The-Homebrew-Cloud/Cemu
2
0
Fork 0
mirror of https://github.com/cemu-project/Cemu.git synced 2024-11-22 17:19:18 +01:00
Code Issues Packages Projects Releases Wiki Activity

Latte: Small refactor for GLSL texture coord handling

Browse Source 
Also adds support for 2D textures coordinates with source as 0.0 or 1.0 literals instead of GPRs. Seen in shaders generated by CafeGLSL
This commit is contained in:
Exzap 2023-12-06 02:29:56 +01:00
parent e7fa8ec0c6
commit dee764473d
1 changed files with 46 additions and 62 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

110
src/Cafe/HW/Latte/LegacyShaderDecompiler/LatteDecompilerEmitGLSL.cpp
View File

@ -507,7 +507,7 @@ void _emitRegisterAccessCode(LatteDecompilerShaderContext* shaderContext, sint32
{ {
_emitTypeConversionPrefix(shaderContext, registerElementDataType, dataType); _emitTypeConversionPrefix(shaderContext, registerElementDataType, dataType);
} }
if(shaderContext->typeTracker.useArrayGPRs ) if (shaderContext->typeTracker.useArrayGPRs)
src->add("R"); src->add("R");
else else
src->addFmt("R{}", gprIndex); src->addFmt("R{}", gprIndex);
@ -540,6 +540,26 @@ void _emitRegisterAccessCode(LatteDecompilerShaderContext* shaderContext, sint32
_emitTypeConversionSuffix(shaderContext, registerElementDataType, dataType); _emitTypeConversionSuffix(shaderContext, registerElementDataType, dataType);
} }
// optimized variant of _emitRegisterAccessCode for raw one channel reads
void _emitRegisterChannelAccessCode(LatteDecompilerShaderContext* shaderContext, sint32 gprIndex, sint32 channel, sint32 dataType)
{
cemu_assert_debug(gprIndex >= 0 && gprIndex <= 127);
cemu_assert_debug(channel >= 0 && channel < 4);
StringBuf* src = shaderContext->shaderSource;
sint32 registerElementDataType = shaderContext->typeTracker.defaultDataType;
_emitTypeConversionPrefix(shaderContext, registerElementDataType, dataType);
if (shaderContext->typeTracker.useArrayGPRs)
src->add("R");
else
src->addFmt("R{}", gprIndex);
_appendRegisterTypeSuffix(src, registerElementDataType);
if (shaderContext->typeTracker.useArrayGPRs)
src->addFmt("[{}]", gprIndex);
src->add(".");
src->add(_getElementStrByIndex(channel));
_emitTypeConversionSuffix(shaderContext, registerElementDataType, dataType);
}
void _emitALURegisterInputAccessCode(LatteDecompilerShaderContext* shaderContext, LatteDecompilerALUInstruction* aluInstruction, sint32 operandIndex) void _emitALURegisterInputAccessCode(LatteDecompilerShaderContext* shaderContext, LatteDecompilerALUInstruction* aluInstruction, sint32 operandIndex)
{ {
StringBuf* src = shaderContext->shaderSource; StringBuf* src = shaderContext->shaderSource;
@ -2129,63 +2149,31 @@ void _emitALUClauseCode(LatteDecompilerShaderContext* shaderContext, LatteDecomp
/* /*
* Emits code to access one component (xyzw) of the texture coordinate input vector * Emits code to access one component (xyzw) of the texture coordinate input vector
*/ */
void _emitTEXSampleCoordInputComponent(LatteDecompilerShaderContext* shaderContext, LatteDecompilerTEXInstruction* texInstruction, sint32 componentIndex, sint32 varType) void _emitTEXSampleCoordInputComponent(LatteDecompilerShaderContext* shaderContext, LatteDecompilerTEXInstruction* texInstruction, sint32 componentIndex, sint32 interpretSrcAsType)
{ {
cemu_assert(componentIndex >= 0 && componentIndex < 4);
cemu_assert_debug(interpretSrcAsType == LATTE_DECOMPILER_DTYPE_SIGNED_INT || interpretSrcAsType == LATTE_DECOMPILER_DTYPE_FLOAT);
StringBuf* src = shaderContext->shaderSource; StringBuf* src = shaderContext->shaderSource;
if( componentIndex >= 4 )
{
debugBreakpoint();
return;
}
sint32 elementSel = texInstruction->textureFetch.srcSel[componentIndex]; sint32 elementSel = texInstruction->textureFetch.srcSel[componentIndex];
const char* resultElemTable[4] = {"x","y","z","w"};
if( varType == LATTE_DECOMPILER_DTYPE_SIGNED_INT )
{
if (elementSel < 4) if (elementSel < 4)
{ {
if (shaderContext->typeTracker.defaultDataType == LATTE_DECOMPILER_DTYPE_SIGNED_INT) _emitRegisterChannelAccessCode(shaderContext, texInstruction->srcGpr, elementSel, interpretSrcAsType);
src->addFmt("{}.{}", _getRegisterVarName(shaderContext, texInstruction->srcGpr), resultElemTable[elementSel]); return;
else if (shaderContext->typeTracker.defaultDataType == LATTE_DECOMPILER_DTYPE_FLOAT) }
src->addFmt("floatBitsToInt({}.{})", _getRegisterVarName(shaderContext, texInstruction->srcGpr), resultElemTable[elementSel]); const char* resultElemTable[4] = {"x","y","z","w"};
else if(interpretSrcAsType == LATTE_DECOMPILER_DTYPE_SIGNED_INT )
{ {
cemu_assert_unimplemented(); if( elementSel == 4 )
}
}
else if( elementSel == 4 )
src->add("floatBitsToInt(0.0)"); src->add("floatBitsToInt(0.0)");
else if( elementSel == 5 ) else if( elementSel == 5 )
src->add("floatBitsToInt(1.0)"); src->add("floatBitsToInt(1.0)");
else
{
cemu_assert_unimplemented();
} }
} else if(interpretSrcAsType == LATTE_DECOMPILER_DTYPE_FLOAT )
else if( varType == LATTE_DECOMPILER_DTYPE_FLOAT )
{ {
if (elementSel < 4) if( elementSel == 4 )
{ src->add("0.0");
if (shaderContext->typeTracker.defaultDataType == LATTE_DECOMPILER_DTYPE_SIGNED_INT)
src->addFmt("intBitsToFloat({}.{})", _getRegisterVarName(shaderContext, texInstruction->srcGpr), resultElemTable[elementSel]);
else if (shaderContext->typeTracker.defaultDataType == LATTE_DECOMPILER_DTYPE_FLOAT)
src->addFmt("{}.{}", _getRegisterVarName(shaderContext, texInstruction->srcGpr), resultElemTable[elementSel]);
else
{
cemu_assert_unimplemented();
}
}
else if( elementSel == 4 )
src->addFmt("0.0");
else if( elementSel == 5 ) else if( elementSel == 5 )
src->addFmt("1.0"); src->add("1.0");
else
{
cemu_assert_unimplemented();
}
}
else
{
cemu_assert_unimplemented();
} }
} }
@ -2430,10 +2418,6 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
cemu_assert_unimplemented(); cemu_assert_unimplemented();
src->add("texture("); src->add("texture(");
} }
if( texInstruction->textureFetch.srcSel[0] >= 4 )
cemu_assert_unimplemented();
if( texInstruction->textureFetch.srcSel[1] >= 4 )
cemu_assert_unimplemented();
src->addFmt("{}{}, ", _getTextureUnitVariablePrefixName(shaderContext->shader->shaderType), texInstruction->textureFetch.textureIndex); src->addFmt("{}{}, ", _getTextureUnitVariablePrefixName(shaderContext->shader->shaderType), texInstruction->textureFetch.textureIndex);
// for textureGather() add shift (todo: depends on rounding mode set in sampler registers?) // for textureGather() add shift (todo: depends on rounding mode set in sampler registers?)
@ -2455,7 +2439,7 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
} }
} }
const sint32 texCoordDataType = (texOpcode == GPU7_TEX_INST_LD) ? LATTE_DECOMPILER_DTYPE_SIGNED_INT : LATTE_DECOMPILER_DTYPE_FLOAT;
if(useTexelCoordinates) if(useTexelCoordinates)
{ {
// handle integer coordinates for texelFetch // handle integer coordinates for texelFetch
@ -2463,9 +2447,9 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
{ {
src->add("ivec2("); src->add("ivec2(");
src->add("vec2("); src->add("vec2(");
_emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 0, (texOpcode == GPU7_TEX_INST_LD) ? LATTE_DECOMPILER_DTYPE_SIGNED_INT : LATTE_DECOMPILER_DTYPE_FLOAT); _emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 0, texCoordDataType);
src->addFmt(", "); src->addFmt(", ");
_emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 1, (texOpcode == GPU7_TEX_INST_LD) ? LATTE_DECOMPILER_DTYPE_SIGNED_INT : LATTE_DECOMPILER_DTYPE_FLOAT); _emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 1, texCoordDataType);
src->addFmt(")*uf_tex{}Scale", texInstruction->textureFetch.textureIndex); // close vec2 and scale src->addFmt(")*uf_tex{}Scale", texInstruction->textureFetch.textureIndex); // close vec2 and scale
@ -2485,7 +2469,7 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
else else
cemu_assert_debug(false); cemu_assert_debug(false);
} }
else else /* useTexelCoordinates == false */
{ {
// float coordinates // float coordinates
if ( (texOpcode == GPU7_TEX_INST_SAMPLE_C || texOpcode == GPU7_TEX_INST_SAMPLE_C_L || texOpcode == GPU7_TEX_INST_SAMPLE_C_LZ) ) if ( (texOpcode == GPU7_TEX_INST_SAMPLE_C || texOpcode == GPU7_TEX_INST_SAMPLE_C_L || texOpcode == GPU7_TEX_INST_SAMPLE_C_LZ) )
@ -2549,10 +2533,8 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
else if( texDim == Latte::E_DIM::DIM_CUBEMAP ) else if( texDim == Latte::E_DIM::DIM_CUBEMAP )
{ {
// 2 coords + faceId // 2 coords + faceId
if( texInstruction->textureFetch.srcSel[0] >= 4 || texInstruction->textureFetch.srcSel[1] >= 4 ) cemu_assert_debug(texInstruction->textureFetch.srcSel[0] < 4);
{ cemu_assert_debug(texInstruction->textureFetch.srcSel[1] < 4);
debugBreakpoint();
}
src->add("vec4("); src->add("vec4(");
src->addFmt("redcCUBEReverse({},", _getTexGPRAccess(shaderContext, texInstruction->srcGpr, LATTE_DECOMPILER_DTYPE_FLOAT, texInstruction->textureFetch.srcSel[0], texInstruction->textureFetch.srcSel[1], -1, -1, tempBuffer0)); src->addFmt("redcCUBEReverse({},", _getTexGPRAccess(shaderContext, texInstruction->srcGpr, LATTE_DECOMPILER_DTYPE_FLOAT, texInstruction->textureFetch.srcSel[0], texInstruction->textureFetch.srcSel[1], -1, -1, tempBuffer0));
_emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 2, LATTE_DECOMPILER_DTYPE_SIGNED_INT); _emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 2, LATTE_DECOMPILER_DTYPE_SIGNED_INT);
@ -2567,8 +2549,11 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
else else
{ {
// 2 coords // 2 coords
src->add(_getTexGPRAccess(shaderContext, texInstruction->srcGpr, LATTE_DECOMPILER_DTYPE_FLOAT, texInstruction->textureFetch.srcSel[0], texInstruction->textureFetch.srcSel[1], -1, -1, tempBuffer0)); src->add("vec2(");
_emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 0, LATTE_DECOMPILER_DTYPE_FLOAT);
src->add(",");
_emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 1, LATTE_DECOMPILER_DTYPE_FLOAT);
src->add(")");
// avoid truncate to effectively round downwards on texel edges // avoid truncate to effectively round downwards on texel edges
if (ActiveSettings::ForceSamplerRoundToPrecision()) if (ActiveSettings::ForceSamplerRoundToPrecision())
src->addFmt("+ vec2(1.0)/vec2(textureSize({}{}, 0))/512.0", _getTextureUnitVariablePrefixName(shaderContext->shader->shaderType), texInstruction->textureFetch.textureIndex); src->addFmt("+ vec2(1.0)/vec2(textureSize({}{}, 0))/512.0", _getTextureUnitVariablePrefixName(shaderContext->shader->shaderType), texInstruction->textureFetch.textureIndex);
@ -2576,9 +2561,8 @@ void _emitTEXSampleTextureCode(LatteDecompilerShaderContext* shaderContext, Latt
// lod or lod bias parameter // lod or lod bias parameter
if( texOpcode == GPU7_TEX_INST_SAMPLE_L || texOpcode == GPU7_TEX_INST_SAMPLE_LB || texOpcode == GPU7_TEX_INST_SAMPLE_C_L) if( texOpcode == GPU7_TEX_INST_SAMPLE_L || texOpcode == GPU7_TEX_INST_SAMPLE_LB || texOpcode == GPU7_TEX_INST_SAMPLE_C_L)
{ {
if( texInstruction->textureFetch.srcSel[3] >= 4 ) src->add(",");
debugBreakpoint(); _emitTEXSampleCoordInputComponent(shaderContext, texInstruction, 3, LATTE_DECOMPILER_DTYPE_FLOAT);
src->addFmt(",{}", _getTexGPRAccess(shaderContext, texInstruction->srcGpr, LATTE_DECOMPILER_DTYPE_FLOAT, texInstruction->textureFetch.srcSel[3], -1, -1, -1, tempBuffer0));
} }
else if( texOpcode == GPU7_TEX_INST_SAMPLE_LZ || texOpcode == GPU7_TEX_INST_SAMPLE_C_LZ ) else if( texOpcode == GPU7_TEX_INST_SAMPLE_LZ || texOpcode == GPU7_TEX_INST_SAMPLE_C_LZ )
{ {