Metal API
1.关于IOS 8 Metal
Metal API是apple 2014年在ios平台上推出的高效底层的3D图形API,它通过减少驱动层的API调用CPU的消耗提高渲染效率。
参考视频解说(百度云网盘链接)
2.创建Metal工程
首先要准备好苹果开发者的账号和一只iPhone 6/6s/iPad Air 2,Metal工程目前的编译和调试都需要真机进行调试。
如果没有连接真机并把目标设为真机的话,会出现Metal Compiler Error问题。
设为真机后,最后link上MetalFramework点播放键就能在iphone6上运行了。
3. 代码概览
a. 创建设备和指令队列
_device = MTLCreateSystemDefaultDevice(); // create a new command queue _commandQueue = [_device newCommandQueue]; _defaultLibrary = [_device newDefaultLibrary];</span>
b.加载资源(Metal Shader 和 ConstBuffer)
id <MTLFunction> fragmentProgram = [_defaultLibrary newFunctionWithName:@"lighting_fragment"]; id <MTLFunction> vertexProgram = [_defaultLibrary newFunctionWithName:@"lighting_vertex"];
_vertexBuffer = [_device newBufferWithBytes:kCubeVertexData length:sizeof(kCubeVertexData) options:MTLResourceOptionCPUCacheModeDefault]; _vertexBuffer.label = @"Vertices";
c.初始化渲染管线
MTLRenderPipelineDescriptor *pipelineStateDescriptor = [[MTLRenderPipelineDescriptor alloc] init]; pipelineStateDescriptor.label = @"MyPipeline"; pipelineStateDescriptor.sampleCount = _sampleCount; pipelineStateDescriptor.vertexFunction = vertexProgram; pipelineStateDescriptor.fragmentFunction = fragmentProgram; pipelineStateDescriptor.colorAttachments[0].pixelFormat = MTLPixelFormatBGRA8Unorm; pipelineStateDescriptor.depthAttachmentPixelFormat = _depthPixelFormat; // create a compiled pipeline state object. Shader functions (from the render pipeline descriptor) // are compiled when this is created unlessed they are obtained from the device's cache NSError *error = nil; _pipelineState = [_device newRenderPipelineStateWithDescriptor:pipelineStateDescriptor error:&error]; if(!_pipelineState) { NSLog(@">> ERROR: Couldnt create a pipeline: %@", error); // assert here if pipeline coudnt be created for any reason assert(0); } MTLDepthStencilDescriptor *depthStateDesc = [[MTLDepthStencilDescriptor alloc] init]; depthStateDesc.depthCompareFunction = MTLCompareFunctionLess; depthStateDesc.depthWriteEnabled = YES; _depthState = [_device newDepthStencilStateWithDescriptor:depthStateDesc]; //renderEncoder需要用到
d.执行渲染
// create a new command buffer for each renderpass to the current drawable id <MTLCommandBuffer> commandBuffer = [_commandQueue commandBuffer]; // create a render command encoder so we can render into something MTLRenderPassDescriptor *renderPassDescriptor = view.renderPassDescriptor; if (renderPassDescriptor) { id <MTLRenderCommandEncoder> renderEncoder = [commandBuffer renderCommandEncoderWithDescriptor:renderPassDescriptor]; [renderEncoder pushDebugGroup:@"Boxes"]; [renderEncoder setDepthStencilState:_depthState]; [renderEncoder setRenderPipelineState:_pipelineState]; [renderEncoder setVertexBuffer:_vertexBuffer offset:0 atIndex:0 ]; // NOTE: this could be alot faster if we render using instancing, but in this case we want to emit lots of draw calls for (int i = 0; i < kNumberOfBoxes; i++) { // set constant buffer for each box [renderEncoder setVertexBuffer:_dynamicConstantBuffer[_constantDataBufferIndex] offset:i*_sizeOfConstantT atIndex:1 ]; // tell the render context we want to draw our primitives [renderEncoder drawPrimitives:MTLPrimitiveTypeTriangle vertexStart:0 vertexCount:36 instanceCount:1]; } [renderEncoder endEncoding]; [renderEncoder popDebugGroup]; // call the view's completion handler which is required by the view since it will signal its semaphore and set up the next buffer __block dispatch_semaphore_t block_sema = _inflight_semaphore; [commandBuffer addCompletedHandler:^(id<MTLCommandBuffer> buffer) { dispatch_semaphore_signal(block_sema); }]; // schedule a present once rendering to the framebuffer is complete [commandBuffer presentDrawable:view.currentDrawable]; // finalize rendering here. this will push the command buffer to the GPU [commandBuffer commit]; }
4.Metal 调试
GPU Trace查看界面
本文版权DsoTsin所有,转载文章请注明出处!
由于缺米,Metal教程等后续Mac到手之后再更新(虚拟机太慢了)