preserve-emulation/ppsspp
1
0
Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2026-07-11 01:25:07 +02:00
Code Issues Packages Projects Releases Wiki Activity

GPU: Constify VertexDecoder pointers, remove in some cases

Browse Source
This commit is contained in:
Henrik Rydgård
2025-05-14 09:38:31 +02:00
parent 3e0fea509d
commit 8ed898c5e6
20 changed files with 58 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files
GPU/Common/DepthRaster.cpp
+3 -3
View File
@@ -264,7 +264,7 @@ void DepthRaster4Triangles(int stats[3], uint16_t *depthBuf, int stride, DepthSc
}
// This will always run on the main thread. Though, might consider moving the transforms out and just storing verts instead?
void DecodeAndTransformForDepthRaster(float *dest, const float *worldviewproj, const void *vertexData, int indexLowerBound, int indexUpperBound, VertexDecoder *dec, u32 vertTypeID) {
void DecodeAndTransformForDepthRaster(float *dest, const float *worldviewproj, const void *vertexData, int indexLowerBound, int indexUpperBound, const VertexDecoder *dec, u32 vertTypeID) {
// TODO: Ditch skinned and morphed prims for now since we don't have a fast way to skin without running the full decoder.
_dbg_assert_((vertTypeID & (GE_VTYPE_WEIGHT_MASK | GE_VTYPE_MORPHCOUNT_MASK)) == 0);
@@ -298,7 +298,7 @@ void DecodeAndTransformForDepthRaster(float *dest, const float *worldviewproj, c
}
}
void TransformPredecodedForDepthRaster(float *dest, const float *worldviewproj, const void *decodedVertexData, VertexDecoder *dec, int count) {
void TransformPredecodedForDepthRaster(float *dest, const float *worldviewproj, const void *decodedVertexData, const VertexDecoder *dec, int count) {
// TODO: Ditch skinned and morphed prims for now since we don't have a fast way to skin without running the full decoder.
_dbg_assert_((dec->VertexType() & (GE_VTYPE_WEIGHT_MASK | GE_VTYPE_MORPHCOUNT_MASK)) == 0);
@@ -315,7 +315,7 @@ void TransformPredecodedForDepthRaster(float *dest, const float *worldviewproj,
}
}
void ConvertPredecodedThroughForDepthRaster(float *dest, const void *decodedVertexData, VertexDecoder *dec, int count) {
void ConvertPredecodedThroughForDepthRaster(float *dest, const void *decodedVertexData, const VertexDecoder *dec, int count) {
// TODO: Ditch skinned and morphed prims for now since we don't have a fast way to skin without running the full decoder.
_dbg_assert_((dec->VertexType() & (GE_VTYPE_WEIGHT_MASK | GE_VTYPE_MORPHCOUNT_MASK)) == 0);
GPU/Common/DepthRaster.h
+3 -3
View File
@@ -53,7 +53,7 @@ struct TransformedVertex;
int DepthRasterClipIndexedTriangles(int *tx, int *ty, float *tz, const float *transformed, const uint16_t *indexBuffer, const DepthDraw &draw, const DepthScissor scissor);
int DepthRasterClipIndexedRectangles(int *tx, int *ty, float *tz, const float *transformed, const uint16_t *indexBuffer, const DepthDraw &draw, const DepthScissor scissor);
void DecodeAndTransformForDepthRaster(float *dest, const float *worldviewproj, const void *vertexData, int indexLowerBound, int indexUpperBound, VertexDecoder *dec, u32 vertTypeID);
void TransformPredecodedForDepthRaster(float *dest, const float *worldviewproj, const void *decodedVertexData, VertexDecoder *dec, int count);
void ConvertPredecodedThroughForDepthRaster(float *dest, const void *decodedVertexData, VertexDecoder *dec, int count);
void DecodeAndTransformForDepthRaster(float *dest, const float *worldviewproj, const void *vertexData, int indexLowerBound, int indexUpperBound, const VertexDecoder *dec, u32 vertTypeID);
void TransformPredecodedForDepthRaster(float *dest, const float *worldviewproj, const void *decodedVertexData, const VertexDecoder *dec, int count);
void ConvertPredecodedThroughForDepthRaster(float *dest, const void *decodedVertexData, const VertexDecoder *dec, int count);
void DepthRasterScreenVerts(uint16_t *depth, int depthStride, const int *tx, const int *ty, const float *tz, int count, const DepthDraw &draw, const DepthScissor scissor, bool lowQ);
GPU/Common/DrawEngineCommon.cpp
+9 -9
View File
@@ -215,7 +215,7 @@ void DrawEngineCommon::UpdatePlanes() {
// * Compute min/max of the verts, and then compute a bounding sphere and check that against the planes.
// - Less accurate, but..
// - Only requires six plane evaluations then.
bool DrawEngineCommon::TestBoundingBox(const void *vdata, const void *inds, int vertexCount, VertexDecoder *dec, u32 vertType) {
bool DrawEngineCommon::TestBoundingBox(const void *vdata, const void *inds, int vertexCount, const VertexDecoder *dec, u32 vertType) {
// Grab temp buffer space from large offsets in decoded_. Not exactly safe for large draws.
if (vertexCount > 1024) {
return true;
@@ -354,7 +354,7 @@ bool DrawEngineCommon::TestBoundingBox(const void *vdata, const void *inds, int
// corners. That way we can cull more draws quite cheaply.
// We could take the min/max during the regular vertex decode, and just skip the draw call if it's trivially culled.
// This would help games like Midnight Club (that one does a lot of out-of-bounds drawing) immensely.
bool DrawEngineCommon::TestBoundingBoxFast(const void *vdata, int vertexCount, VertexDecoder *dec, u32 vertType) {
bool DrawEngineCommon::TestBoundingBoxFast(const void *vdata, int vertexCount, const VertexDecoder *dec, u32 vertType) {
SimpleVertex *corners = (SimpleVertex *)(decoded_ + 65536 * 12);
float *verts = (float *)(decoded_ + 65536 * 18);
@@ -527,7 +527,7 @@ bool DrawEngineCommon::TestBoundingBoxFast(const void *vdata, int vertexCount, V
// 2D bounding box test against scissor. No indexing yet.
// Only supports non-indexed draws with float positions.
bool DrawEngineCommon::TestBoundingBoxThrough(const void *vdata, int vertexCount, VertexDecoder *dec, u32 vertType) {
bool DrawEngineCommon::TestBoundingBoxThrough(const void *vdata, int vertexCount, const VertexDecoder *dec, u32 vertType) {
// Grab temp buffer space from large offsets in decoded_. Not exactly safe for large draws.
if (vertexCount > 16) {
return true;
@@ -607,7 +607,7 @@ int DrawEngineCommon::ComputeNumVertsToDecode() const {
// Takes a list of consecutive PRIM opcodes, and extends the current draw call to include them.
// This is just a performance optimization.
int DrawEngineCommon::ExtendNonIndexedPrim(const uint32_t *cmd, const uint32_t *stall, VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead, bool isTriangle) {
int DrawEngineCommon::ExtendNonIndexedPrim(const uint32_t *cmd, const uint32_t *stall, const VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead, bool isTriangle) {
const uint32_t *start = cmd;
int prevDrawVerts = numDrawVerts_ - 1;
DeferredVerts &dv = drawVerts_[prevDrawVerts];
@@ -654,7 +654,7 @@ int DrawEngineCommon::ExtendNonIndexedPrim(const uint32_t *cmd, const uint32_t *
return cmd - start;
}
void DrawEngineCommon::SkipPrim(GEPrimitiveType prim, int vertexCount, VertexDecoder *dec, u32 vertTypeID, int *bytesRead) {
void DrawEngineCommon::SkipPrim(GEPrimitiveType prim, int vertexCount, const VertexDecoder *dec, u32 vertTypeID, int *bytesRead) {
if (!indexGen.PrimCompatible(prevPrim_, prim)) {
Flush();
}
@@ -674,7 +674,7 @@ void DrawEngineCommon::SkipPrim(GEPrimitiveType prim, int vertexCount, VertexDec
}
// vertTypeID is the vertex type but with the UVGen mode smashed into the top bits.
bool DrawEngineCommon::SubmitPrim(const void *verts, const void *inds, GEPrimitiveType prim, int vertexCount, VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead) {
bool DrawEngineCommon::SubmitPrim(const void *verts, const void *inds, GEPrimitiveType prim, int vertexCount, const VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead) {
if (!indexGen.PrimCompatible(prevPrim_, prim) || numDrawVerts_ >= MAX_DEFERRED_DRAW_VERTS || numDrawInds_ >= MAX_DEFERRED_DRAW_INDS || vertexCountInDrawCalls_ + vertexCount > VERTEX_BUFFER_MAX) {
Flush();
}
@@ -771,7 +771,7 @@ void DrawEngineCommon::BeginFrame() {
applySkinInDecode_ = g_Config.bSoftwareSkinning;
}
void DrawEngineCommon::DecodeVerts(VertexDecoder *dec, u8 *dest) {
void DrawEngineCommon::DecodeVerts(const VertexDecoder *dec, u8 *dest) {
if (!numDrawVerts_) {
return;
}
@@ -1025,7 +1025,7 @@ bool DrawEngineCommon::CalculateDepthDraw(DepthDraw *draw, GEPrimitiveType prim,
return true;
}
void DrawEngineCommon::DepthRasterSubmitRaw(GEPrimitiveType prim, VertexDecoder *dec, uint32_t vertTypeID, int vertexCount) {
void DrawEngineCommon::DepthRasterSubmitRaw(GEPrimitiveType prim, const VertexDecoder *dec, uint32_t vertTypeID, int vertexCount) {
if (!gstate.isModeClear() && (!gstate.isDepthTestEnabled() || !gstate.isDepthWriteEnabled())) {
return;
}
@@ -1076,7 +1076,7 @@ void DrawEngineCommon::DepthRasterSubmitRaw(GEPrimitiveType prim, VertexDecoder
// FlushQueuedDepth();
}
void DrawEngineCommon::DepthRasterPredecoded(GEPrimitiveType prim, const void *inVerts, int numDecoded, VertexDecoder *dec, int vertexCount) {
void DrawEngineCommon::DepthRasterPredecoded(GEPrimitiveType prim, const void *inVerts, int numDecoded, const VertexDecoder *dec, int vertexCount) {
if (!gstate.isModeClear() && (!gstate.isDepthTestEnabled() || !gstate.isDepthWriteEnabled())) {
return;
}
GPU/Common/DrawEngineCommon.h
+10 -10
View File
@@ -99,12 +99,12 @@ public:
virtual void DispatchSubmitImm(GEPrimitiveType prim, TransformedVertex *buffer, int vertexCount, int cullMode, bool continuation);
bool TestBoundingBox(const void *control_points, const void *inds, int vertexCount, VertexDecoder *dec, u32 vertType);
bool TestBoundingBox(const void *control_points, const void *inds, int vertexCount, const VertexDecoder *dec, u32 vertType);
// This is a less accurate version of TestBoundingBox, but faster. Can have more false positives.
// Doesn't support indexing.
bool TestBoundingBoxFast(const void *control_points, int vertexCount, VertexDecoder *dec, u32 vertType);
bool TestBoundingBoxThrough(const void *vdata, int vertexCount, VertexDecoder *dec, u32 vertType);
bool TestBoundingBoxFast(const void *control_points, int vertexCount, const VertexDecoder *dec, u32 vertType);
bool TestBoundingBoxThrough(const void *vdata, int vertexCount, const VertexDecoder *dec, u32 vertType);
void FlushPartialDecode() {
DecodeVerts(dec_, decoded_);
@@ -116,9 +116,9 @@ public:
}
}
int ExtendNonIndexedPrim(const uint32_t *cmd, const uint32_t *stall, VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead, bool isTriangle);
bool SubmitPrim(const void *verts, const void *inds, GEPrimitiveType prim, int vertexCount, VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead);
void SkipPrim(GEPrimitiveType prim, int vertexCount, VertexDecoder *dec, u32 vertTypeID, int *bytesRead);
int ExtendNonIndexedPrim(const uint32_t *cmd, const uint32_t *stall, const VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead, bool isTriangle);
bool SubmitPrim(const void *verts, const void *inds, GEPrimitiveType prim, int vertexCount, const VertexDecoder *dec, u32 vertTypeID, bool clockwise, int *bytesRead);
void SkipPrim(GEPrimitiveType prim, int vertexCount, const VertexDecoder *dec, u32 vertTypeID, int *bytesRead);
template<class Surface>
void SubmitCurve(const void *control_points, const void *indices, Surface &surface, u32 vertType, int *bytesRead, const char *scope);
@@ -170,7 +170,7 @@ protected:
virtual bool UpdateUseHWTessellation(bool enabled) const { return enabled; }
void UpdatePlanes();
void DecodeVerts(VertexDecoder *dec, u8 *dest);
void DecodeVerts(const VertexDecoder *dec, u8 *dest);
int DecodeInds();
int ComputeNumVertsToDecode() const;
@@ -179,8 +179,8 @@ protected:
void InitDepthRaster();
void ShutdownDepthRaster();
void DepthRasterSubmitRaw(GEPrimitiveType prim, VertexDecoder *dec, uint32_t vertTypeID, int vertexCount);
void DepthRasterPredecoded(GEPrimitiveType prim, const void *inVerts, int numDecoded, VertexDecoder *dec, int vertexCount);
void DepthRasterSubmitRaw(GEPrimitiveType prim, const VertexDecoder *dec, uint32_t vertTypeID, int vertexCount);
void DepthRasterPredecoded(GEPrimitiveType prim, const void *inVerts, int numDecoded, const VertexDecoder *dec, int vertexCount);
bool CalculateDepthDraw(DepthDraw *draw, GEPrimitiveType prim, int vertexCount);
static inline int IndexSize(u32 vtype) {
@@ -319,7 +319,7 @@ protected:
uint32_t drawVertexOffsets_[MAX_DEFERRED_DRAW_VERTS];
DeferredInds drawInds_[MAX_DEFERRED_DRAW_INDS];
VertexDecoder *dec_ = nullptr;
const VertexDecoder *dec_ = nullptr;
u32 lastVType_ = -1; // corresponds to dec_. Could really just pick it out of dec_...
int numDrawVerts_ = 0;
int numDrawInds_ = 0;
GPU/Common/ShaderId.cpp
+1 -4
View File
@@ -66,9 +66,7 @@ std::string VertexShaderDesc(const VShaderID &id) {
return desc.str();
}
void ComputeVertexShaderID(VShaderID *id_out, VertexDecoder *vertexDecoder, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode) {
u32 vertType = vertexDecoder->VertexType();
void ComputeVertexShaderID(VShaderID *id_out, u32 vertType, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode) {
bool isModeThrough = (vertType & GE_VTYPE_THROUGH) != 0;
bool doTexture = gstate.isTextureMapEnabled() && !gstate.isModeClear();
bool doShadeMapping = doTexture && (gstate.getUVGenMode() == GE_TEXMAP_ENVIRONMENT_MAP);
@@ -117,7 +115,6 @@ void ComputeVertexShaderID(VShaderID *id_out, VertexDecoder *vertexDecoder, bool
}
// Bones.
u32 vertType = vertexDecoder->VertexType();
bool enableBones = !useSkinInDecode && vertTypeIsSkinningEnabled(vertType);
id.SetBit(VS_BIT_ENABLE_BONES, enableBones);
if (enableBones) {
GPU/Common/ShaderId.h
+1 -3
View File
@@ -280,9 +280,7 @@ namespace Draw {
class Bugs;
}
class VertexDecoder;
void ComputeVertexShaderID(VShaderID *id, VertexDecoder *vertexDecoder, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode);
void ComputeVertexShaderID(VShaderID *id, u32 vertType, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode);
// Generates a compact string that describes the shader. Useful in a list to get an overview
// of the current flora of shaders.
std::string VertexShaderDesc(const VShaderID &id);
GPU/Common/SoftwareTransformCommon.cpp
+1 -1
View File
@@ -932,7 +932,7 @@ bool SoftwareTransform::ExpandPoints(int vertexCount, int &maxIndex, int vertsSi
// The rest of the transform pipeline like lighting will go as normal, either hardware or software.
// The implementation is initially a bit inefficient but shouldn't be a big deal.
// An intermediate buffer of not-easy-to-predict size is stored at bufPtr.
u32 NormalizeVertices(SimpleVertex *sverts, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, VertexDecoder *dec, u32 vertType) {
u32 NormalizeVertices(SimpleVertex *sverts, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, const VertexDecoder *dec, u32 vertType) {
// First, decode the vertices into a GPU compatible format. This step can be eliminated but will need a separate
// implementation of the vertex decoder.
dec->DecodeVerts(bufPtr, inPtr, &gstate_c.uv, lowerBound, upperBound);
GPU/Common/SoftwareTransformCommon.h
+1 -1
View File
@@ -88,5 +88,5 @@ protected:
};
// Slow. See description in the cpp file.
u32 NormalizeVertices(SimpleVertex *sverts, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, VertexDecoder *dec, u32 vertType);
u32 NormalizeVertices(SimpleVertex *sverts, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, const VertexDecoder *dec, u32 vertType);
bool GetCurrentDrawAsDebugVertices(DrawEngineCommon *drawEngine, int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices);
GPU/D3D11/DrawEngineD3D11.cpp
+3 -3
View File
@@ -284,7 +284,7 @@ void DrawEngineD3D11::Flush() {
D3D11VertexShader *vshader;
D3D11FragmentShader *fshader;
shaderManager_->GetShaders(prim, dec_, &vshader, &fshader, pipelineState_, useHWTransform, useHWTessellation_, decOptions_.expandAllWeightsToFloat, applySkinInDecode_);
shaderManager_->GetShaders(prim, dec_->VertexType(), &vshader, &fshader, pipelineState_, useHWTransform, useHWTessellation_, decOptions_.expandAllWeightsToFloat, applySkinInDecode_);
ComPtr<ID3D11InputLayout> inputLayout;
SetupDecFmtForDraw(vshader, dec_->GetDecVtxFmt(), dec_->VertexType(), &inputLayout);
context_->PSSetShader(fshader->GetShader(), nullptr, 0);
@@ -331,7 +331,7 @@ void DrawEngineD3D11::Flush() {
}
} else {
PROFILE_THIS_SCOPE("soft");
VertexDecoder *swDec = dec_;
const VertexDecoder *swDec = dec_;
if (swDec->nweights != 0) {
u32 withSkinning = lastVType_ | (1 << 26);
if (withSkinning != lastVType_) {
@@ -420,7 +420,7 @@ void DrawEngineD3D11::Flush() {
if (result.action == SW_DRAW_INDEXED) {
D3D11VertexShader *vshader;
D3D11FragmentShader *fshader;
shaderManager_->GetShaders(prim, swDec, &vshader, &fshader, pipelineState_, false, false, decOptions_.expandAllWeightsToFloat, true);
shaderManager_->GetShaders(prim, swDec->VertexType(), &vshader, &fshader, pipelineState_, false, false, decOptions_.expandAllWeightsToFloat, true);
context_->PSSetShader(fshader->GetShader(), nullptr, 0);
context_->VSSetShader(vshader->GetShader(), nullptr, 0);
shaderManager_->UpdateUniforms(framebufferManager_->UseBufferedRendering());
GPU/D3D11/ShaderManagerD3D11.cpp
+3 -2
View File
@@ -27,6 +27,7 @@
#include "Common/CommonTypes.h"
#include "GPU/GPUState.h"
#include "GPU/Common/VertexShaderGenerator.h"
#include "GPU/Common/VertexDecoderCommon.h"
#include "GPU/D3D11/ShaderManagerD3D11.h"
#include "GPU/D3D11/D3D11Util.h"
@@ -162,13 +163,13 @@ void ShaderManagerD3D11::BindUniforms() {
context_->PSSetConstantBuffers(0, 1, ps_cbs);
}
void ShaderManagerD3D11::GetShaders(int prim, VertexDecoder *decoder, D3D11VertexShader **vshader, D3D11FragmentShader **fshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode) {
void ShaderManagerD3D11::GetShaders(int prim, u32 vertexType, D3D11VertexShader **vshader, D3D11FragmentShader **fshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode) {
VShaderID VSID;
FShaderID FSID;
if (gstate_c.IsDirty(DIRTY_VERTEXSHADER_STATE)) {
gstate_c.Clean(DIRTY_VERTEXSHADER_STATE);
ComputeVertexShaderID(&VSID, decoder, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
ComputeVertexShaderID(&VSID, vertexType, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
} else {
VSID = lastVSID_;
}
GPU/D3D11/ShaderManagerD3D11.h
+1 -1
View File
@@ -86,7 +86,7 @@ public:
ShaderManagerD3D11(Draw::DrawContext *draw, ID3D11Device *device, ID3D11DeviceContext *context, D3D_FEATURE_LEVEL featureLevel);
~ShaderManagerD3D11();
void GetShaders(int prim, VertexDecoder *decoder, D3D11VertexShader **vshader, D3D11FragmentShader **fshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode);
void GetShaders(int prim, u32 vertexType, D3D11VertexShader **vshader, D3D11FragmentShader **fshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode);
void ClearShaders() override;
void DirtyLastShader() override;
GPU/Directx9/DrawEngineDX9.cpp
+3 -3
View File
@@ -268,7 +268,7 @@ void DrawEngineDX9::Flush() {
ApplyDrawState(prim);
ApplyDrawStateLate();
VSShader *vshader = shaderManager_->ApplyShader(true, useHWTessellation_, dec_, decOptions_.expandAllWeightsToFloat, applySkinInDecode_, pipelineState_);
VSShader *vshader = shaderManager_->ApplyShader(true, useHWTessellation_, dec_->VertexType(), decOptions_.expandAllWeightsToFloat, applySkinInDecode_, pipelineState_);
ComPtr<IDirect3DVertexDeclaration9> pHardwareVertexDecl;
SetupDecFmtForDraw(dec_->GetDecVtxFmt(), dec_->VertexType(), &pHardwareVertexDecl);
@@ -296,7 +296,7 @@ void DrawEngineDX9::Flush() {
DepthRasterSubmitRaw(prim, dec_, dec_->VertexType(), vertexCount);
}
} else {
VertexDecoder *swDec = dec_;
const VertexDecoder *swDec = dec_;
if (swDec->nweights != 0) {
u32 withSkinning = lastVType_ | (1 << 26);
if (withSkinning != lastVType_) {
@@ -386,7 +386,7 @@ void DrawEngineDX9::Flush() {
ApplyDrawStateLate();
VSShader *vshader = shaderManager_->ApplyShader(false, false, swDec, decOptions_.expandAllWeightsToFloat, true, pipelineState_);
VSShader *vshader = shaderManager_->ApplyShader(false, false, swDec->VertexType(), decOptions_.expandAllWeightsToFloat, true, pipelineState_);
if (result.action == SW_DRAW_INDEXED) {
if (result.setStencil) {
GPU/Directx9/ShaderManagerDX9.cpp
+3 -3
View File
@@ -544,11 +544,11 @@ void ShaderManagerDX9::DirtyLastShader() {
gstate_c.Dirty(DIRTY_ALL_UNIFORMS | DIRTY_VERTEXSHADER_STATE | DIRTY_FRAGMENTSHADER_STATE);
}
VSShader *ShaderManagerDX9::ApplyShader(bool useHWTransform, bool useHWTessellation, VertexDecoder *decoder, bool weightsAsFloat, bool useSkinInDecode, const ComputedPipelineState &pipelineState) {
VSShader *ShaderManagerDX9::ApplyShader(bool useHWTransform, bool useHWTessellation, u32 vertexType, bool weightsAsFloat, bool useSkinInDecode, const ComputedPipelineState &pipelineState) {
VShaderID VSID;
if (gstate_c.IsDirty(DIRTY_VERTEXSHADER_STATE)) {
gstate_c.Clean(DIRTY_VERTEXSHADER_STATE);
ComputeVertexShaderID(&VSID, decoder, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
ComputeVertexShaderID(&VSID, vertexType, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
} else {
VSID = lastVSID_;
}
@@ -598,7 +598,7 @@ VSShader *ShaderManagerDX9::ApplyShader(bool useHWTransform, bool useHWTessellat
}
delete vs;
ComputeVertexShaderID(&VSID, decoder, false, false, weightsAsFloat, useSkinInDecode);
ComputeVertexShaderID(&VSID, vertexType, false, false, weightsAsFloat, useSkinInDecode);
// TODO: Look for existing shader with the appropriate ID, use that instead of generating a new one - however, need to make sure
// that that shader ID is not used when computing the linked shader ID below, because then IDs won't match
GPU/Directx9/ShaderManagerDX9.h
+1 -1
View File
@@ -80,7 +80,7 @@ public:
~ShaderManagerDX9();
void ClearShaders() override;
VSShader *ApplyShader(bool useHWTransform, bool useHWTessellation, VertexDecoder *decoder, bool weightsAsFloat, bool useSkinInDecode, const ComputedPipelineState &pipelineState);
VSShader *ApplyShader(bool useHWTransform, bool useHWTessellation, u32 vertexType, bool weightsAsFloat, bool useSkinInDecode, const ComputedPipelineState &pipelineState);
void DirtyLastShader() override;
int GetNumVertexShaders() const { return (int)vsCache_.size(); }
GPU/GLES/DrawEngineGLES.cpp
+2 -2
View File
@@ -253,7 +253,7 @@ void DrawEngineGLES::Flush() {
GEPrimitiveType prim = prevPrim_;
Shader *vshader = shaderManager_->ApplyVertexShader(CanUseHardwareTransform(prim), useHWTessellation_, dec_, decOptions_.expandAllWeightsToFloat, applySkinInDecode_ || !CanUseHardwareTransform(prim), &vsid);
Shader *vshader = shaderManager_->ApplyVertexShader(CanUseHardwareTransform(prim), useHWTessellation_, dec_->VertexType(), decOptions_.expandAllWeightsToFloat, applySkinInDecode_ || !CanUseHardwareTransform(prim), &vsid);
GLRBuffer *vertexBuffer = nullptr;
GLRBuffer *indexBuffer = nullptr;
@@ -320,7 +320,7 @@ void DrawEngineGLES::Flush() {
}
} else {
PROFILE_THIS_SCOPE("soft");
VertexDecoder *swDec = dec_;
const VertexDecoder *swDec = dec_;
if (swDec->nweights != 0) {
u32 withSkinning = lastVType_ | (1 << 26);
if (withSkinning != lastVType_) {
GPU/GLES/ShaderManagerGLES.cpp
+3 -3
View File
@@ -778,10 +778,10 @@ Shader *ShaderManagerGLES::CompileVertexShader(VShaderID VSID) {
return new Shader(render_, codeBuffer_, desc, params);
}
Shader *ShaderManagerGLES::ApplyVertexShader(bool useHWTransform, bool useHWTessellation, VertexDecoder *decoder, bool weightsAsFloat, bool useSkinInDecode, VShaderID *VSID) {
Shader *ShaderManagerGLES::ApplyVertexShader(bool useHWTransform, bool useHWTessellation, u32 vertexType, bool weightsAsFloat, bool useSkinInDecode, VShaderID *VSID) {
if (gstate_c.IsDirty(DIRTY_VERTEXSHADER_STATE)) {
gstate_c.Clean(DIRTY_VERTEXSHADER_STATE);
ComputeVertexShaderID(VSID, decoder, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
ComputeVertexShaderID(VSID, vertexType, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
} else {
*VSID = lastVSID_;
}
@@ -814,7 +814,7 @@ Shader *ShaderManagerGLES::ApplyVertexShader(bool useHWTransform, bool useHWTess
// Can still work with software transform.
VShaderID vsidTemp;
ComputeVertexShaderID(&vsidTemp, decoder, false, false, weightsAsFloat, true);
ComputeVertexShaderID(&vsidTemp, vertexType, false, false, weightsAsFloat, true);
vs = CompileVertexShader(vsidTemp);
}
GPU/GLES/ShaderManagerGLES.h
+3 -1
View File
@@ -157,6 +157,8 @@ private:
uint64_t uniformMask_;
};
class VertexDecoder;
class ShaderManagerGLES : public ShaderManagerCommon {
public:
ShaderManagerGLES(Draw::DrawContext *draw);
@@ -166,7 +168,7 @@ public:
// This is the old ApplyShader split into two parts, because of annoying information dependencies.
// If you call ApplyVertexShader, you MUST call ApplyFragmentShader soon afterwards.
Shader *ApplyVertexShader(bool useHWTransform, bool useHWTessellation, VertexDecoder *vertexDecoder, bool weightsAsFloat, bool useSkinInDecode, VShaderID *VSID);
Shader *ApplyVertexShader(bool useHWTransform, bool useHWTessellation, u32 vertexType, bool weightsAsFloat, bool useSkinInDecode, VShaderID *VSID);
LinkedShader *ApplyFragmentShader(VShaderID VSID, Shader *vs, const ComputedPipelineState &pipelineState, bool useBufferedRendering);
void DeviceLost() override;
GPU/Vulkan/DrawEngineVulkan.cpp
+3 -3
View File
@@ -293,7 +293,7 @@ void DrawEngineVulkan::Flush() {
VulkanFragmentShader *fshader = nullptr;
VulkanGeometryShader *gshader = nullptr;
shaderManager_->GetShaders(prim, dec_, &vshader, &fshader, &gshader, pipelineState_, true, useHWTessellation_, decOptions_.expandAllWeightsToFloat, applySkinInDecode_);
shaderManager_->GetShaders(prim, dec_->VertexType(), &vshader, &fshader, &gshader, pipelineState_, true, useHWTessellation_, decOptions_.expandAllWeightsToFloat, applySkinInDecode_);
_dbg_assert_msg_(vshader->UseHWTransform(), "Bad vshader");
VulkanPipeline *pipeline = pipelineManager_->GetOrCreatePipeline(renderManager, pipelineLayout_, pipelineKey_, &dec_->decFmt, vshader, fshader, gshader, true, 0, framebufferManager_->GetMSAALevel(), false);
if (!pipeline || !pipeline->pipeline) {
@@ -375,7 +375,7 @@ void DrawEngineVulkan::Flush() {
}
} else {
PROFILE_THIS_SCOPE("soft");
VertexDecoder *swDec = dec_;
const VertexDecoder *swDec = dec_;
if (swDec->nweights != 0) {
u32 withSkinning = lastVType_ | (1 << 26);
if (withSkinning != lastVType_) {
@@ -484,7 +484,7 @@ void DrawEngineVulkan::Flush() {
VulkanFragmentShader *fshader = nullptr;
VulkanGeometryShader *gshader = nullptr;
shaderManager_->GetShaders(prim, swDec, &vshader, &fshader, &gshader, pipelineState_, false, false, decOptions_.expandAllWeightsToFloat, true);
shaderManager_->GetShaders(prim, swDec->VertexType(), &vshader, &fshader, &gshader, pipelineState_, false, false, decOptions_.expandAllWeightsToFloat, true);
_dbg_assert_msg_(!vshader->UseHWTransform(), "Bad vshader");
VulkanPipeline *pipeline = pipelineManager_->GetOrCreatePipeline(renderManager, pipelineLayout_, pipelineKey_, &swDec->decFmt, vshader, fshader, gshader, false, 0, framebufferManager_->GetMSAALevel(), false);
if (!pipeline || !pipeline->pipeline) {
GPU/Vulkan/ShaderManagerVulkan.cpp
+3 -2
View File
@@ -27,6 +27,7 @@
#include "Common/StringUtils.h"
#include "Common/GPU/Vulkan/VulkanContext.h"
#include "Common/Log.h"
#include "Common/TimeUtil.h"
#include "GPU/GPUState.h"
#include "GPU/Common/FragmentShaderGenerator.h"
#include "GPU/Common/VertexShaderGenerator.h"
@@ -279,14 +280,14 @@ uint64_t ShaderManagerVulkan::UpdateUniforms(bool useBufferedRendering) {
return dirty;
}
void ShaderManagerVulkan::GetShaders(int prim, VertexDecoder *decoder, VulkanVertexShader **vshader, VulkanFragmentShader **fshader, VulkanGeometryShader **gshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode) {
void ShaderManagerVulkan::GetShaders(int prim, u32 vertexType, VulkanVertexShader **vshader, VulkanFragmentShader **fshader, VulkanGeometryShader **gshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode) {
VulkanContext *vulkan = (VulkanContext *)draw_->GetNativeObject(Draw::NativeObject::CONTEXT);
VShaderID VSID;
VulkanVertexShader *vs = nullptr;
if (gstate_c.IsDirty(DIRTY_VERTEXSHADER_STATE)) {
gstate_c.Clean(DIRTY_VERTEXSHADER_STATE);
ComputeVertexShaderID(&VSID, decoder, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
ComputeVertexShaderID(&VSID, vertexType, useHWTransform, useHWTessellation, weightsAsFloat, useSkinInDecode);
if (VSID == lastVSID_) {
_dbg_assert_(lastVShader_ != nullptr);
vs = lastVShader_;
GPU/Vulkan/ShaderManagerVulkan.h
+1 -1
View File
@@ -118,7 +118,7 @@ public:
void DeviceLost() override;
void DeviceRestore(Draw::DrawContext *draw) override;
void GetShaders(int prim, VertexDecoder *decoder, VulkanVertexShader **vshader, VulkanFragmentShader **fshader, VulkanGeometryShader **gshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode);
void GetShaders(int prim, u32 vertexType, VulkanVertexShader **vshader, VulkanFragmentShader **fshader, VulkanGeometryShader **gshader, const ComputedPipelineState &pipelineState, bool useHWTransform, bool useHWTessellation, bool weightsAsFloat, bool useSkinInDecode);
void ClearShaders() override;
void DirtyLastShader() override;