Assorted fixes to UV unwrapping and GPU lightmapper

Various fixes to UV2 unwrapping and the GPU lightmapper. Listed here for
context in case of git blame/bisect:

* Fix UV2 unwrapping on import, also cleaned up the unwrap cache code.
* Fix saving of RGBA images in EXR format.
* Fixes to the GPU lightmapper:
	- Added padding between atlas elements, avoids bleeding.
	- Remove old SDF generation code.
	- Fix baked attenuation for Omni/Spot lights.
	- Fix baking of material properties onto UV2 (wireframe was
	  wrongly used before).
	- Disable statically baked lights for objects that have a
	  lightmap texture to avoid applying the same light twice.
	- Fix lightmap pairing in RendererSceneCull.
	- Fix UV2 array generated from `RenderingServer::mesh_surface_get_arrays()`.
	- Port autoexposure fix for OIDN from 3.x.
	- Save debug textures as EXR when using floating point format.

6 files changed, 154 insertions, 174 deletions

diff --git a/modules/lightmapper_rd/lightmapper_rd.cpp b/modules/lightmapper_rd/lightmapper_rd.cpp
index 9394e5c47e..3b0fbb1c47 100644
--- a/modules/lightmapper_rd/lightmapper_rd.cpp
+++ b/modules/lightmapper_rd/lightmapper_rd.cpp
@@ -162,8 +162,8 @@ Lightmapper::BakeError LightmapperRD::_blit_meshes_into_atlas(int p_max_texture_
 		MeshInstance &mi = mesh_instances.write[m_i];
 		Size2i s = Size2i(mi.data.albedo_on_uv2->get_width(), mi.data.albedo_on_uv2->get_height());
 		sizes.push_back(s);
-		atlas_size.width = MAX(atlas_size.width, s.width);
-		atlas_size.height = MAX(atlas_size.height, s.height);
+		atlas_size.width = MAX(atlas_size.width, s.width + 2);
+		atlas_size.height = MAX(atlas_size.height, s.height + 2);
 	}
 
 	int max = nearest_power_of_2_templated(atlas_size.width);
@@ -186,10 +186,12 @@ Lightmapper::BakeError LightmapperRD::_blit_meshes_into_atlas(int p_max_texture_
 
 	//determine best texture array atlas size by bruteforce fitting
 	while (atlas_size.x <= p_max_texture_size && atlas_size.y <= p_max_texture_size) {
-		Vector<Vector2i> source_sizes = sizes;
+		Vector<Vector2i> source_sizes;
 		Vector<int> source_indices;
-		source_indices.resize(source_sizes.size());
+		source_sizes.resize(sizes.size());
+		source_indices.resize(sizes.size());
 		for (int i = 0; i < source_indices.size(); i++) {
+			source_sizes.write[i] = sizes[i] + Vector2i(2, 2); // Add padding between lightmaps
 			source_indices.write[i] = i;
 		}
 		Vector<Vector3i> atlas_offsets;
@@ -207,7 +209,7 @@ Lightmapper::BakeError LightmapperRD::_blit_meshes_into_atlas(int p_max_texture_
 				if (ofs.z > 0) {
 					//valid
 					ofs.z = slices;
-					atlas_offsets.write[sidx] = ofs;
+					atlas_offsets.write[sidx] = ofs + Vector3i(1, 1, 0); // Center lightmap in the reserved oversized region
 				} else {
 					new_indices.push_back(sidx);
 					new_sources.push_back(source_sizes[i]);
@@ -272,7 +274,7 @@ Lightmapper::BakeError LightmapperRD::_blit_meshes_into_atlas(int p_max_texture_
 	return BAKE_OK;
 }
 
-void LightmapperRD::_create_acceleration_structures(RenderingDevice *rd, Size2i atlas_size, int atlas_slices, AABB &bounds, int grid_size, Vector<Probe> &probe_positions, GenerateProbes p_generate_probes, Vector<int> &slice_triangle_count, Vector<int> &slice_seam_count, RID &vertex_buffer, RID &triangle_buffer, RID &box_buffer, RID &lights_buffer, RID &triangle_cell_indices_buffer, RID &probe_positions_buffer, RID &grid_texture, RID &grid_texture_sdf, RID &seams_buffer, BakeStepFunc p_step_function, void *p_bake_userdata) {
+void LightmapperRD::_create_acceleration_structures(RenderingDevice *rd, Size2i atlas_size, int atlas_slices, AABB &bounds, int grid_size, Vector<Probe> &probe_positions, GenerateProbes p_generate_probes, Vector<int> &slice_triangle_count, Vector<int> &slice_seam_count, RID &vertex_buffer, RID &triangle_buffer, RID &box_buffer, RID &lights_buffer, RID &triangle_cell_indices_buffer, RID &probe_positions_buffer, RID &grid_texture, RID &seams_buffer, BakeStepFunc p_step_function, void *p_bake_userdata) {
 	HashMap<Vertex, uint32_t, VertexHash> vertex_map;
 
 	//fill triangles array and vertex array
@@ -482,14 +484,6 @@ void LightmapperRD::_create_acceleration_structures(RenderingDevice *rd, Size2i
 		img->save_png("res://grid_layer_" + itos(1000 + i).substr(1, 3) + ".png");
 	}
 #endif
-	if (p_step_function) {
-		p_step_function(0.45, TTR("Generating Signed Distance Field"), p_bake_userdata, true);
-	}
-
-	//generate SDF for raytracing
-	Vector<uint32_t> euclidean_pos = Geometry3D::generate_edf(solid, Vector3i(grid_size, grid_size, grid_size), false);
-	Vector<uint32_t> euclidean_neg = Geometry3D::generate_edf(solid, Vector3i(grid_size, grid_size, grid_size), true);
-	Vector<int8_t> sdf8 = Geometry3D::generate_sdf8(euclidean_pos, euclidean_neg);
 
 	/*****************************/
 	/*** CREATE GPU STRUCTURES ***/
@@ -551,10 +545,6 @@ void LightmapperRD::_create_acceleration_structures(RenderingDevice *rd, Size2i
 		tf.format = RD::DATA_FORMAT_R32G32_UINT;
 		texdata.write[0] = grid_indices.to_byte_array();
 		grid_texture = rd->texture_create(tf, RD::TextureView(), texdata);
-		//sdf
-		tf.format = RD::DATA_FORMAT_R8_SNORM;
-		texdata.write[0] = sdf8.to_byte_array();
-		grid_texture_sdf = rd->texture_create(tf, RD::TextureView(), texdata);
 	}
 }
 
@@ -755,8 +745,7 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 			light_environment_tex = rd->texture_create(tfp, RD::TextureView(), tdata);
 
 #ifdef DEBUG_TEXTURES
-			panorama_tex->convert(Image::FORMAT_RGB8);
-			panorama_tex->save_png("res://0_panorama.png");
+			panorama_tex->save_exr("res://0_panorama.exr", false);
 #endif
 		}
 	}
@@ -770,7 +759,6 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 	RID lights_buffer;
 	RID triangle_cell_indices_buffer;
 	RID grid_texture;
-	RID grid_texture_sdf;
 	RID seams_buffer;
 	RID probe_positions_buffer;
 
@@ -783,11 +771,10 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 	rd->free(lights_buffer);                \
 	rd->free(triangle_cell_indices_buffer); \
 	rd->free(grid_texture);                 \
-	rd->free(grid_texture_sdf);             \
 	rd->free(seams_buffer);                 \
 	rd->free(probe_positions_buffer);
 
-	_create_acceleration_structures(rd, atlas_size, atlas_slices, bounds, grid_size, probe_positions, p_generate_probes, slice_triangle_count, slice_seam_count, vertex_buffer, triangle_buffer, box_buffer, lights_buffer, triangle_cell_indices_buffer, probe_positions_buffer, grid_texture, grid_texture_sdf, seams_buffer, p_step_function, p_bake_userdata);
+	_create_acceleration_structures(rd, atlas_size, atlas_slices, bounds, grid_size, probe_positions, p_generate_probes, slice_triangle_count, slice_seam_count, vertex_buffer, triangle_buffer, box_buffer, lights_buffer, triangle_cell_indices_buffer, probe_positions_buffer, grid_texture, seams_buffer, p_step_function, p_bake_userdata);
 
 	if (p_step_function) {
 		p_step_function(0.47, TTR("Preparing shaders"), p_bake_userdata, true);
@@ -883,27 +870,20 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 			RD::Uniform u;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 			u.binding = 9;
-			u.ids.push_back(grid_texture_sdf);
-			base_uniforms.push_back(u);
-		}
-		{
-			RD::Uniform u;
-			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
-			u.binding = 10;
 			u.ids.push_back(albedo_array_tex);
 			base_uniforms.push_back(u);
 		}
 		{
 			RD::Uniform u;
 			u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
-			u.binding = 11;
+			u.binding = 10;
 			u.ids.push_back(emission_array_tex);
 			base_uniforms.push_back(u);
 		}
 		{
 			RD::Uniform u;
 			u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
-			u.binding = 12;
+			u.binding = 11;
 			u.ids.push_back(sampler);
 			base_uniforms.push_back(u);
 		}
@@ -937,13 +917,11 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 		Ref<Image> img;
 		img.instance();
 		img->create(atlas_size.width, atlas_size.height, false, Image::FORMAT_RGBAF, s);
-		img->convert(Image::FORMAT_RGBA8);
-		img->save_png("res://1_position_" + itos(i) + ".png");
+		img->save_exr("res://1_position_" + itos(i) + ".exr", false);
 
 		s = rd->texture_get_data(normal_tex, i);
 		img->create(atlas_size.width, atlas_size.height, false, Image::FORMAT_RGBAH, s);
-		img->convert(Image::FORMAT_RGBA8);
-		img->save_png("res://1_normal_" + itos(i) + ".png");
+		img->save_exr("res://1_normal_" + itos(i) + ".exr", false);
 	}
 #endif
 
@@ -966,27 +944,27 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 	}
 	ERR_FAIL_COND_V(err != OK, BAKE_ERROR_LIGHTMAP_CANT_PRE_BAKE_MESHES);
 
-	//unoccluder
+	// Unoccluder
 	RID compute_shader_unocclude = rd->shader_create_from_bytecode(compute_shader->get_bytecode("unocclude"));
 	ERR_FAIL_COND_V(compute_shader_unocclude.is_null(), BAKE_ERROR_LIGHTMAP_CANT_PRE_BAKE_MESHES); // internal check, should not happen
 	RID compute_shader_unocclude_pipeline = rd->compute_pipeline_create(compute_shader_unocclude);
 
-	//direct light
+	// Direct light
 	RID compute_shader_primary = rd->shader_create_from_bytecode(compute_shader->get_bytecode("primary"));
 	ERR_FAIL_COND_V(compute_shader_primary.is_null(), BAKE_ERROR_LIGHTMAP_CANT_PRE_BAKE_MESHES); // internal check, should not happen
 	RID compute_shader_primary_pipeline = rd->compute_pipeline_create(compute_shader_primary);
 
-	//indirect light
+	// Indirect light
 	RID compute_shader_secondary = rd->shader_create_from_bytecode(compute_shader->get_bytecode("secondary"));
 	ERR_FAIL_COND_V(compute_shader_secondary.is_null(), BAKE_ERROR_LIGHTMAP_CANT_PRE_BAKE_MESHES); //internal check, should not happen
 	RID compute_shader_secondary_pipeline = rd->compute_pipeline_create(compute_shader_secondary);
 
-	//dilate
+	// Dilate
 	RID compute_shader_dilate = rd->shader_create_from_bytecode(compute_shader->get_bytecode("dilate"));
 	ERR_FAIL_COND_V(compute_shader_dilate.is_null(), BAKE_ERROR_LIGHTMAP_CANT_PRE_BAKE_MESHES); //internal check, should not happen
 	RID compute_shader_dilate_pipeline = rd->compute_pipeline_create(compute_shader_dilate);
 
-	//dilate
+	// Light probes
 	RID compute_shader_light_probes = rd->shader_create_from_bytecode(compute_shader->get_bytecode("light_probes"));
 	ERR_FAIL_COND_V(compute_shader_light_probes.is_null(), BAKE_ERROR_LIGHTMAP_CANT_PRE_BAKE_MESHES); //internal check, should not happen
 	RID compute_shader_light_probes_pipeline = rd->compute_pipeline_create(compute_shader_light_probes);
@@ -1153,8 +1131,7 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 		Ref<Image> img;
 		img.instance();
 		img->create(atlas_size.width, atlas_size.height, false, Image::FORMAT_RGBAH, s);
-		img->convert(Image::FORMAT_RGBA8);
-		img->save_png("res://2_light_primary_" + itos(i) + ".png");
+		img->save_exr("res://2_light_primary_" + itos(i) + ".exr", false);
 	}
 #endif
 
@@ -1212,7 +1189,7 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 				RD::Uniform u;
 				u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
 				u.binding = 6;
-				u.ids.push_back(light_environment_tex); //reuse unocclude tex
+				u.ids.push_back(light_environment_tex);
 				uniforms.push_back(u);
 			}
 		}
@@ -1298,7 +1275,15 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 					}
 				}
 			}
+
+			if (b == 0) {
+				// This disables the environment for subsequent bounces
+				push_constant.environment_xform[3] = -99.0f;
+			}
 		}
+
+		// Restore the correct environment transform
+		push_constant.environment_xform[3] = 0.0f;
 	}
 
 	/* LIGHPROBES */
@@ -1449,8 +1434,7 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 		Ref<Image> img;
 		img.instance();
 		img->create(atlas_size.width, atlas_size.height, false, Image::FORMAT_RGBAH, s);
-		img->convert(Image::FORMAT_RGBA8);
-		img->save_png("res://4_light_secondary_" + itos(i) + ".png");
+		img->save_exr("res://4_light_secondary_" + itos(i) + ".exr", false);
 	}
 #endif
 
@@ -1582,6 +1566,11 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 		clear_colors.push_back(Color(0, 0, 0, 1));
 		for (int i = 0; i < atlas_slices; i++) {
 			int subslices = (p_bake_sh ? 4 : 1);
+
+			if (slice_seam_count[i] == 0) {
+				continue;
+			}
+
 			for (int k = 0; k < subslices; k++) {
 				RasterSeamsPushConstant seams_push_constant;
 				seams_push_constant.slice = uint32_t(i * subslices + k);
@@ -1654,8 +1643,7 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 		Ref<Image> img;
 		img.instance();
 		img->create(atlas_size.width, atlas_size.height, false, Image::FORMAT_RGBAH, s);
-		img->convert(Image::FORMAT_RGBA8);
-		img->save_png("res://5_blendseams" + itos(i) + ".png");
+		img->save_exr("res://5_blendseams" + itos(i) + ".exr", false);
 	}
 #endif
 	if (p_step_function) {
@@ -1682,7 +1670,7 @@ LightmapperRD::BakeError LightmapperRD::bake(BakeQuality p_quality, bool p_use_d
 			Ref<Image> img2;
 			img2.instance();
 			img2->create(probe_values.size(), 1, false, Image::FORMAT_RGBAF, probe_data);
-			img2->save_png("res://6_lightprobes.png");
+			img2->save_exr("res://6_lightprobes.exr", false);
 		}
 #endif
 	}
diff --git a/modules/lightmapper_rd/lightmapper_rd.h b/modules/lightmapper_rd/lightmapper_rd.h
index f2a826a447..7ab7f34464 100644
--- a/modules/lightmapper_rd/lightmapper_rd.h
+++ b/modules/lightmapper_rd/lightmapper_rd.h
@@ -231,7 +231,7 @@ class LightmapperRD : public Lightmapper {
 	Vector<Color> probe_values;
 
 	BakeError _blit_meshes_into_atlas(int p_max_texture_size, Vector<Ref<Image>> &albedo_images, Vector<Ref<Image>> &emission_images, AABB &bounds, Size2i &atlas_size, int &atlas_slices, BakeStepFunc p_step_function, void *p_bake_userdata);
-	void _create_acceleration_structures(RenderingDevice *rd, Size2i atlas_size, int atlas_slices, AABB &bounds, int grid_size, Vector<Probe> &probe_positions, GenerateProbes p_generate_probes, Vector<int> &slice_triangle_count, Vector<int> &slice_seam_count, RID &vertex_buffer, RID &triangle_buffer, RID &box_buffer, RID &lights_buffer, RID &triangle_cell_indices_buffer, RID &probe_positions_buffer, RID &grid_texture, RID &grid_texture_sdf, RID &seams_buffer, BakeStepFunc p_step_function, void *p_bake_userdata);
+	void _create_acceleration_structures(RenderingDevice *rd, Size2i atlas_size, int atlas_slices, AABB &bounds, int grid_size, Vector<Probe> &probe_positions, GenerateProbes p_generate_probes, Vector<int> &slice_triangle_count, Vector<int> &slice_seam_count, RID &vertex_buffer, RID &triangle_buffer, RID &box_buffer, RID &lights_buffer, RID &triangle_cell_indices_buffer, RID &probe_positions_buffer, RID &grid_texture, RID &seams_buffer, BakeStepFunc p_step_function, void *p_bake_userdata);
 	void _raster_geometry(RenderingDevice *rd, Size2i atlas_size, int atlas_slices, int grid_size, AABB bounds, float p_bias, Vector<int> slice_triangle_count, RID position_tex, RID unocclude_tex, RID normal_tex, RID raster_depth_buffer, RID rasterize_shader, RID raster_base_uniform);
 
 public:
diff --git a/modules/lightmapper_rd/lm_common_inc.glsl b/modules/lightmapper_rd/lm_common_inc.glsl
index f8a0cd16de..1581639036 100644
--- a/modules/lightmapper_rd/lm_common_inc.glsl
+++ b/modules/lightmapper_rd/lm_common_inc.glsl
@@ -84,9 +84,8 @@ layout(set = 0, binding = 7, std430) restrict readonly buffer Probes {
 probe_positions;
 
 layout(set = 0, binding = 8) uniform utexture3D grid;
-layout(set = 0, binding = 9) uniform texture3D grid_sdf;
 
-layout(set = 0, binding = 10) uniform texture2DArray albedo_tex;
-layout(set = 0, binding = 11) uniform texture2DArray emission_tex;
+layout(set = 0, binding = 9) uniform texture2DArray albedo_tex;
+layout(set = 0, binding = 10) uniform texture2DArray emission_tex;
 
-layout(set = 0, binding = 12) uniform sampler linear_sampler;
+layout(set = 0, binding = 11) uniform sampler linear_sampler;
diff --git a/modules/lightmapper_rd/lm_compute.glsl b/modules/lightmapper_rd/lm_compute.glsl
index 3dd96893fb..9ca40535f9 100644
--- a/modules/lightmapper_rd/lm_compute.glsl
+++ b/modules/lightmapper_rd/lm_compute.glsl
@@ -96,15 +96,22 @@ params;
 bool ray_hits_triangle(vec3 from, vec3 dir, float max_dist, vec3 p0, vec3 p1, vec3 p2, out float r_distance, out vec3 r_barycentric) {
 	const vec3 e0 = p1 - p0;
 	const vec3 e1 = p0 - p2;
-	vec3 triangleNormal = cross(e1, e0);
+	vec3 triangle_normal = cross(e1, e0);
 
-	const vec3 e2 = (1.0 / dot(triangleNormal, dir)) * (p0 - from);
+	float n_dot_dir = dot(triangle_normal, dir);
+
+	if (abs(n_dot_dir) < 0.01) {
+		return false;
+	}
+
+	const vec3 e2 = (p0 - from) / n_dot_dir;
 	const vec3 i = cross(dir, e2);
 
 	r_barycentric.y = dot(i, e1);
 	r_barycentric.z = dot(i, e0);
 	r_barycentric.x = 1.0 - (r_barycentric.z + r_barycentric.y);
-	r_distance = dot(triangleNormal, e2);
+	r_distance = dot(triangle_normal, e2);
+
 	return (r_distance > params.bias) && (r_distance < max_dist) && all(greaterThanEqual(r_barycentric, vec3(0.0)));
 }
 
@@ -307,8 +314,6 @@ void main() {
 				continue;
 			}
 
-			d /= lights.data[i].range;
-
 			attenuation = get_omni_attenuation(d, 1.0 / lights.data[i].range, lights.data[i].attenuation);
 
 			if (lights.data[i].type == LIGHT_TYPE_SPOT) {
@@ -410,7 +415,7 @@ void main() {
 		uint tidx;
 		vec3 barycentric;
 
-		vec3 light;
+		vec3 light = vec3(0.0);
 		if (trace_ray(position + ray_dir * params.bias, position + ray_dir * length(params.world_size), tidx, barycentric)) {
 			//hit a triangle
 			vec2 uv0 = vertices.data[triangles.data[tidx].indices.x].uv;
@@ -419,8 +424,8 @@ void main() {
 			vec3 uvw = vec3(barycentric.x * uv0 + barycentric.y * uv1 + barycentric.z * uv2, float(triangles.data[tidx].slice));
 
 			light = textureLod(sampler2DArray(source_light, linear_sampler), uvw, 0.0).rgb;
-		} else {
-			//did not hit a triangle, reach out for the sky
+		} else if (params.env_transform[0][3] == 0.0) { // Use env_transform[0][3] to indicate when we are computing the first bounce
+			// Did not hit a triangle, reach out for the sky
 			vec3 sky_dir = normalize(mat3(params.env_transform) * ray_dir);
 
 			vec2 st = vec2(
diff --git a/modules/tinyexr/image_saver_tinyexr.cpp b/modules/tinyexr/image_saver_tinyexr.cpp
index f747763248..6a2fb0f666 100644
--- a/modules/tinyexr/image_saver_tinyexr.cpp
+++ b/modules/tinyexr/image_saver_tinyexr.cpp
@@ -169,7 +169,7 @@ Error save_exr(const String &p_path, const Ref<Image> &p_img, bool p_grayscale)
 		{ 0 }, // R
 		{ 1, 0 }, // GR
 		{ 2, 1, 0 }, // BGR
-		{ 2, 1, 0, 3 } // BGRA
+		{ 3, 2, 1, 0 } // ABGR
 	};
 
 	int channel_count = get_channel_count(format);
diff --git a/modules/xatlas_unwrap/register_types.cpp b/modules/xatlas_unwrap/register_types.cpp
index e1f9521a48..8913ef1b65 100644
--- a/modules/xatlas_unwrap/register_types.cpp
+++ b/modules/xatlas_unwrap/register_types.cpp
@@ -29,26 +29,19 @@
 /*************************************************************************/
 
 #include "register_types.h"
-
-#include "core/error/error_macros.h"
-
 #include "core/crypto/crypto_core.h"
-
 #include "thirdparty/xatlas/xatlas.h"
 
-#include <stdio.h>
-#include <stdlib.h>
+extern bool (*array_mesh_lightmap_unwrap_callback)(float p_texel_size, const float *p_vertices, const float *p_normals, int p_vertex_count, const int *p_indices, int p_index_count, const uint8_t *p_cache_data, bool *r_use_cache, uint8_t **r_mesh_cache, int *r_mesh_cache_size, float **r_uv, int **r_vertex, int *r_vertex_count, int **r_index, int *r_index_count, int *r_size_hint_x, int *r_size_hint_y);
 
-extern bool (*array_mesh_lightmap_unwrap_callback)(float p_texel_size, const float *p_vertices, const float *p_normals, int p_vertex_count, const int *p_indices, int p_index_count, float **r_uv, int **r_vertex, int *r_vertex_count, int **r_index, int *r_index_count, int *r_size_hint_x, int *r_size_hint_y, int *&r_cache_data, unsigned int &r_cache_size, bool &r_used_cache);
-
-bool xatlas_mesh_lightmap_unwrap_callback(float p_texel_size, const float *p_vertices, const float *p_normals, int p_vertex_count, const int *p_indices, int p_index_count, float **r_uvs, int **r_vertices, int *r_vertex_count, int **r_indices, int *r_index_count, int *r_size_hint_x, int *r_size_hint_y, int *&r_cache_data, unsigned int &r_cache_size, bool &r_used_cache) {
+bool xatlas_mesh_lightmap_unwrap_callback(float p_texel_size, const float *p_vertices, const float *p_normals, int p_vertex_count, const int *p_indices, int p_index_count, const uint8_t *p_cache_data, bool *r_use_cache, uint8_t **r_mesh_cache, int *r_mesh_cache_size, float **r_uv, int **r_vertex, int *r_vertex_count, int **r_index, int *r_index_count, int *r_size_hint_x, int *r_size_hint_y) {
 	CryptoCore::MD5Context ctx;
 	ctx.start();
 
 	ctx.update((unsigned char *)&p_texel_size, sizeof(float));
 	ctx.update((unsigned char *)p_indices, sizeof(int) * p_index_count);
-	ctx.update((unsigned char *)p_vertices, sizeof(float) * p_vertex_count);
-	ctx.update((unsigned char *)p_normals, sizeof(float) * p_vertex_count);
+	ctx.update((unsigned char *)p_vertices, sizeof(float) * p_vertex_count * 3);
+	ctx.update((unsigned char *)p_normals, sizeof(float) * p_vertex_count * 3);
 
 	unsigned char hash[16];
 	ctx.finish(hash);
@@ -56,38 +49,37 @@ bool xatlas_mesh_lightmap_unwrap_callback(float p_texel_size, const float *p_ver
 	bool cached = false;
 	unsigned int cache_idx = 0;
 
-	if (r_used_cache && r_cache_size) {
-		//Check if hash is in cache data
+	*r_mesh_cache = nullptr;
+	*r_mesh_cache_size = 0;
 
-		int *cache_data = r_cache_data;
+	if (p_cache_data) {
+		//Check if hash is in cache data
+		int *cache_data = (int *)p_cache_data;
 		int n_entries = cache_data[0];
-		unsigned int r_idx = 1;
+		unsigned int read_idx = 1;
 		for (int i = 0; i < n_entries; ++i) {
-			if (memcmp(&cache_data[r_idx], hash, 16) == 0) {
+			if (memcmp(&cache_data[read_idx], hash, 16) == 0) {
 				cached = true;
-				cache_idx = r_idx;
+				cache_idx = read_idx;
 				break;
 			}
 
-			r_idx += 4; // hash
-			r_idx += 2; // size hint
+			read_idx += 4; // hash
+			read_idx += 2; // size hint
 
-			int vertex_count = cache_data[r_idx];
-			r_idx += 1; // vertex count
-			r_idx += vertex_count; // vertex
-			r_idx += vertex_count * 2; // uvs
+			int vertex_count = cache_data[read_idx];
+			read_idx += 1; // vertex count
+			read_idx += vertex_count; // vertex
+			read_idx += vertex_count * 2; // uvs
 
-			int index_count = cache_data[r_idx];
-			r_idx += 1; // index count
-			r_idx += index_count; // indices
+			int index_count = cache_data[read_idx];
+			read_idx += 1; // index count
+			read_idx += index_count; // indices
 		}
 	}
 
-	if (r_used_cache && cached) {
-		int *cache_data = r_cache_data;
-
-		// Return cache data pointer to the caller
-		r_cache_data = &cache_data[cache_idx];
+	if (cached) {
+		int *cache_data = (int *)p_cache_data;
 
 		cache_idx += 4;
 
@@ -99,96 +91,92 @@ bool xatlas_mesh_lightmap_unwrap_callback(float p_texel_size, const float *p_ver
 		// Load vertices
 		*r_vertex_count = cache_data[cache_idx];
 		cache_idx++;
-		*r_vertices = &cache_data[cache_idx];
+		*r_vertex = &cache_data[cache_idx];
 		cache_idx += *r_vertex_count;
 
 		// Load UVs
-		*r_uvs = (float *)&cache_data[cache_idx];
+		*r_uv = (float *)&cache_data[cache_idx];
 		cache_idx += *r_vertex_count * 2;
 
 		// Load indices
 		*r_index_count = cache_data[cache_idx];
 		cache_idx++;
-		*r_indices = &cache_data[cache_idx];
-
-		// Return cache data size to the caller
-		r_cache_size = sizeof(int) * (4 + 2 + 1 + *r_vertex_count + (*r_vertex_count * 2) + 1 + *r_index_count); // hash + size hint + vertex_count + vertices + uvs + index_count + indices
-		r_used_cache = true;
-		return true;
-	}
-
-	//set up input mesh
-	xatlas::MeshDecl input_mesh;
-	input_mesh.indexData = p_indices;
-	input_mesh.indexCount = p_index_count;
-	input_mesh.indexFormat = xatlas::IndexFormat::UInt32;
-
-	input_mesh.vertexCount = p_vertex_count;
-	input_mesh.vertexPositionData = p_vertices;
-	input_mesh.vertexPositionStride = sizeof(float) * 3;
-	input_mesh.vertexNormalData = p_normals;
-	input_mesh.vertexNormalStride = sizeof(uint32_t) * 3;
-	input_mesh.vertexUvData = nullptr;
-	input_mesh.vertexUvStride = 0;
-
-	xatlas::ChartOptions chart_options;
-	xatlas::PackOptions pack_options;
-
-	pack_options.maxChartSize = 4096;
-	pack_options.blockAlign = true;
-	pack_options.padding = 1;
-	pack_options.texelsPerUnit = 1.0 / p_texel_size;
+		*r_index = &cache_data[cache_idx];
+	} else {
+		// set up input mesh
+		xatlas::MeshDecl input_mesh;
+		input_mesh.indexData = p_indices;
+		input_mesh.indexCount = p_index_count;
+		input_mesh.indexFormat = xatlas::IndexFormat::UInt32;
+
+		input_mesh.vertexCount = p_vertex_count;
+		input_mesh.vertexPositionData = p_vertices;
+		input_mesh.vertexPositionStride = sizeof(float) * 3;
+		input_mesh.vertexNormalData = p_normals;
+		input_mesh.vertexNormalStride = sizeof(uint32_t) * 3;
+		input_mesh.vertexUvData = NULL;
+		input_mesh.vertexUvStride = 0;
+
+		xatlas::ChartOptions chart_options;
+		chart_options.fixWinding = true;
+
+		xatlas::PackOptions pack_options;
+		pack_options.padding = 1;
+		pack_options.maxChartSize = 4094; // Lightmap atlassing needs 2 for padding between meshes, so 4096-2
+		pack_options.blockAlign = true;
+		pack_options.texelsPerUnit = 1.0 / p_texel_size;
+
+		xatlas::Atlas *atlas = xatlas::Create();
+
+		xatlas::AddMeshError err = xatlas::AddMesh(atlas, input_mesh, 1);
+		ERR_FAIL_COND_V_MSG(err != xatlas::AddMeshError::Success, false, xatlas::StringForEnum(err));
+
+		xatlas::Generate(atlas, chart_options, pack_options);
+
+		*r_size_hint_x = atlas->width;
+		*r_size_hint_y = atlas->height;
+
+		float w = *r_size_hint_x;
+		float h = *r_size_hint_y;
+
+		if (w == 0 || h == 0) {
+			xatlas::Destroy(atlas);
+			return false; //could not bake because there is no area
+		}
 
-	xatlas::Atlas *atlas = xatlas::Create();
-	printf("Adding mesh..\n");
-	xatlas::AddMeshError err = xatlas::AddMesh(atlas, input_mesh, 1);
-	ERR_FAIL_COND_V_MSG(err != xatlas::AddMeshError::Success, false, xatlas::StringForEnum(err));
+		const xatlas::Mesh &output = atlas->meshes[0];
+
+		*r_vertex = (int *)memalloc(sizeof(int) * output.vertexCount);
+		ERR_FAIL_NULL_V_MSG(*r_vertex, false, "Out of memory.");
+		*r_uv = (float *)memalloc(sizeof(float) * output.vertexCount * 2);
+		ERR_FAIL_NULL_V_MSG(*r_uv, false, "Out of memory.");
+		*r_index = (int *)memalloc(sizeof(int) * output.indexCount);
+		ERR_FAIL_NULL_V_MSG(*r_index, false, "Out of memory.");
+
+		float max_x = 0;
+		float max_y = 0;
+		for (uint32_t i = 0; i < output.vertexCount; i++) {
+			(*r_vertex)[i] = output.vertexArray[i].xref;
+			(*r_uv)[i * 2 + 0] = output.vertexArray[i].uv[0] / w;
+			(*r_uv)[i * 2 + 1] = output.vertexArray[i].uv[1] / h;
+			max_x = MAX(max_x, output.vertexArray[i].uv[0]);
+			max_y = MAX(max_y, output.vertexArray[i].uv[1]);
+		}
 
-	printf("Generate..\n");
-	xatlas::Generate(atlas, chart_options, pack_options);
+		*r_vertex_count = output.vertexCount;
 
-	*r_size_hint_x = atlas->width;
-	*r_size_hint_y = atlas->height;
+		for (uint32_t i = 0; i < output.indexCount; i++) {
+			(*r_index)[i] = output.indexArray[i];
+		}
 
-	float w = *r_size_hint_x;
-	float h = *r_size_hint_y;
+		*r_index_count = output.indexCount;
 
-	if (w == 0 || h == 0) {
 		xatlas::Destroy(atlas);
-		return false; //could not bake because there is no area
-	}
-
-	const xatlas::Mesh &output = atlas->meshes[0];
-
-	*r_vertices = (int *)malloc(sizeof(int) * output.vertexCount);
-	ERR_FAIL_NULL_V_MSG(*r_vertices, false, "Out of memory.");
-	*r_uvs = (float *)malloc(sizeof(float) * output.vertexCount * 2);
-	ERR_FAIL_NULL_V_MSG(*r_uvs, false, "Out of memory.");
-	*r_indices = (int *)malloc(sizeof(int) * output.indexCount);
-	ERR_FAIL_NULL_V_MSG(*r_indices, false, "Out of memory.");
-
-	float max_x = 0.0;
-	float max_y = 0.0;
-	for (uint32_t i = 0; i < output.vertexCount; i++) {
-		(*r_vertices)[i] = output.vertexArray[i].xref;
-		(*r_uvs)[i * 2 + 0] = output.vertexArray[i].uv[0] / w;
-		(*r_uvs)[i * 2 + 1] = output.vertexArray[i].uv[1] / h;
-		max_x = MAX(max_x, output.vertexArray[i].uv[0]);
-		max_y = MAX(max_y, output.vertexArray[i].uv[1]);
 	}
 
-	printf("Final texture size: %f,%f - max %f,%f\n", w, h, max_x, max_y);
-	*r_vertex_count = output.vertexCount;
+	if (*r_use_cache) {
+		// Build cache data for current mesh
 
-	for (uint32_t i = 0; i < output.indexCount; i++) {
-		(*r_indices)[i] = output.indexArray[i];
-	}
-
-	*r_index_count = output.indexCount;
-
-	xatlas::Destroy(atlas);
-
-	if (r_used_cache) {
 		unsigned int new_cache_size = 4 + 2 + 1 + *r_vertex_count + (*r_vertex_count * 2) + 1 + *r_index_count; // hash + size hint + vertex_count + vertices + uvs + index_count + indices
 		new_cache_size *= sizeof(int);
 		int *new_cache_data = (int *)memalloc(new_cache_size);
@@ -208,11 +196,11 @@ bool xatlas_mesh_lightmap_unwrap_callback(float p_texel_size, const float *p_ver
 		new_cache_idx++;
 
 		// vertices
-		memcpy(&new_cache_data[new_cache_idx], *r_vertices, sizeof(int) * *r_vertex_count);
+		memcpy(&new_cache_data[new_cache_idx], *r_vertex, sizeof(int) * (*r_vertex_count));
 		new_cache_idx += *r_vertex_count;
 
 		// uvs
-		memcpy(&new_cache_data[new_cache_idx], *r_uvs, sizeof(float) * *r_vertex_count * 2);
+		memcpy(&new_cache_data[new_cache_idx], *r_uv, sizeof(float) * (*r_vertex_count) * 2);
 		new_cache_idx += *r_vertex_count * 2;
 
 		// index count
@@ -220,15 +208,15 @@ bool xatlas_mesh_lightmap_unwrap_callback(float p_texel_size, const float *p_ver
 		new_cache_idx++;
 
 		// indices
-		memcpy(&new_cache_data[new_cache_idx], *r_indices, sizeof(int) * *r_index_count);
-		new_cache_idx += *r_index_count;
+		memcpy(&new_cache_data[new_cache_idx], *r_index, sizeof(int) * (*r_index_count));
 
 		// Return cache data to the caller
-		r_cache_data = new_cache_data;
-		r_cache_size = new_cache_size;
-		r_used_cache = false;
+		*r_mesh_cache = (uint8_t *)new_cache_data;
+		*r_mesh_cache_size = new_cache_size;
 	}
 
+	*r_use_cache = cached; // Return whether cache was used.
+
 	return true;
 }