY") |