Add Betsy to speed up BC6 compression

Author: BlueCube3310

core/io/image.cpp

@@ -30,6 +30,7 @@
 
 #include "image.h"
 
+#include "core/config/project_settings.h"
 #include "core/error/error_list.h"
 #include "core/error/error_macros.h"
 #include "core/io/image_loader.h"
@@ -2734,6 +2735,27 @@ Error Image::compress(CompressMode p_mode, CompressSource p_source, ASTCFormat p
 Error Image::compress_from_channels(CompressMode p_mode, UsedChannels p_channels, ASTCFormat p_astc_format) {
 	ERR_FAIL_COND_V(data.is_empty(), ERR_INVALID_DATA);
 
+	// RenderingDevice only.
+	if (GLOBAL_GET("rendering/textures/vram_compression/compress_with_gpu")) {
+		switch (p_mode) {
+			case COMPRESS_BPTC: {
+				// BC7 is unsupported currently.
+				if ((format >= FORMAT_RF && format <= FORMAT_RGBE9995) && _image_compress_bptc_rd_func) {
+					Error result = _image_compress_bptc_rd_func(this, p_channels);
+
+					// If the image was compressed successfully, we return here. If not, we fall back to the default compression scheme.
+					if (result == OK) {
+						return OK;
+					}
+				}
+
+			} break;
+
+			default: {
+			}
+		}
+	}
+
 	switch (p_mode) {
 		case COMPRESS_S3TC: {
 			ERR_FAIL_NULL_V(_image_compress_bc_func, ERR_UNAVAILABLE);
@@ -3115,6 +3137,7 @@ void (*Image::_image_compress_bptc_func)(Image *, Image::UsedChannels) = nullptr
 void (*Image::_image_compress_etc1_func)(Image *) = nullptr;
 void (*Image::_image_compress_etc2_func)(Image *, Image::UsedChannels) = nullptr;
 void (*Image::_image_compress_astc_func)(Image *, Image::ASTCFormat) = nullptr;
+Error (*Image::_image_compress_bptc_rd_func)(Image *, Image::UsedChannels) = nullptr;
 void (*Image::_image_decompress_bc)(Image *) = nullptr;
 void (*Image::_image_decompress_bptc)(Image *) = nullptr;
 void (*Image::_image_decompress_etc1)(Image *) = nullptr;

core/io/image.h

@@ -159,6 +159,8 @@ class Image : public Resource {
 	static void (*_image_compress_etc2_func)(Image *, UsedChannels p_channels);
 	static void (*_image_compress_astc_func)(Image *, ASTCFormat p_format);
 
+	static Error (*_image_compress_bptc_rd_func)(Image *, UsedChannels p_channels);
+
 	static void (*_image_decompress_bc)(Image *);
 	static void (*_image_decompress_bptc)(Image *);
 	static void (*_image_decompress_etc1)(Image *);

doc/classes/ProjectSettings.xml

@@ -2875,6 +2875,11 @@
 		<member name="rendering/textures/lossless_compression/force_png" type="bool" setter="" getter="" default="false">
 			If [code]true[/code], the texture importer will import lossless textures using the PNG format. Otherwise, it will default to using WebP.
 		</member>
+		<member name="rendering/textures/vram_compression/compress_with_gpu" type="bool" setter="" getter="" default="true">
+			If [code]true[/code], the texture importer will utilize the GPU for compressing textures, which makes large textures import significantly faster.
+			[b]Note:[/b] This setting requires either Vulkan or D3D12 available as a rendering backend.
+			[b]Note:[/b] Currently this only affects BC6H compression, which is used on Desktop and Console for HDR images.
+		</member>
 		<member name="rendering/textures/vram_compression/import_etc2_astc" type="bool" setter="" getter="" default="false">
 			If [code]true[/code], the texture importer will import VRAM-compressed textures using the Ericsson Texture Compression 2 algorithm for lower quality textures and normal maps and Adaptable Scalable Texture Compression algorithm for high quality textures (in 4×4 block size).
 			[b]Note:[/b] This setting is an override. The texture importer will always import the format the host platform needs, even if this is set to [code]false[/code].

editor/import/resource_importer_layered_texture.cpp

@@ -341,8 +341,6 @@ Error ResourceImporterLayeredTexture::import(const String &p_source_file, const
 	}
 
 	if (compress_mode == COMPRESS_VRAM_COMPRESSED) {
-		mipmaps = true;
-
 		//if using video ram, optimize
 		if (channel_pack == 0) {
 			//remove alpha if not needed, so compression is more efficient

modules/betsy/CrossPlatformSettings_piece_all.glsl

@@ -0,0 +1,76 @@
+
+#define min3(a, b, c) min(a, min(b, c))
+#define max3(a, b, c) max(a, max(b, c))
+
+#define float2 vec2
+#define float3 vec3
+#define float4 vec4
+
+#define int2 ivec2
+#define int3 ivec3
+#define int4 ivec4
+
+#define uint2 uvec2
+#define uint3 uvec3
+#define uint4 uvec4
+
+#define float2x2 mat2
+#define float3x3 mat3
+#define float4x4 mat4
+#define ogre_float4x3 mat3x4
+
+#define ushort uint
+#define ushort3 uint3
+#define ushort4 uint4
+
+//Short used for read operations. It's an int in GLSL & HLSL. An ushort in Metal
+#define rshort int
+#define rshort2 int2
+#define rint int
+//Short used for write operations. It's an int in GLSL. An ushort in HLSL & Metal
+#define wshort2 int2
+#define wshort3 int3
+
+#define toFloat3x3(x) mat3(x)
+#define buildFloat3x3(row0, row1, row2) mat3(row0, row1, row2)
+
+#define mul(x, y) ((x) * (y))
+#define saturate(x) clamp((x), 0.0, 1.0)
+#define lerp mix
+#define rsqrt inversesqrt
+#define INLINE
+#define NO_INTERPOLATION_PREFIX flat
+#define NO_INTERPOLATION_SUFFIX
+
+#define PARAMS_ARG_DECL
+#define PARAMS_ARG
+
+#define reversebits bitfieldReverse
+
+#define OGRE_Sample(tex, sampler, uv) texture(tex, uv)
+#define OGRE_SampleLevel(tex, sampler, uv, lod) textureLod(tex, uv, lod)
+#define OGRE_SampleArray2D(tex, sampler, uv, arrayIdx) texture(tex, vec3(uv, arrayIdx))
+#define OGRE_SampleArray2DLevel(tex, sampler, uv, arrayIdx, lod) textureLod(tex, vec3(uv, arrayIdx), lod)
+#define OGRE_SampleArrayCubeLevel(tex, sampler, uv, arrayIdx, lod) textureLod(tex, vec4(uv, arrayIdx), lod)
+#define OGRE_SampleGrad(tex, sampler, uv, ddx, ddy) textureGrad(tex, uv, ddx, ddy)
+#define OGRE_SampleArray2DGrad(tex, sampler, uv, arrayIdx, ddx, ddy) textureGrad(tex, vec3(uv, arrayIdx), ddx, ddy)
+#define OGRE_ddx(val) dFdx(val)
+#define OGRE_ddy(val) dFdy(val)
+#define OGRE_Load2D(tex, iuv, lod) texelFetch(tex, iuv, lod)
+#define OGRE_LoadArray2D(tex, iuv, arrayIdx, lod) texelFetch(tex, ivec3(iuv, arrayIdx), lod)
+#define OGRE_Load2DMS(tex, iuv, subsample) texelFetch(tex, iuv, subsample)
+
+#define OGRE_Load3D(tex, iuv, lod) texelFetch(tex, ivec3(iuv), lod)
+
+#define OGRE_GatherRed(tex, sampler, uv) textureGather(tex, uv, 0)
+#define OGRE_GatherGreen(tex, sampler, uv) textureGather(tex, uv, 1)
+#define OGRE_GatherBlue(tex, sampler, uv) textureGather(tex, uv, 2)
+
+#define bufferFetch1(buffer, idx) texelFetch(buffer, idx).x
+
+#define OGRE_SAMPLER_ARG_DECL(samplerName)
+#define OGRE_SAMPLER_ARG(samplerName)
+
+#define OGRE_Texture3D_float4 sampler3D
+#define OGRE_OUT_REF(declType, variableName) out declType variableName
+#define OGRE_INOUT_REF(declType, variableName) inout declType variableName

modules/betsy/SCsub

@@ -0,0 +1,24 @@
+# !/ usr / bin / env python
+Import("env")
+Import("env_modules")
+
+env_betsy = env_modules.Clone()
+env_betsy.GLSL_HEADER("bc6h.glsl")
+env_betsy.Depends(Glob("*.glsl.gen.h"), ["#glsl_builders.py"])
+
+# Thirdparty source files
+thirdparty_obj = []
+thirdparty_dir = "#thirdparty/betsy/"
+env_betsy.Prepend(CPPPATH=[thirdparty_dir])
+
+env_thirdparty = env_betsy.Clone()
+env_thirdparty.disable_warnings()
+env.modules_sources += thirdparty_obj
+
+# Godot source files
+module_obj = []
+env_betsy.add_source_files(module_obj, "*.cpp")
+env.modules_sources += module_obj
+
+# Needed to force rebuilding the module files when the thirdparty library is updated.
+env.Depends(module_obj, thirdparty_obj)

modules/betsy/UavCrossPlatform_piece_all.glsl

@@ -0,0 +1,17 @@
+
+#define OGRE_imageLoad2D(inImage, iuv) imageLoad(inImage, int2(iuv))
+#define OGRE_imageLoad2DArray(inImage, iuvw) imageLoad(inImage, int3(iuvw))
+
+#define OGRE_imageWrite2D1(outImage, iuv, value) imageStore(outImage, int2(iuv), float4(value, 0, 0, 0))
+#define OGRE_imageWrite2D2(outImage, iuv, value) imageStore(outImage, int2(iuv), float4(value, 0, 0))
+#define OGRE_imageWrite2D4(outImage, iuv, value) imageStore(outImage, int2(iuv), value)
+
+#define OGRE_imageLoad3D(inImage, iuv) imageLoad(inImage, int3(iuv))
+
+#define OGRE_imageWrite3D1(outImage, iuv, value) imageStore(outImage, int3(iuv), value)
+#define OGRE_imageWrite3D4(outImage, iuv, value) imageStore(outImage, int3(iuv), value)
+
+#define OGRE_imageWrite2DArray1(outImage, iuvw, value) imageStore(outImage, int3(iuvw), value)
+#define OGRE_imageWrite2DArray4(outImage, iuvw, value) imageStore(outImage, int3(iuvw), value)
+
+//#define sharedOnlyBarrier memoryBarrierShared();barrier();