Porting cubemap compute shaders to raster for the mobile renderer

7 files changed, 626 insertions, 141 deletions

diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl
index 9fa84657d1..63f0ce690e 100644
--- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl
+++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl
@@ -32,53 +32,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cubemap;
 
 layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap;
 
-layout(push_constant, binding = 1, std430) uniform Params {
-	uint face_size;
-}
-params;
-
-#define M_PI 3.14159265359
-
-void get_dir_0(out vec3 dir, in float u, in float v) {
-	dir[0] = 1.0;
-	dir[1] = v;
-	dir[2] = -u;
-}
-
-void get_dir_1(out vec3 dir, in float u, in float v) {
-	dir[0] = -1.0;
-	dir[1] = v;
-	dir[2] = u;
-}
-
-void get_dir_2(out vec3 dir, in float u, in float v) {
-	dir[0] = u;
-	dir[1] = 1.0;
-	dir[2] = -v;
-}
-
-void get_dir_3(out vec3 dir, in float u, in float v) {
-	dir[0] = u;
-	dir[1] = -1.0;
-	dir[2] = v;
-}
-
-void get_dir_4(out vec3 dir, in float u, in float v) {
-	dir[0] = u;
-	dir[1] = v;
-	dir[2] = 1.0;
-}
-
-void get_dir_5(out vec3 dir, in float u, in float v) {
-	dir[0] = -u;
-	dir[1] = v;
-	dir[2] = -1.0;
-}
-
-float calcWeight(float u, float v) {
-	float val = u * u + v * v + 1.0;
-	return val * sqrt(val);
-}
+#include "cubemap_downsampler_inc.glsl"
 
 void main() {
 	uvec3 id = gl_GlobalInvocationID;
diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl
new file mode 100644
index 0000000000..b329e67293
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl
@@ -0,0 +1,48 @@
+layout(push_constant, binding = 1, std430) uniform Params {
+	uint face_size;
+	uint face_id; // only used in raster shader
+}
+params;
+
+#define M_PI 3.14159265359
+
+void get_dir_0(out vec3 dir, in float u, in float v) {
+	dir[0] = 1.0;
+	dir[1] = v;
+	dir[2] = -u;
+}
+
+void get_dir_1(out vec3 dir, in float u, in float v) {
+	dir[0] = -1.0;
+	dir[1] = v;
+	dir[2] = u;
+}
+
+void get_dir_2(out vec3 dir, in float u, in float v) {
+	dir[0] = u;
+	dir[1] = 1.0;
+	dir[2] = -v;
+}
+
+void get_dir_3(out vec3 dir, in float u, in float v) {
+	dir[0] = u;
+	dir[1] = -1.0;
+	dir[2] = v;
+}
+
+void get_dir_4(out vec3 dir, in float u, in float v) {
+	dir[0] = u;
+	dir[1] = v;
+	dir[2] = 1.0;
+}
+
+void get_dir_5(out vec3 dir, in float u, in float v) {
+	dir[0] = -u;
+	dir[1] = v;
+	dir[2] = -1.0;
+}
+
+float calcWeight(float u, float v) {
+	float val = u * u + v * v + 1.0;
+	return val * sqrt(val);
+}
diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl
new file mode 100644
index 0000000000..0828ffd921
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl
@@ -0,0 +1,163 @@
+// Copyright 2016 Activision Publishing, Inc.
+//
+// Permission is hereby granted, free of charge, to any person obtaining
+// a copy of this software and associated documentation files (the "Software"),
+// to deal in the Software without restriction, including without limitation
+// the rights to use, copy, modify, merge, publish, distribute, sublicense,
+// and/or sell copies of the Software, and to permit persons to whom the Software
+// is furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+
+/* clang-format off */
+#[vertex]
+
+#version 450
+
+#VERSION_DEFINES
+
+#include "cubemap_downsampler_inc.glsl"
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+	vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	uv_interp = base_arr[gl_VertexIndex] * float(params.face_size);
+	gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0);
+}
+
+/* clang-format off */
+#[fragment]
+
+#version 450
+
+#VERSION_DEFINES
+
+#include "cubemap_downsampler_inc.glsl"
+
+layout(set = 0, binding = 0) uniform samplerCube source_cubemap;
+
+layout(location = 0) in vec2 uv_interp;
+layout(location = 0) out vec4 frag_color;
+/* clang-format on */
+
+void main() {
+	// Converted from compute shader which uses absolute coordinates.
+	// Could possibly simplify this
+	float face_size = float(params.face_size);
+
+	if (uv_interp.x < face_size && uv_interp.y < face_size) {
+		float inv_face_size = 1.0 / face_size;
+
+		float u0 = (uv_interp.x * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0;
+		float u1 = (uv_interp.x * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0;
+
+		float v0 = (uv_interp.y * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0;
+		float v1 = (uv_interp.y * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0;
+
+		float weights[4];
+		weights[0] = calcWeight(u0, v0);
+		weights[1] = calcWeight(u1, v0);
+		weights[2] = calcWeight(u0, v1);
+		weights[3] = calcWeight(u1, v1);
+
+		const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]);
+		for (int i = 0; i < 4; i++) {
+			weights[i] = weights[i] * wsum + .125;
+		}
+
+		vec3 dir;
+		vec4 color;
+		switch (params.face_id) {
+			case 0:
+				get_dir_0(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_0(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_0(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_0(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 1:
+				get_dir_1(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_1(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_1(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_1(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 2:
+				get_dir_2(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_2(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_2(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_2(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 3:
+				get_dir_3(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_3(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_3(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_3(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 4:
+				get_dir_4(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_4(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_4(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_4(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			default:
+				get_dir_5(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_5(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_5(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_5(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+		}
+		frag_color = color;
+	}
+}
diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl
new file mode 100644
index 0000000000..324d306218
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl
@@ -0,0 +1,256 @@
+// Copyright 2016 Activision Publishing, Inc.
+//
+// Permission is hereby granted, free of charge, to any person obtaining
+// a copy of this software and associated documentation files (the "Software"),
+// to deal in the Software without restriction, including without limitation
+// the rights to use, copy, modify, merge, publish, distribute, sublicense,
+// and/or sell copies of the Software, and to permit persons to whom the Software
+// is furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+
+/* clang-format off */
+#[vertex]
+
+#version 450
+
+#VERSION_DEFINES
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	int mip_level;
+	uint face_id;
+}
+params;
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+	vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	uv_interp = base_arr[gl_VertexIndex];
+	gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0);
+}
+
+/* clang-format off */
+#[fragment]
+
+#version 450
+
+#VERSION_DEFINES
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	int mip_level;
+	uint face_id;
+}
+params;
+
+layout(set = 0, binding = 0) uniform samplerCube source_cubemap;
+
+layout(location = 0) in vec2 uv_interp;
+layout(location = 0) out vec4 frag_color;
+
+/* clang-format on */
+
+#ifdef USE_HIGH_QUALITY
+#define NUM_TAPS 32
+#else
+#define NUM_TAPS 8
+#endif
+
+#define BASE_RESOLUTION 128
+
+#ifdef USE_HIGH_QUALITY
+layout(set = 1, binding = 0, std430) buffer restrict readonly Data {
+	vec4[7][5][3][24] coeffs;
+}
+data;
+#else
+layout(set = 1, binding = 0, std430) buffer restrict readonly Data {
+	vec4[7][5][6] coeffs;
+}
+data;
+#endif
+
+void get_dir(out vec3 dir, in vec2 uv, in uint face) {
+	switch (face) {
+		case 0:
+			dir = vec3(1.0, uv[1], -uv[0]);
+			break;
+		case 1:
+			dir = vec3(-1.0, uv[1], uv[0]);
+			break;
+		case 2:
+			dir = vec3(uv[0], 1.0, -uv[1]);
+			break;
+		case 3:
+			dir = vec3(uv[0], -1.0, uv[1]);
+			break;
+		case 4:
+			dir = vec3(uv[0], uv[1], 1.0);
+			break;
+		default:
+			dir = vec3(-uv[0], uv[1], -1.0);
+			break;
+	}
+}
+
+void main() {
+	// determine dir / pos for the texel
+	vec3 dir, adir, frameZ;
+	{
+		vec2 uv;
+		uv.x = uv_interp.x;
+		uv.y = 1.0 - uv_interp.y;
+		uv = uv * 2.0 - 1.0;
+
+		get_dir(dir, uv, params.face_id);
+		frameZ = normalize(dir);
+
+		adir = abs(dir);
+	}
+
+	// determine which texel this is
+	// NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name.
+	int mip_level = 0;
+
+	if (params.mip_level < 0) {
+		// return as is
+		frag_color.rgb = textureLod(source_cubemap, frameZ, 0.0).rgb;
+		frag_color.a = 1.0;
+		return;
+	} else if (params.mip_level > 6) {
+		// maximum level
+		mip_level = 6;
+	} else {
+		mip_level = params.mip_level;
+	}
+
+	// GGX gather colors
+	vec4 color = vec4(0.0);
+	for (int axis = 0; axis < 3; axis++) {
+		const int otherAxis0 = 1 - (axis & 1) - (axis >> 1);
+		const int otherAxis1 = 2 - (axis >> 1);
+
+		float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25;
+		if (frameweight > 0.0) {
+			// determine frame
+			vec3 UpVector;
+			switch (axis) {
+				case 0:
+					UpVector = vec3(1, 0, 0);
+					break;
+				case 1:
+					UpVector = vec3(0, 1, 0);
+					break;
+				default:
+					UpVector = vec3(0, 0, 1);
+					break;
+			}
+
+			vec3 frameX = normalize(cross(UpVector, frameZ));
+			vec3 frameY = cross(frameZ, frameX);
+
+			// calculate parametrization for polynomial
+			float Nx = dir[otherAxis0];
+			float Ny = dir[otherAxis1];
+			float Nz = adir[axis];
+
+			float NmaxXY = max(abs(Ny), abs(Nx));
+			Nx /= NmaxXY;
+			Ny /= NmaxXY;
+
+			float theta;
+			if (Ny < Nx) {
+				if (Ny <= -0.999)
+					theta = Nx;
+				else
+					theta = Ny;
+			} else {
+				if (Ny >= 0.999)
+					theta = -Nx;
+				else
+					theta = -Ny;
+			}
+
+			float phi;
+			if (Nz <= -0.999)
+				phi = -NmaxXY;
+			else if (Nz >= 0.999)
+				phi = NmaxXY;
+			else
+				phi = Nz;
+
+			float theta2 = theta * theta;
+			float phi2 = phi * phi;
+
+			// sample
+			for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) {
+				const int index = (NUM_TAPS / 4) * axis + iSuperTap;
+
+#ifdef USE_HIGH_QUALITY
+				vec4 coeffsDir0[3];
+				vec4 coeffsDir1[3];
+				vec4 coeffsDir2[3];
+				vec4 coeffsLevel[3];
+				vec4 coeffsWeight[3];
+
+				for (int iCoeff = 0; iCoeff < 3; iCoeff++) {
+					coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index];
+					coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index];
+					coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index];
+					coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index];
+					coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index];
+				}
+
+				for (int iSubTap = 0; iSubTap < 4; iSubTap++) {
+					// determine sample attributes (dir, weight, mip_level)
+					vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2);
+
+					float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2;
+
+					float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2;
+#else
+				vec4 coeffsDir0 = data.coeffs[mip_level][0][index];
+				vec4 coeffsDir1 = data.coeffs[mip_level][1][index];
+				vec4 coeffsDir2 = data.coeffs[mip_level][2][index];
+				vec4 coeffsLevel = data.coeffs[mip_level][3][index];
+				vec4 coeffsWeight = data.coeffs[mip_level][4][index];
+
+				for (int iSubTap = 0; iSubTap < 4; iSubTap++) {
+					// determine sample attributes (dir, weight, mip_level)
+					vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap];
+
+					float sample_level = coeffsLevel[iSubTap];
+
+					float sample_weight = coeffsWeight[iSubTap];
+#endif
+
+					sample_weight *= frameweight;
+
+					// adjust for jacobian
+					sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2])));
+					sample_level += 0.75 * log2(dot(sample_dir, sample_dir));
+					// sample cubemap
+					color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight;
+					color.w += sample_weight;
+				}
+			}
+		}
+	}
+	color /= color.w;
+
+	// write color
+	color.xyz = max(vec3(0.0), color.xyz);
+	color.w = 1.0;
+
+	frag_color = color;
+}
diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl
index ce7c03c1d4..28f4dc59ec 100644
--- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl
+++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl
@@ -12,100 +12,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube;
 
 layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap;
 
-layout(push_constant, binding = 1, std430) uniform Params {
-	uint face_id;
-	uint sample_count;
-	float roughness;
-	bool use_direct_write;
-	float face_size;
-}
-params;
-
-#define M_PI 3.14159265359
-
-vec3 texelCoordToVec(vec2 uv, uint faceID) {
-	mat3 faceUvVectors[6];
-
-	// -x
-	faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z
-	faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y
-	faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face
-
-	// +x
-	faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z
-	faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y
-	faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face
-
-	// -y
-	faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x
-	faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z
-	faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face
-
-	// +y
-	faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x
-	faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z
-	faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face
-
-	// -z
-	faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x
-	faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y
-	faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face
-
-	// +z
-	faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x
-	faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y
-	faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face
-
-	// out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2].
-	vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2];
-	return normalize(result);
-}
-
-vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) {
-	float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph]
-
-	// Compute distribution direction
-	float Phi = 2.0 * M_PI * Xi.x;
-	float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y));
-	float SinTheta = sqrt(1.0 - CosTheta * CosTheta);
-
-	// Convert to spherical direction
-	vec3 H;
-	H.x = SinTheta * cos(Phi);
-	H.y = SinTheta * sin(Phi);
-	H.z = CosTheta;
-
-	vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0);
-	vec3 TangentX = normalize(cross(UpVector, N));
-	vec3 TangentY = cross(N, TangentX);
-
-	// Tangent to world space
-	return TangentX * H.x + TangentY * H.y + N * H.z;
-}
-
-// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
-float GGX(float NdotV, float a) {
-	float k = a / 2.0;
-	return NdotV / (NdotV * (1.0 - k) + k);
-}
-
-// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
-float G_Smith(float a, float nDotV, float nDotL) {
-	return GGX(nDotL, a * a) * GGX(nDotV, a * a);
-}
-
-float radicalInverse_VdC(uint bits) {
-	bits = (bits << 16u) | (bits >> 16u);
-	bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u);
-	bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u);
-	bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u);
-	bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u);
-	return float(bits) * 2.3283064365386963e-10; // / 0x100000000
-}
-
-vec2 Hammersley(uint i, uint N) {
-	return vec2(float(i) / float(N), radicalInverse_VdC(i));
-}
+#include "cubemap_roughness_inc.glsl"
 
 void main() {
 	uvec3 id = gl_GlobalInvocationID;
diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl
new file mode 100644
index 0000000000..80c0ac4fb4
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl
@@ -0,0 +1,94 @@
+#define M_PI 3.14159265359
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	uint face_id;
+	uint sample_count;
+	float roughness;
+	bool use_direct_write;
+	float face_size;
+}
+params;
+
+vec3 texelCoordToVec(vec2 uv, uint faceID) {
+	mat3 faceUvVectors[6];
+
+	// -x
+	faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z
+	faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face
+
+	// +x
+	faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z
+	faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face
+
+	// -y
+	faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x
+	faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z
+	faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face
+
+	// +y
+	faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x
+	faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z
+	faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face
+
+	// -z
+	faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x
+	faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face
+
+	// +z
+	faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x
+	faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face
+
+	// out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2].
+	vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2];
+	return normalize(result);
+}
+
+vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) {
+	float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph]
+
+	// Compute distribution direction
+	float Phi = 2.0 * M_PI * Xi.x;
+	float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y));
+	float SinTheta = sqrt(1.0 - CosTheta * CosTheta);
+
+	// Convert to spherical direction
+	vec3 H;
+	H.x = SinTheta * cos(Phi);
+	H.y = SinTheta * sin(Phi);
+	H.z = CosTheta;
+
+	vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0);
+	vec3 TangentX = normalize(cross(UpVector, N));
+	vec3 TangentY = cross(N, TangentX);
+
+	// Tangent to world space
+	return TangentX * H.x + TangentY * H.y + N * H.z;
+}
+
+// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
+float GGX(float NdotV, float a) {
+	float k = a / 2.0;
+	return NdotV / (NdotV * (1.0 - k) + k);
+}
+
+// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
+float G_Smith(float a, float nDotV, float nDotL) {
+	return GGX(nDotL, a * a) * GGX(nDotV, a * a);
+}
+
+float radicalInverse_VdC(uint bits) {
+	bits = (bits << 16u) | (bits >> 16u);
+	bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u);
+	bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u);
+	bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u);
+	bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u);
+	return float(bits) * 2.3283064365386963e-10; // / 0x100000000
+}
+
+vec2 Hammersley(uint i, uint N) {
+	return vec2(float(i) / float(N), radicalInverse_VdC(i));
+}
diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl
new file mode 100644
index 0000000000..2570308816
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl
@@ -0,0 +1,63 @@
+/* clang-format off */
+#[vertex]
+
+#version 450
+
+#VERSION_DEFINES
+
+#include "cubemap_roughness_inc.glsl"
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+	vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	uv_interp = base_arr[gl_VertexIndex];
+	gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0);
+}
+
+/* clang-format off */
+#[fragment]
+
+#version 450
+
+#VERSION_DEFINES
+
+#include "cubemap_roughness_inc.glsl"
+
+layout(location = 0) in vec2 uv_interp;
+
+layout(set = 0, binding = 0) uniform samplerCube source_cube;
+
+layout(location = 0) out vec4 frag_color;
+/* clang-format on */
+
+void main() {
+	vec3 N = texelCoordToVec(uv_interp * 2.0 - 1.0, params.face_id);
+
+	//vec4 color = color_interp;
+
+	if (params.use_direct_write) {
+		frag_color = vec4(texture(source_cube, N).rgb, 1.0);
+	} else {
+		vec4 sum = vec4(0.0, 0.0, 0.0, 0.0);
+
+		for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) {
+			vec2 xi = Hammersley(sampleNum, params.sample_count);
+
+			vec3 H = ImportanceSampleGGX(xi, params.roughness, N);
+			vec3 V = N;
+			vec3 L = (2.0 * dot(V, H) * H - V);
+
+			float ndotl = clamp(dot(N, L), 0.0, 1.0);
+
+			if (ndotl > 0.0) {
+				sum.rgb += textureLod(source_cube, L, 0.0).rgb * ndotl;
+				sum.a += ndotl;
+			}
+		}
+		sum /= sum.a;
+
+		frag_color = vec4(sum.rgb, 1.0);
+	}
+}