You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
cocos_lib/cocos/renderer/gfx-metal/MTLPipelineState.mm

402 lines
18 KiB

/****************************************************************************
Copyright (c) 2019-2022 Xiamen Yaji Software Co., Ltd.
http://www.cocos.com
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated engine source code (the "Software"), a limited,
worldwide, royalty-free, non-assignable, revocable and non-exclusive license
to use Cocos Creator solely to develop games on your target platforms. You shall
not use Cocos Creator software for developing other software or tools that's
used for developing games. You are not granted to publish, distribute,
sublicense, and/or sell copies of Cocos Creator.
The software or tools in this License Agreement are licensed, not sold.
Xiamen Yaji Software Co., Ltd. reserves all rights not expressly granted to you.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
****************************************************************************/
#import "base/std/container/set.h"
#import "base/Log.h"
#import "MTLDevice.h"
#import "MTLGPUObjects.h"
#import "MTLPipelineLayout.h"
#import "MTLPipelineState.h"
#import "MTLRenderPass.h"
#import "MTLSampler.h"
#import "MTLShader.h"
#import "MTLTexture.h"
#import "MTLUtils.h"
#import <Metal/MTLComputePipeline.h>
#import <Metal/MTLDevice.h>
#import <Metal/MTLVertexDescriptor.h>
namespace cc {
namespace gfx {
CCMTLPipelineState::CCMTLPipelineState() : PipelineState() {
_typedID = generateObjectID<decltype(this)>();
}
CCMTLPipelineState::~CCMTLPipelineState() {
destroy();
}
void CCMTLPipelineState::doInit(const PipelineStateInfo &info) {
createGPUPipelineState();
}
void CCMTLPipelineState::doDestroy() {
CC_SAFE_DELETE(_GPUPipelineState);
id<MTLRenderPipelineState> renderPipelineState = _mtlRenderPipelineState;
_mtlRenderPipelineState = nil;
id<MTLDepthStencilState> depthStencilState = _mtlDepthStencilState;
_mtlDepthStencilState = nil;
std::function<void(void)> destroyFunc = [=]() {
if (renderPipelineState) {
[renderPipelineState release];
}
if (depthStencilState) {
[depthStencilState release];
}
};
CCMTLGPUGarbageCollectionPool::getInstance()->collect(destroyFunc);
}
bool CCMTLPipelineState::initRenderPipeline() {
if (!createMTLRenderPipelineState()) {
return false;
}
// Application can run with wrong depth/stencil state.
if (!createMTLDepthStencilState()) {
return false;
}
_GPUPipelineState->mtlDepthStencilState = _mtlDepthStencilState;
_GPUPipelineState->mtlRenderPipelineState = _mtlRenderPipelineState;
_GPUPipelineState->cullMode = static_cast<MTLCullMode>(mu::toMTLCullMode(_rasterizerState.cullMode));
_GPUPipelineState->fillMode = static_cast<MTLTriangleFillMode>(mu::toMTLTriangleFillMode(_rasterizerState.polygonMode));
_GPUPipelineState->depthClipMode = static_cast<MTLDepthClipMode>(mu::toMTLDepthClipMode(_rasterizerState.isDepthClip != 0));
_GPUPipelineState->winding = static_cast<MTLWinding>(mu::toMTLWinding(_rasterizerState.isFrontFaceCCW != 0));
_GPUPipelineState->stencilRefFront = _depthStencilState.stencilRefFront;
_GPUPipelineState->stencilRefBack = _depthStencilState.stencilRefBack;
_GPUPipelineState->primitiveType = mu::toMTLPrimitiveType(_primitive);
if (_pipelineLayout)
_GPUPipelineState->gpuPipelineLayout = static_cast<CCMTLPipelineLayout*>(_pipelineLayout)->gpuPipelineLayout();
_GPUPipelineState->gpuShader = static_cast<CCMTLShader *>(_shader)->gpuShader(
static_cast<CCMTLRenderPass*>(_renderPass), _subpass);
_renderPipelineReady = true;
return true;
}
void CCMTLPipelineState::check(CCMTLRenderPass *renderPass) {
if (renderPass)
_renderPass = renderPass;
if (!_renderPipelineReady) {
initRenderPipeline();
}
}
bool CCMTLPipelineState::createGPUPipelineState() {
_GPUPipelineState = ccnew CCMTLGPUPipelineState;
if (!_GPUPipelineState) {
CC_LOG_ERROR("CCMTLPipelineState: new CCMTLGPUPipelineState failed.");
return false;
}
if (_bindPoint == PipelineBindPoint::GRAPHICS) {
if (_renderPass->getSubpasses().empty()) {
initRenderPipeline();
}
} else if (_bindPoint == PipelineBindPoint::COMPUTE) {
if (!createMTLComputePipelineState()) {
return false;
}
_GPUPipelineState->mtlComputePipelineState = _mtlComputePipeline;
_GPUPipelineState->gpuShader = static_cast<CCMTLShader *>(_shader)->gpuShader(nullptr, 0);
if (_pipelineLayout)
_GPUPipelineState->gpuPipelineLayout = static_cast<CCMTLPipelineLayout *>(_pipelineLayout)->gpuPipelineLayout();
}
return true;
}
bool CCMTLPipelineState::createMTLComputePipelineState() {
//create with function
id<MTLDevice> mtlDevice = id<MTLDevice>(CCMTLDevice::getInstance()->getMTLDevice());
NSError *err;
_mtlComputePipeline = [mtlDevice newComputePipelineStateWithFunction:((CCMTLShader *)_shader)->getComputeMTLFunction()
error:&err];
if (!_mtlComputePipeline) {
CC_LOG_ERROR("Create compute pipeline failed: %s", [err.localizedDescription UTF8String]);
return false;
}
return true;
}
bool CCMTLPipelineState::createMTLDepthStencilState() {
const auto& colorAttachments = _renderPass->getColorAttachments();
bool hasDS = std::any_of(colorAttachments.begin(), colorAttachments.end(), [](const ColorAttachment& attachemnt){
return attachemnt.format == Format::DEPTH ||attachemnt.format == Format::DEPTH_STENCIL;
});
hasDS |= _renderPass->getDepthStencilAttachment().format != Format::UNKNOWN;
if(!hasDS) {
// default nil
return true;
}
MTLDepthStencilDescriptor *descriptor = [[MTLDepthStencilDescriptor alloc] init];
if (descriptor == nil) {
CC_LOG_ERROR("CCMTLPipelineState: MTLDepthStencilDescriptor could not be allocated.");
return false;
}
descriptor.depthWriteEnabled = _depthStencilState.depthWrite != 0;
if (!_depthStencilState.depthTest)
descriptor.depthCompareFunction = MTLCompareFunctionAlways;
else
descriptor.depthCompareFunction = mu::toMTLCompareFunction(_depthStencilState.depthFunc);
if (_depthStencilState.stencilTestFront) {
descriptor.frontFaceStencil.stencilCompareFunction = mu::toMTLCompareFunction(_depthStencilState.stencilFuncFront);
descriptor.frontFaceStencil.readMask = _depthStencilState.stencilReadMaskFront;
descriptor.frontFaceStencil.writeMask = _depthStencilState.stencilWriteMaskFront;
descriptor.frontFaceStencil.stencilFailureOperation = mu::toMTLStencilOperation(_depthStencilState.stencilFailOpFront);
descriptor.frontFaceStencil.depthFailureOperation = mu::toMTLStencilOperation(_depthStencilState.stencilZFailOpFront);
descriptor.frontFaceStencil.depthStencilPassOperation = mu::toMTLStencilOperation(_depthStencilState.stencilPassOpFront);
} else {
descriptor.frontFaceStencil = nil;
}
if (_depthStencilState.stencilTestBack) {
descriptor.backFaceStencil.stencilCompareFunction = mu::toMTLCompareFunction(_depthStencilState.stencilFuncBack);
descriptor.backFaceStencil.readMask = _depthStencilState.stencilReadMaskBack;
descriptor.backFaceStencil.writeMask = _depthStencilState.stencilWriteMaskBack;
descriptor.backFaceStencil.stencilFailureOperation = mu::toMTLStencilOperation(_depthStencilState.stencilFailOpBack);
descriptor.backFaceStencil.depthFailureOperation = mu::toMTLStencilOperation(_depthStencilState.stencilZFailOpBack);
descriptor.backFaceStencil.depthStencilPassOperation = mu::toMTLStencilOperation(_depthStencilState.stencilPassOpBack);
} else {
descriptor.backFaceStencil = nil;
}
id<MTLDevice> mtlDevice = id<MTLDevice>(CCMTLDevice::getInstance()->getMTLDevice());
_mtlDepthStencilState = [mtlDevice newDepthStencilStateWithDescriptor:descriptor];
[descriptor release];
if (!_mtlDepthStencilState) {
CC_LOG_ERROR("Failed to create MTLDepthStencilState.");
return false;
}
return true;
}
bool CCMTLPipelineState::createMTLRenderPipelineState() {
bool ret = true;
MTLRenderPipelineDescriptor *descriptor = [[MTLRenderPipelineDescriptor alloc] init];
if (descriptor == nil) {
CC_LOG_ERROR("CCMTLPipelineState: MTLRenderPipelineDescriptor could not be allocated.");
ret = false;
}
if(ret) ret = setMTLFunctionsAndFormats(descriptor);
if(ret) ret = setVertexDescriptor(descriptor);
if(ret) ret = setBlendStates(descriptor);
if(ret) ret = createMTLRenderPipeline(descriptor);
[descriptor release];
if(!ret) {
CC_LOG_ERROR("Failed to create pipeline state, please check if shader/pileinelayout match with each other!");
}
return ret;
}
//TODO: reconstruction
bool CCMTLPipelineState::setVertexDescriptor(MTLRenderPipelineDescriptor *descriptor) {
bool res = true;
auto activeAttributes = static_cast<CCMTLShader *>(_shader)->getAttributes();
ccstd::vector<std::tuple<int /**vertexBufferBindingIndex*/, uint32_t /**stream*/>> layouts;
ccstd::unordered_map<int /**vertexBufferBindingIndex*/, std::tuple<uint32_t /**stride*/, bool /**isInstanced*/>> map;
ccstd::vector<uint32_t> streamOffsets(CCMTLDevice::getInstance()->getCapabilities().maxVertexAttributes, 0u);
ccstd::vector<bool> activeAttribIdx(activeAttributes.size(), false);
for (const auto &inputAttrib : _inputState.attributes) {
auto bufferIndex = static_cast<CCMTLShader *>(_shader)->getAvailableBufferBindingIndex(ShaderStageFlagBit::VERTEX, inputAttrib.stream);
for (auto i = 0; i < activeAttributes.size(); i++) {
const auto &activeAttribute = activeAttributes[i];
if (inputAttrib.name == activeAttribute.name) {
descriptor.vertexDescriptor.attributes[activeAttribute.location].format = mu::toMTLVertexFormat(inputAttrib.format, inputAttrib.isNormalized);
descriptor.vertexDescriptor.attributes[activeAttribute.location].offset = streamOffsets[inputAttrib.stream];
descriptor.vertexDescriptor.attributes[activeAttribute.location].bufferIndex = bufferIndex;
auto tuple = std::make_tuple(bufferIndex, inputAttrib.stream);
if (std::find(layouts.begin(), layouts.end(), tuple) == layouts.end())
layouts.emplace_back(tuple);
activeAttribIdx[i] = true;
break;
}
}
uint32_t attributeSize = GFX_FORMAT_INFOS[(int)inputAttrib.format].size;
// NOTE: Metal requires 4 bytes alignment for attribute
if (attributeSize > 0) {
attributeSize = utils::alignTo(attributeSize, 4U);
}
streamOffsets[inputAttrib.stream] += attributeSize;
map[bufferIndex] = std::make_tuple(streamOffsets[inputAttrib.stream], inputAttrib.isInstanced);
}
for (auto i = 0; i < activeAttribIdx.size(); i++) {
if (activeAttribIdx[i]) continue;
const auto &dummy = activeAttributes[i];
descriptor.vertexDescriptor.attributes[dummy.location].format = MTLVertexFormatFloat;
descriptor.vertexDescriptor.attributes[dummy.location].offset = 0;
descriptor.vertexDescriptor.attributes[dummy.location].bufferIndex = static_cast<CCMTLShader *>(_shader)->getAvailableBufferBindingIndex(ShaderStageFlagBit::VERTEX, dummy.stream);
CC_LOG_WARNING("Attribute %s is missing, add a dummy data for it.", dummy.name.c_str());
}
if(layouts.empty()) {
res = false;
}
// layouts
for (const auto &layout : layouts) {
auto index = std::get<0>(layout);
descriptor.vertexDescriptor.layouts[index].stride = std::get<0>(map[index]);
descriptor.vertexDescriptor.layouts[index].stepFunction = std::get<1>(map[index]) ? MTLVertexStepFunctionPerInstance : MTLVertexStepFunctionPerVertex;
descriptor.vertexDescriptor.layouts[index].stepRate = 1;
//to improve performance: https://developer.apple.com/documentation/metal/mtlpipelinebufferdescriptor?language=objc
if (@available(iOS 11.0, macOS 10.13, *)) {
descriptor.vertexBuffers[index].mutability = MTLMutabilityImmutable;
}
}
_GPUPipelineState->vertexBufferBindingInfo = std::move(layouts);
return res;
}
bool CCMTLPipelineState::setMTLFunctionsAndFormats(MTLRenderPipelineDescriptor *descriptor) {
auto *mtlPass = static_cast<CCMTLRenderPass*>(_renderPass);
const SubpassInfoList &subpasses = _renderPass->getSubpasses();
const ColorAttachmentList &colorAttachments = _renderPass->getColorAttachments();
const auto &ccShader = static_cast<CCMTLShader *>(_shader);
const CCMTLGPUShader *gpuShader = ccShader->gpuShader(mtlPass, _subpass);
MTLPixelFormat mtlPixelFormat;
ccstd::set<uint32_t> inputs;
uint32_t depthStencilTexIndex = INVALID_BINDING;
if (!subpasses.empty()) {
for (size_t passIndex = 0; passIndex < subpasses.size(); ++passIndex) {
const SubpassInfo &subpass = subpasses[passIndex];
depthStencilTexIndex = subpass.depthStencil;
for (size_t i = 0; i < subpass.inputs.size(); ++i) {
uint32_t input = subpass.inputs[i];
if (inputs.find(input) == inputs.end()) {
inputs.insert(input);
if(_renderPass->getColorAttachments()[input].format == Format::DEPTH ||
_renderPass->getColorAttachments()[input].format == Format::DEPTH_STENCIL) {
continue;
}
mtlPixelFormat = mu::toMTLPixelFormat(colorAttachments[input].format);
descriptor.colorAttachments[input].pixelFormat = mtlPixelFormat;
}
}
for (size_t i = 0; i < subpass.colors.size(); ++i) {
uint32_t output = subpass.colors[i];
if(output >= colorAttachments.size()) {
depthStencilTexIndex = output;
continue;
}
mtlPixelFormat = mu::toMTLPixelFormat(colorAttachments[output].format);
descriptor.colorAttachments[output].pixelFormat = mtlPixelFormat;
}
}
} else {
for (size_t i = 0; i < colorAttachments.size(); ++i) {
mtlPixelFormat = mu::toMTLPixelFormat(colorAttachments[i].format);
descriptor.colorAttachments[i].pixelFormat = mtlPixelFormat;
depthStencilTexIndex = INVALID_BINDING;
}
}
SampleCount sample = SampleCount::X1;
Format depthStencilFormat;
if (depthStencilTexIndex != INVALID_BINDING && depthStencilTexIndex < _renderPass->getColorAttachments().size()) {
sample = _renderPass->getColorAttachments()[depthStencilTexIndex].sampleCount;
depthStencilFormat = _renderPass->getColorAttachments()[depthStencilTexIndex].format;
} else {
sample = _renderPass->getDepthStencilAttachment().sampleCount;
depthStencilFormat = _renderPass->getDepthStencilAttachment().format;
}
descriptor.sampleCount = mu::toMTLSampleCount(sample);
auto *ccMTLShader = static_cast<CCMTLShader *>(_shader);
descriptor.vertexFunction = ccMTLShader->getVertMTLFunction();
descriptor.fragmentFunction = ccMTLShader->getFragmentMTLFunction();
mtlPixelFormat = mu::toMTLPixelFormat(depthStencilFormat);
if (mtlPixelFormat != MTLPixelFormatInvalid) {
descriptor.depthAttachmentPixelFormat = mtlPixelFormat;
if (depthStencilFormat == Format::DEPTH_STENCIL)
descriptor.stencilAttachmentPixelFormat = mtlPixelFormat;
}
return true;
}
bool CCMTLPipelineState::setBlendStates(MTLRenderPipelineDescriptor *descriptor) {
//FIXME: how to handle these two attributes?
// BlendState::isIndepend
// BlendState::blendColor;
descriptor.alphaToCoverageEnabled = _blendState.isA2C != 0;
int i = 0;
for (const auto blendTarget : _blendState.targets) {
MTLRenderPipelineColorAttachmentDescriptor *colorDescriptor = descriptor.colorAttachments[i];
colorDescriptor.writeMask = mu::toMTLColorWriteMask(blendTarget.blendColorMask);
colorDescriptor.blendingEnabled = blendTarget.blend != 0;
if (!blendTarget.blend)
continue;
colorDescriptor.sourceRGBBlendFactor = mu::toMTLBlendFactor(blendTarget.blendSrc);
colorDescriptor.destinationRGBBlendFactor = mu::toMTLBlendFactor(blendTarget.blendDst);
colorDescriptor.rgbBlendOperation = mu::toMTLBlendOperation(blendTarget.blendEq);
colorDescriptor.sourceAlphaBlendFactor = mu::toMTLBlendFactor(blendTarget.blendSrcAlpha);
colorDescriptor.destinationAlphaBlendFactor = mu::toMTLBlendFactor(blendTarget.blendDstAlpha);
colorDescriptor.alphaBlendOperation = mu::toMTLBlendOperation(blendTarget.blendAlphaEq);
++i;
}
return true;
}
bool CCMTLPipelineState::createMTLRenderPipeline(MTLRenderPipelineDescriptor *descriptor) {
id<MTLDevice> mtlDevice = id<MTLDevice>(CCMTLDevice::getInstance()->getMTLDevice());
NSError *nsError = nil;
_mtlRenderPipelineState = [mtlDevice newRenderPipelineStateWithDescriptor:descriptor error:&nsError];
if (!_mtlRenderPipelineState) {
CC_LOG_ERROR("Failed to create MTLRenderPipelineState: %s", [nsError.localizedDescription UTF8String]);
return false;
}
return true;
}
} // namespace gfx
} // namespace cc