mirror/rt64
1
0
Fork 0
mirror of https://github.com/rt64/rt64.git synced 2024-12-26 03:15:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

Move clampWrapMirrorSample and sampleTextureLevel calls into loops to significantly improve ubershader size

Browse Source
This commit is contained in:
Mr-Wiseguy 2024-11-02 19:18:10 -04:00
parent d450596417
commit 2b5b379448
1 changed files with 47 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
src/shaders/TextureSampler.hlsli
View File

@ -118,12 +118,19 @@ float4 sampleTextureLevel(const RDPTile rdpTile, const GPUTile gpuTile, bool fil
tcScale /= mipScale;
int2 texelBaseInt = floor(uvCoord);
float4 sample00 = clampWrapMirrorSample(rdpTile, gpuTile, tcScale, texelBaseInt + int2(0, 0), textureIndex, tlut, canDecodeTMEM, mipLevel, usesHDR);
if (filterBilerp || linearFiltering) {
bool filtering = or(filterBilerp, linearFiltering);
int numSamples = select_uint(filtering, 4, 1);
float4 samples[4];
[unroll]
for (int i = 0; i < numSamples; i++) {
samples[i] = clampWrapMirrorSample(rdpTile, gpuTile, tcScale, texelBaseInt + int2(i >> 1, i & 1), textureIndex, tlut, canDecodeTMEM, mipLevel, usesHDR);
}
float4 sample00 = samples[0];
if (filtering) {
float2 fracPart = uvCoord - texelBaseInt;
float4 sample01 = clampWrapMirrorSample(rdpTile, gpuTile, tcScale, texelBaseInt + int2(0, 1), textureIndex, tlut, canDecodeTMEM, mipLevel, usesHDR);
float4 sample10 = clampWrapMirrorSample(rdpTile, gpuTile, tcScale, texelBaseInt + int2(1, 0), textureIndex, tlut, canDecodeTMEM, mipLevel, usesHDR);
float4 sample11 = clampWrapMirrorSample(rdpTile, gpuTile, tcScale, texelBaseInt + int2(1, 1), textureIndex, tlut, canDecodeTMEM, mipLevel, usesHDR);
float4 sample01 = samples[1];
float4 sample10 = samples[2];
float4 sample11 = samples[3];
if (linearFiltering) {
return lerp(lerp(sample00, sample10, fracPart.x), lerp(sample01, sample11, fracPart.x), fracPart.y);
}
@ -187,27 +194,37 @@ float4 sampleTexture(OtherMode otherMode, RenderFlags renderFlags, float2 inputU
const bool canDecodeTMEM = renderFlagCanDecodeTMEM(renderFlags);
const bool usesHDR = renderFlagUsesHDR(renderFlags);
const uint nativeSampler = rdpTile.nativeSampler;
if (!gpuTileFlagHasMipmaps(gpuTile.flags)) {
return sampleTextureLevel(rdpTile, gpuTile, filterBilerp, filterAverage, linearFiltering, uvCoord, textureIndex, tlut, canDecodeTMEM, 0, usesHDR);
const bool flagHasMipmaps = gpuTileFlagHasMipmaps(gpuTile.flags);
uint numRDPSamples = 0;
uint RDPMipLevels[2];
RDPMipLevels[0] = 0;
RDPMipLevels[1] = 0;
float mip;
// Determine the RDP sample count and mip levels.
if (!flagHasMipmaps) {
numRDPSamples = 1;
}
else {
// Retrieve the dimensions of the texture for either type of sampler.
Texture2D texture = gTextures[NonUniformResourceIndex(textureIndex)];
uint textureWidth, textureHeight, textureLevels;
texture.GetDimensions(0, textureWidth, textureHeight, textureLevels);
if (nativeSampler == NATIVE_SAMPLER_NONE) {
float2 ddxUVxScaled = ddxUVx * gpuTile.tcScale;
float2 ddxUVyScaled = ddyUVy * gpuTile.tcScale;
float ddMax = max(dot(ddxUVxScaled, ddxUVxScaled), dot(ddxUVyScaled, ddxUVyScaled));
float mipBias = -0.5f;
float mip = 0.5 * log2(ddMax) + mipBias;
mip = 0.5 * log2(ddMax) + mipBias;
float maxMip = float(textureLevels - 1);
float4 hiLevel = sampleTextureLevel(rdpTile, gpuTile, filterBilerp, filterAverage, linearFiltering, uvCoord, textureIndex, tlut, canDecodeTMEM, min(floor(mip), maxMip), usesHDR);
float4 loLevel = sampleTextureLevel(rdpTile, gpuTile, filterBilerp, filterAverage, linearFiltering, uvCoord, textureIndex, tlut, canDecodeTMEM, min(floor(mip) + 1, maxMip), usesHDR);
return lerp(hiLevel, loLevel, frac(mip));
RDPMipLevels[0] = min(floor(mip), maxMip);
RDPMipLevels[1] = min(floor(mip) + 1, maxMip);
numRDPSamples = 2;
}
else {
// Native sampling, just return the native sampling result directly.
numRDPSamples = 0;
// Must normalize the texture coordinate and the derivatives.
float2 nativeUVCoord = uvCoord / float2(textureWidth, textureHeight);
float2 originalSize = float2(textureWidth, textureHeight) / gpuTile.tcScale;
@ -238,4 +255,22 @@ float4 sampleTexture(OtherMode otherMode, RenderFlags renderFlags, float2 inputU
}
}
}
float4 textureSamples[2];
// Perform the RDP sampling.
[unroll]
for (uint i = 0; i < numRDPSamples; i++) {
textureSamples[i] = sampleTextureLevel(rdpTile, gpuTile, filterBilerp, filterAverage, linearFiltering, uvCoord, textureIndex, tlut, canDecodeTMEM, RDPMipLevels[i], usesHDR);
}
// Compute the result from the RDP samples.
if (!flagHasMipmaps) {
return textureSamples[0];
}
else {
float4 hiLevel = textureSamples[0];
float4 loLevel = textureSamples[1];
return lerp(hiLevel, loLevel, frac(mip));
}
}