Output to an rgba16ui image for the BC1 and BC4 shaders

Increases compatibility with GLSL ES version 310. The ARB_shader_image_load_store extension says, For textures allocated by the GL, an image unit format is compatible with a texture internal format if they match by size. The textures were allocated by the GL implementation, so this format substitution is safe to do.
darksylinc · Oct 4, 2022 · dfc01a9 · dfc01a9
1 parent 4d32a7a
commit dfc01a9
Show file tree

Hide file tree

Showing 4 changed files with 16 additions and 13 deletions.
diff --git a/bin/Data/bc1.glsl b/bin/Data/bc1.glsl
@@ -19,7 +19,7 @@ uniform uint p_numRefinements;
 
 uniform sampler2D srcTex;
 
-layout( rg32ui ) uniform restrict writeonly highp uimage2D dstTexture;
+layout( rgba16ui ) uniform restrict writeonly mediump uimage2D dstTexture;
 
 layout( std430, binding = 1 ) readonly restrict buffer globalBuffer
 {
@@ -513,10 +513,12 @@ void main()
 		mask ^= 0x55555555u;
 	}
 
-	uint2 outputBytes;
-	outputBytes.x = uint( maxEndp16 ) | ( uint( minEndp16 ) << 16u );
-	outputBytes.y = mask;
+	uint4 outputBytes;
+	outputBytes.x = uint( maxEndp16 );
+	outputBytes.y = uint( minEndp16 );
+	outputBytes.z = mask & 0xFFFFu;
+	outputBytes.w = mask >> 16u;
 
 	uint2 dstUV = gl_GlobalInvocationID.xy;
-	imageStore( dstTexture, int2( dstUV ), uint4( outputBytes.xy, 0u, 0u ) );
+	imageStore( dstTexture, int2( dstUV ), outputBytes );
 }
diff --git a/bin/Data/bc4.glsl b/bin/Data/bc4.glsl
@@ -23,7 +23,7 @@ layout( location = 0 ) uniform float2 params;
 
 uniform sampler2D srcTex;
 
-layout( rg32ui ) uniform restrict writeonly highp uimage2D dstTexture;
+layout( rgba16ui ) uniform restrict writeonly mediump uimage2D dstTexture;
 
 layout( local_size_x = 4,  //
 		local_size_y = 4,  //
@@ -142,7 +142,7 @@ void main()
 	if( blockThreadId == 0u )
 	{
 		// Save data
-		uint2 outputBytes;
+		uint4 outputBytes;
 
 		if( p_useSNorm != 0.0f )
 		{
@@ -155,10 +155,11 @@ void main()
 			outputBytes.x = packUnorm4x8(
 				float4( maxVal * ( 1.0f / 255.0f ), minVal * ( 1.0f / 255.0f ), 0.0f, 0.0f ) );
 		}
-		outputBytes.x |= g_mask[maskIdxBase].x;
-		outputBytes.y = g_mask[maskIdxBase].y;
+		outputBytes.y = g_mask[maskIdxBase].x >> 16u;
+		outputBytes.z = g_mask[maskIdxBase].y & 0xFFFFu;
+		outputBytes.w = g_mask[maskIdxBase].y >> 16u;
 
 		uint2 dstUV = gl_GlobalInvocationID.yz;
-		imageStore( dstTexture, int2( dstUV ), uint4( outputBytes.xy, 0u, 0u ) );
+		imageStore( dstTexture, int2( dstUV ), outputBytes );
 	}
 }
diff --git a/src/betsy/EncoderBC1.cpp b/src/betsy/EncoderBC1.cpp
@@ -116,7 +116,7 @@ namespace betsy
 	{
 		bindTexture( 0u, m_srcTexture );
 		bindComputePso( m_bc1Pso );
-		bindUav( 0u, m_bc1TargetRes, PFG_RG32_UINT, ResourceAccess::Write );
+		bindUav( 0u, m_bc1TargetRes, PFG_RGBA16_UINT, ResourceAccess::Write );
 		bindUavBuffer( 1u, m_bc1TablesSsbo, 0u, sizeof( Bc1Tables ) );
 
 		glUniform1ui( 0, 2u );
@@ -128,7 +128,7 @@ namespace betsy
 		{
 			// Compress Alpha too (using BC4)
 			bindComputePso( m_bc4Pso );
-			bindUav( 0u, m_bc4TargetRes, PFG_RG32_UINT, ResourceAccess::Write );
+			bindUav( 0u, m_bc4TargetRes, PFG_RGBA16_UINT, ResourceAccess::Write );
 
 			// p_channelIdx, p_useSNorm
 			glUniform2f( 0, 3.0f, 0.0f );

diff --git a/src/betsy/EncoderBC4.cpp b/src/betsy/EncoderBC4.cpp
@@ -107,7 +107,7 @@ namespace betsy
 		const size_t numChannels = m_bc4TargetRes[1] ? 2u : 1u;
 		for( size_t i = 0u; i < numChannels; ++i )
 		{
-			bindUav( 0u, m_bc4TargetRes[i], PFG_RG32_UINT, ResourceAccess::Write );
+			bindUav( 0u, m_bc4TargetRes[i], PFG_RGBA16_UINT, ResourceAccess::Write );
 
 			// p_channelIdx, p_useSNorm
 			glUniform2f( 0, i == 0u ? 0.0f : 1.0f, m_encodeSNorm ? 1.0f : 0.0f );