blob: fb61346ba717d9e226ecaaf2c74d9202ec992767 [file] [log] [blame]
/*
* Copyright 2018 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "src/gpu/mtl/GrMtlPipelineStateBuilder.h"
#include "include/gpu/GrDirectContext.h"
#include "src/core/SkReadBuffer.h"
#include "src/core/SkTraceEvent.h"
#include "src/gpu/GrAutoLocaleSetter.h"
#include "src/gpu/GrDirectContextPriv.h"
#include "src/gpu/GrPersistentCacheUtils.h"
#include "src/gpu/GrRenderTarget.h"
#include "src/gpu/GrShaderUtils.h"
#include "src/gpu/mtl/GrMtlGpu.h"
#include "src/gpu/mtl/GrMtlPipelineState.h"
#include "src/gpu/mtl/GrMtlUtil.h"
#import <simd/simd.h>
#if !__has_feature(objc_arc)
#error This file must be compiled with Arc. Use -fobjc-arc flag
#endif
GrMtlPipelineState* GrMtlPipelineStateBuilder::CreatePipelineState(
GrMtlGpu* gpu,
GrRenderTarget* renderTarget,
const GrProgramDesc& desc,
const GrProgramInfo& programInfo) {
GrAutoLocaleSetter als("C");
GrMtlPipelineStateBuilder builder(gpu, renderTarget, desc, programInfo);
if (!builder.emitAndInstallProcs()) {
return nullptr;
}
return builder.finalize(renderTarget, desc, programInfo);
}
GrMtlPipelineStateBuilder::GrMtlPipelineStateBuilder(GrMtlGpu* gpu,
GrRenderTarget* renderTarget,
const GrProgramDesc& desc,
const GrProgramInfo& programInfo)
: INHERITED(renderTarget, desc, programInfo)
, fGpu(gpu)
, fUniformHandler(this)
, fVaryingHandler(this) {
}
const GrCaps* GrMtlPipelineStateBuilder::caps() const {
return fGpu->caps();
}
void GrMtlPipelineStateBuilder::finalizeFragmentOutputColor(GrShaderVar& outputColor) {
outputColor.addLayoutQualifier("location = 0, index = 0");
}
void GrMtlPipelineStateBuilder::finalizeFragmentSecondaryColor(GrShaderVar& outputColor) {
outputColor.addLayoutQualifier("location = 0, index = 1");
}
static constexpr SkFourByteTag kMSL_Tag = SkSetFourByteTag('M', 'S', 'L', ' ');
static constexpr SkFourByteTag kSKSL_Tag = SkSetFourByteTag('S', 'K', 'S', 'L');
bool GrMtlPipelineStateBuilder::loadShadersFromCache(SkReadBuffer* cached,
__strong id<MTLLibrary> outLibraries[]) {
SkSL::String shaders[kGrShaderTypeCount];
SkSL::Program::Inputs inputs[kGrShaderTypeCount];
if (!GrPersistentCacheUtils::UnpackCachedShaders(cached, shaders, inputs, kGrShaderTypeCount)) {
return false;
}
outLibraries[kVertex_GrShaderType] = this->compileMtlShaderLibrary(
shaders[kVertex_GrShaderType],
inputs[kVertex_GrShaderType]);
outLibraries[kFragment_GrShaderType] = this->compileMtlShaderLibrary(
shaders[kFragment_GrShaderType],
inputs[kFragment_GrShaderType]);
return outLibraries[kVertex_GrShaderType] &&
outLibraries[kFragment_GrShaderType] &&
shaders[kGeometry_GrShaderType].empty(); // Geometry shaders are not supported
}
void GrMtlPipelineStateBuilder::storeShadersInCache(const SkSL::String shaders[],
const SkSL::Program::Inputs inputs[],
bool isSkSL) {
// Here we shear off the Mtl-specific portion of the Desc in order to create the
// persistent key. This is because Mtl only caches the MSL code, not the fully compiled
// program, and that only depends on the base GrProgramDesc data.
sk_sp<SkData> key = SkData::MakeWithoutCopy(this->desc().asKey(),
this->desc().initialKeyLength());
sk_sp<SkData> data = GrPersistentCacheUtils::PackCachedShaders(isSkSL ? kSKSL_Tag : kMSL_Tag,
shaders,
inputs, kGrShaderTypeCount);
fGpu->getContext()->priv().getPersistentCache()->store(*key, *data);
}
id<MTLLibrary> GrMtlPipelineStateBuilder::generateMtlShaderLibrary(
const SkSL::String& shader,
SkSL::Program::Kind kind,
const SkSL::Program::Settings& settings,
SkSL::String* msl,
SkSL::Program::Inputs* inputs) {
id<MTLLibrary> shaderLibrary = GrGenerateMtlShaderLibrary(fGpu, shader,
kind, settings, msl, inputs);
if (shaderLibrary != nil && inputs->fRTHeight) {
this->addRTHeightUniform(SKSL_RTHEIGHT_NAME);
}
return shaderLibrary;
}
id<MTLLibrary> GrMtlPipelineStateBuilder::compileMtlShaderLibrary(const SkSL::String& shader,
SkSL::Program::Inputs inputs) {
id<MTLLibrary> shaderLibrary = GrCompileMtlShaderLibrary(fGpu, shader);
if (shaderLibrary != nil && inputs.fRTHeight) {
this->addRTHeightUniform(SKSL_RTHEIGHT_NAME);
}
return shaderLibrary;
}
static inline MTLVertexFormat attribute_type_to_mtlformat(GrVertexAttribType type) {
switch (type) {
case kFloat_GrVertexAttribType:
return MTLVertexFormatFloat;
case kFloat2_GrVertexAttribType:
return MTLVertexFormatFloat2;
case kFloat3_GrVertexAttribType:
return MTLVertexFormatFloat3;
case kFloat4_GrVertexAttribType:
return MTLVertexFormatFloat4;
case kHalf_GrVertexAttribType:
if (@available(macOS 10.13, iOS 11.0, *)) {
return MTLVertexFormatHalf;
} else {
return MTLVertexFormatInvalid;
}
case kHalf2_GrVertexAttribType:
return MTLVertexFormatHalf2;
case kHalf4_GrVertexAttribType:
return MTLVertexFormatHalf4;
case kInt2_GrVertexAttribType:
return MTLVertexFormatInt2;
case kInt3_GrVertexAttribType:
return MTLVertexFormatInt3;
case kInt4_GrVertexAttribType:
return MTLVertexFormatInt4;
case kByte_GrVertexAttribType:
if (@available(macOS 10.13, iOS 11.0, *)) {
return MTLVertexFormatChar;
} else {
return MTLVertexFormatInvalid;
}
case kByte2_GrVertexAttribType:
return MTLVertexFormatChar2;
case kByte4_GrVertexAttribType:
return MTLVertexFormatChar4;
case kUByte_GrVertexAttribType:
if (@available(macOS 10.13, iOS 11.0, *)) {
return MTLVertexFormatUChar;
} else {
return MTLVertexFormatInvalid;
}
case kUByte2_GrVertexAttribType:
return MTLVertexFormatUChar2;
case kUByte4_GrVertexAttribType:
return MTLVertexFormatUChar4;
case kUByte_norm_GrVertexAttribType:
if (@available(macOS 10.13, iOS 11.0, *)) {
return MTLVertexFormatUCharNormalized;
} else {
return MTLVertexFormatInvalid;
}
case kUByte4_norm_GrVertexAttribType:
return MTLVertexFormatUChar4Normalized;
case kShort2_GrVertexAttribType:
return MTLVertexFormatShort2;
case kShort4_GrVertexAttribType:
return MTLVertexFormatShort4;
case kUShort2_GrVertexAttribType:
return MTLVertexFormatUShort2;
case kUShort2_norm_GrVertexAttribType:
return MTLVertexFormatUShort2Normalized;
case kInt_GrVertexAttribType:
return MTLVertexFormatInt;
case kUint_GrVertexAttribType:
return MTLVertexFormatUInt;
case kUShort_norm_GrVertexAttribType:
if (@available(macOS 10.13, iOS 11.0, *)) {
return MTLVertexFormatUShortNormalized;
} else {
return MTLVertexFormatInvalid;
}
case kUShort4_norm_GrVertexAttribType:
return MTLVertexFormatUShort4Normalized;
}
SK_ABORT("Unknown vertex attribute type");
}
static MTLVertexDescriptor* create_vertex_descriptor(const GrPrimitiveProcessor& primProc) {
uint32_t vertexBinding = 0, instanceBinding = 0;
int nextBinding = GrMtlUniformHandler::kLastUniformBinding + 1;
if (primProc.hasVertexAttributes()) {
vertexBinding = nextBinding++;
}
if (primProc.hasInstanceAttributes()) {
instanceBinding = nextBinding;
}
auto vertexDescriptor = [[MTLVertexDescriptor alloc] init];
int attributeIndex = 0;
int vertexAttributeCount = primProc.numVertexAttributes();
size_t vertexAttributeOffset = 0;
for (const auto& attribute : primProc.vertexAttributes()) {
MTLVertexAttributeDescriptor* mtlAttribute = vertexDescriptor.attributes[attributeIndex];
mtlAttribute.format = attribute_type_to_mtlformat(attribute.cpuType());
SkASSERT(MTLVertexFormatInvalid != mtlAttribute.format);
mtlAttribute.offset = vertexAttributeOffset;
mtlAttribute.bufferIndex = vertexBinding;
vertexAttributeOffset += attribute.sizeAlign4();
attributeIndex++;
}
SkASSERT(vertexAttributeOffset == primProc.vertexStride());
if (vertexAttributeCount) {
MTLVertexBufferLayoutDescriptor* vertexBufferLayout =
vertexDescriptor.layouts[vertexBinding];
vertexBufferLayout.stepFunction = MTLVertexStepFunctionPerVertex;
vertexBufferLayout.stepRate = 1;
vertexBufferLayout.stride = vertexAttributeOffset;
}
int instanceAttributeCount = primProc.numInstanceAttributes();
size_t instanceAttributeOffset = 0;
for (const auto& attribute : primProc.instanceAttributes()) {
MTLVertexAttributeDescriptor* mtlAttribute = vertexDescriptor.attributes[attributeIndex];
mtlAttribute.format = attribute_type_to_mtlformat(attribute.cpuType());
mtlAttribute.offset = instanceAttributeOffset;
mtlAttribute.bufferIndex = instanceBinding;
instanceAttributeOffset += attribute.sizeAlign4();
attributeIndex++;
}
SkASSERT(instanceAttributeOffset == primProc.instanceStride());
if (instanceAttributeCount) {
MTLVertexBufferLayoutDescriptor* instanceBufferLayout =
vertexDescriptor.layouts[instanceBinding];
instanceBufferLayout.stepFunction = MTLVertexStepFunctionPerInstance;
instanceBufferLayout.stepRate = 1;
instanceBufferLayout.stride = instanceAttributeOffset;
}
return vertexDescriptor;
}
static MTLBlendFactor blend_coeff_to_mtl_blend(GrBlendCoeff coeff) {
switch (coeff) {
case kZero_GrBlendCoeff:
return MTLBlendFactorZero;
case kOne_GrBlendCoeff:
return MTLBlendFactorOne;
case kSC_GrBlendCoeff:
return MTLBlendFactorSourceColor;
case kISC_GrBlendCoeff:
return MTLBlendFactorOneMinusSourceColor;
case kDC_GrBlendCoeff:
return MTLBlendFactorDestinationColor;
case kIDC_GrBlendCoeff:
return MTLBlendFactorOneMinusDestinationColor;
case kSA_GrBlendCoeff:
return MTLBlendFactorSourceAlpha;
case kISA_GrBlendCoeff:
return MTLBlendFactorOneMinusSourceAlpha;
case kDA_GrBlendCoeff:
return MTLBlendFactorDestinationAlpha;
case kIDA_GrBlendCoeff:
return MTLBlendFactorOneMinusDestinationAlpha;
case kConstC_GrBlendCoeff:
return MTLBlendFactorBlendColor;
case kIConstC_GrBlendCoeff:
return MTLBlendFactorOneMinusBlendColor;
case kS2C_GrBlendCoeff:
if (@available(macOS 10.12, iOS 11.0, *)) {
return MTLBlendFactorSource1Color;
} else {
return MTLBlendFactorZero;
}
case kIS2C_GrBlendCoeff:
if (@available(macOS 10.12, iOS 11.0, *)) {
return MTLBlendFactorOneMinusSource1Color;
} else {
return MTLBlendFactorZero;
}
case kS2A_GrBlendCoeff:
if (@available(macOS 10.12, iOS 11.0, *)) {
return MTLBlendFactorSource1Alpha;
} else {
return MTLBlendFactorZero;
}
case kIS2A_GrBlendCoeff:
if (@available(macOS 10.12, iOS 11.0, *)) {
return MTLBlendFactorOneMinusSource1Alpha;
} else {
return MTLBlendFactorZero;
}
case kIllegal_GrBlendCoeff:
return MTLBlendFactorZero;
}
SK_ABORT("Unknown blend coefficient");
}
static MTLBlendOperation blend_equation_to_mtl_blend_op(GrBlendEquation equation) {
static const MTLBlendOperation gTable[] = {
MTLBlendOperationAdd, // kAdd_GrBlendEquation
MTLBlendOperationSubtract, // kSubtract_GrBlendEquation
MTLBlendOperationReverseSubtract, // kReverseSubtract_GrBlendEquation
};
static_assert(SK_ARRAY_COUNT(gTable) == kFirstAdvancedGrBlendEquation);
static_assert(0 == kAdd_GrBlendEquation);
static_assert(1 == kSubtract_GrBlendEquation);
static_assert(2 == kReverseSubtract_GrBlendEquation);
SkASSERT((unsigned)equation < kGrBlendEquationCnt);
return gTable[equation];
}
static MTLRenderPipelineColorAttachmentDescriptor* create_color_attachment(
MTLPixelFormat format, const GrPipeline& pipeline) {
auto mtlColorAttachment = [[MTLRenderPipelineColorAttachmentDescriptor alloc] init];
// pixel format
mtlColorAttachment.pixelFormat = format;
// blending
const GrXferProcessor::BlendInfo& blendInfo = pipeline.getXferProcessor().getBlendInfo();
GrBlendEquation equation = blendInfo.fEquation;
GrBlendCoeff srcCoeff = blendInfo.fSrcBlend;
GrBlendCoeff dstCoeff = blendInfo.fDstBlend;
bool blendOff = GrBlendShouldDisable(equation, srcCoeff, dstCoeff);
mtlColorAttachment.blendingEnabled = !blendOff;
if (!blendOff) {
mtlColorAttachment.sourceRGBBlendFactor = blend_coeff_to_mtl_blend(srcCoeff);
mtlColorAttachment.destinationRGBBlendFactor = blend_coeff_to_mtl_blend(dstCoeff);
mtlColorAttachment.rgbBlendOperation = blend_equation_to_mtl_blend_op(equation);
mtlColorAttachment.sourceAlphaBlendFactor = blend_coeff_to_mtl_blend(srcCoeff);
mtlColorAttachment.destinationAlphaBlendFactor = blend_coeff_to_mtl_blend(dstCoeff);
mtlColorAttachment.alphaBlendOperation = blend_equation_to_mtl_blend_op(equation);
}
if (!blendInfo.fWriteColor) {
mtlColorAttachment.writeMask = MTLColorWriteMaskNone;
} else {
mtlColorAttachment.writeMask = MTLColorWriteMaskAll;
}
return mtlColorAttachment;
}
uint32_t buffer_size(uint32_t offset, uint32_t maxAlignment) {
// Metal expects the buffer to be padded at the end according to the alignment
// of the largest element in the buffer.
uint32_t offsetDiff = offset & maxAlignment;
if (offsetDiff != 0) {
offsetDiff = maxAlignment - offsetDiff + 1;
}
return offset + offsetDiff;
}
GrMtlPipelineState* GrMtlPipelineStateBuilder::finalize(GrRenderTarget* renderTarget,
const GrProgramDesc& desc,
const GrProgramInfo& programInfo) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
auto pipelineDescriptor = [[MTLRenderPipelineDescriptor alloc] init];
id<MTLLibrary> shaderLibraries[kGrShaderTypeCount];
fVS.extensions().appendf("#extension GL_ARB_separate_shader_objects : enable\n");
fFS.extensions().appendf("#extension GL_ARB_separate_shader_objects : enable\n");
fVS.extensions().appendf("#extension GL_ARB_shading_language_420pack : enable\n");
fFS.extensions().appendf("#extension GL_ARB_shading_language_420pack : enable\n");
this->finalizeShaders();
SkSL::Program::Settings settings;
settings.fFlipY = this->origin() != kTopLeft_GrSurfaceOrigin;
settings.fSharpenTextures = fGpu->getContext()->priv().options().fSharpenMipmappedTextures;
SkASSERT(!this->fragColorIsInOut());
sk_sp<SkData> cached;
SkReadBuffer reader;
SkFourByteTag shaderType = 0;
auto persistentCache = fGpu->getContext()->priv().getPersistentCache();
if (persistentCache) {
// Here we shear off the Mtl-specific portion of the Desc in order to create the
// persistent key. This is because Mtl only caches the MSL code, not the fully compiled
// program, and that only depends on the base GrProgramDesc data.
sk_sp<SkData> key = SkData::MakeWithoutCopy(desc.asKey(), desc.initialKeyLength());
cached = persistentCache->load(*key);
if (cached) {
reader.setMemory(cached->data(), cached->size());
shaderType = GrPersistentCacheUtils::GetType(&reader);
}
}
if (kMSL_Tag == shaderType && this->loadShadersFromCache(&reader, shaderLibraries)) {
// We successfully loaded and compiled MSL
} else {
SkSL::String shaders[kGrShaderTypeCount];
SkSL::Program::Inputs inputs[kGrShaderTypeCount];
SkSL::String* sksl[kGrShaderTypeCount] = {
&fVS.fCompilerString,
nullptr, // geometry shaders not supported
&fFS.fCompilerString,
};
SkSL::String cached_sksl[kGrShaderTypeCount];
if (kSKSL_Tag == shaderType) {
if (GrPersistentCacheUtils::UnpackCachedShaders(&reader, cached_sksl, inputs,
kGrShaderTypeCount)) {
for (int i = 0; i < kGrShaderTypeCount; ++i) {
sksl[i] = &cached_sksl[i];
}
}
}
shaderLibraries[kVertex_GrShaderType] = this->generateMtlShaderLibrary(
*sksl[kVertex_GrShaderType],
SkSL::Program::kVertex_Kind,
settings,
&shaders[kVertex_GrShaderType],
&inputs[kVertex_GrShaderType]);
shaderLibraries[kFragment_GrShaderType] = this->generateMtlShaderLibrary(
*sksl[kFragment_GrShaderType],
SkSL::Program::kFragment_Kind,
settings,
&shaders[kFragment_GrShaderType],
&inputs[kFragment_GrShaderType]);
// Geometry shaders are not supported
SkASSERT(!this->primitiveProcessor().willUseGeoShader());
if (!shaderLibraries[kVertex_GrShaderType] || !shaderLibraries[kFragment_GrShaderType]) {
return nullptr;
}
if (persistentCache && !cached) {
bool isSkSL = false;
if (fGpu->getContext()->priv().options().fShaderCacheStrategy ==
GrContextOptions::ShaderCacheStrategy::kSkSL) {
for (int i = 0; i < kGrShaderTypeCount; ++i) {
if (sksl[i]) {
shaders[i] = GrShaderUtils::PrettyPrint(*sksl[i]);
}
}
isSkSL = true;
}
this->storeShadersInCache(shaders, inputs, isSkSL);
}
}
id<MTLFunction> vertexFunction =
[shaderLibraries[kVertex_GrShaderType] newFunctionWithName: @"vertexMain"];
id<MTLFunction> fragmentFunction =
[shaderLibraries[kFragment_GrShaderType] newFunctionWithName: @"fragmentMain"];
if (vertexFunction == nil) {
SkDebugf("Couldn't find vertexMain() in library\n");
return nullptr;
}
if (fragmentFunction == nil) {
SkDebugf("Couldn't find fragmentMain() in library\n");
return nullptr;
}
pipelineDescriptor.vertexFunction = vertexFunction;
pipelineDescriptor.fragmentFunction = fragmentFunction;
pipelineDescriptor.vertexDescriptor = create_vertex_descriptor(programInfo.primProc());
MTLPixelFormat pixelFormat = GrBackendFormatAsMTLPixelFormat(renderTarget->backendFormat());
if (pixelFormat == MTLPixelFormatInvalid) {
return nullptr;
}
pipelineDescriptor.colorAttachments[0] = create_color_attachment(pixelFormat,
programInfo.pipeline());
pipelineDescriptor.sampleCount = programInfo.numRasterSamples();
bool hasStencilAttachment = SkToBool(renderTarget->getStencilAttachment());
GrMtlCaps* mtlCaps = (GrMtlCaps*)this->caps();
pipelineDescriptor.stencilAttachmentPixelFormat =
hasStencilAttachment ? mtlCaps->preferredStencilFormat() : MTLPixelFormatInvalid;
SkASSERT(pipelineDescriptor.vertexFunction);
SkASSERT(pipelineDescriptor.fragmentFunction);
SkASSERT(pipelineDescriptor.vertexDescriptor);
SkASSERT(pipelineDescriptor.colorAttachments[0]);
NSError* error = nil;
#if GR_METAL_SDK_VERSION >= 230
if (@available(macOS 11.0, iOS 14.0, *)) {
id<MTLBinaryArchive> archive = fGpu->binaryArchive();
if (archive) {
NSArray* archiveArray = [NSArray arrayWithObjects:archive, nil];
pipelineDescriptor.binaryArchives = archiveArray;
BOOL result;
{
TRACE_EVENT0("skia.gpu", "addRenderPipelineFunctionsWithDescriptor");
result = [archive addRenderPipelineFunctionsWithDescriptor: pipelineDescriptor
error: &error];
}
if (!result && error) {
SkDebugf("Error storing pipeline: %s\n",
[[error localizedDescription] cStringUsingEncoding: NSASCIIStringEncoding]);
}
}
}
#endif
id<MTLRenderPipelineState> pipelineState;
{
TRACE_EVENT0("skia.gpu", "newRenderPipelineStateWithDescriptor");
#if defined(SK_BUILD_FOR_MAC)
pipelineState = GrMtlNewRenderPipelineStateWithDescriptor(
fGpu->device(), pipelineDescriptor, &error);
#else
pipelineState =
[fGpu->device() newRenderPipelineStateWithDescriptor: pipelineDescriptor
error: &error];
#endif
}
if (error) {
SkDebugf("Error creating pipeline: %s\n",
[[error localizedDescription] cStringUsingEncoding: NSASCIIStringEncoding]);
return nullptr;
}
if (!pipelineState) {
return nullptr;
}
uint32_t bufferSize = buffer_size(fUniformHandler.fCurrentUBOOffset,
fUniformHandler.fCurrentUBOMaxAlignment);
return new GrMtlPipelineState(fGpu,
pipelineState,
pipelineDescriptor.colorAttachments[0].pixelFormat,
fUniformHandles,
fUniformHandler.fUniforms,
bufferSize,
(uint32_t)fUniformHandler.numSamplers(),
std::move(fGeometryProcessor),
std::move(fXferProcessor),
std::move(fFragmentProcessors));
}
//////////////////////////////////////////////////////////////////////////////