Dunfey · Hotel WWDC as data, est. 1983
Front desk everything
Years
Topics

WWDC22 · 14 min · Graphics & Games

Target and optimize GPU binaries with Metal 3

Discover how you can reduce in-app stutters, first launch times, and new level load times when you generate your GPU binaries entirely at project build time with offline compilation. We’ll also show you how to improve total compile time and binary size for larger GPU programs using the "Optimize for size" compiler option.

Watch at developer.apple.com ↗

Transcript all transcripts

Code shown on screen · 9 snippets

Using a JSON editor: render pipeline descriptor objectivec · at 4:47 ↗ 
// An existing Obj-C render pipeline descriptor
NSError *error = nil;
id<MTLDevice> device = MTLCreateSystemDefaultDevice();

id<MTLLibrary> library = [device newLibraryWithFile:@"default.metallib" error:&error];

MTLRenderPipelineDescriptor *desc = [MTLRenderPipelineDescriptor new];
desc.vertexFunction = [library newFunctionWithName:@"vert_main"];
desc.fragmentFunction = [library newFunctionWithName:@"frag_main"];
desc.rasterSampleCount = 2;
desc.colorAttachments[0].pixelFormat = MTLPixelFormatBGRA8Unorm;
desc.depthAttachmentPixelFormat = MTLPixelFormatDepth32Float;
Using a JSON editor: pipelines script json · at 4:47 ↗ 
{
  "//comment": "Its equivalent new JSON script",
  "libraries": {
    "paths": [
      {
        "path": "default.metallib"
      }
    ]
  },
  "pipelines": {
    "render_pipelines": [
      {
        "vertex_function": "vert_main",
        "fragment_function": "frag_main",
        "raster_sample_count": 2,
        "color_attachments": [
          {
            "pixel_format": "BGRA8Unorm"
          },
        ],
        "depth_attachment_pixel_format": "Depth32Float"
      }
    ]
  }
}
Harvesting sample objectivec · at 5:33 ↗ 
// Create pipeline descriptor
MTLRenderPipelineDescriptor *pipeline_desc = [MTLRenderPipelineDescriptor new];
pipeline_desc.vertexFunction = [library newFunctionWithName:@"vert_main"];
pipeline_desc.fragmentFunction = [library newFunctionWithName:@"frag_main"];
pipeline_desc.rasterSampleCount = 2;
pipeline_desc.colorAttachments[0].pixelFormat = MTLPixelFormatBGRA8Unorm;
pipeline_desc.depthAttachmentPixelFormat = MTLPixelFormatDepth32Float;

// Add pipeline descriptor to new archive
MTLBinaryArchiveDescriptor* archive_desc = [MTLBinaryArchiveDescriptor new];
id<MTLBinaryArchive> archive = [device newBinaryArchiveWithDescriptor:archive_desc error:&error];
bool success = [archive addRenderPipelineFunctionsWithDescriptor:pipeline_desc error:&error];

// Serialize archive to file system
NSURL *url = [NSURL fileURLWithPath:@"harvested-binaryArchive.metallib"];
success = [archive serializeToURL:url error:&error];
Extracting a JSON script bash · at 6:01 ↗ 
metal-source -flatbuffers=json harvested-binaryArchive.metallib -o /tmp/descriptors.mtlp-json
Generate a GPU binary from source bash · at 6:24 ↗ 
metal shaders.metal -N descriptors.mtlp-json -o archive.metallib
Generate a GPU binary from Metal library bash · at 6:48 ↗ 
metal-tt shaders.metallib descriptors.mtlp-json -o archive.metallib
Load GPU binaries via the runtime API objectivec · at 7:07 ↗ 
MTLBinaryArchiveDescriptor *desc = [MTLBinaryArchiveDescriptor new];
desc.url = [NSURL fileURLWithPath:@"archive.metallib"];
NSError *error = nil;
id<MTLDevice> device = MTLCreateSystemDefaultDevice();
id<MTLBinaryArchive> binaryArchive = [device newBinaryArchiveWithDescriptor:desc error:&error];
Enable optimize for size in command lines bash · at 12:11 ↗ 
xcrun metal -Os large_shader.metal

# or

xcrun metal -c -Os large_shader.metal
xcrun metal -c     more_shaders.metal
xcrun metal large_shader.air more_shaders.air
Enable optimize for size with Metal framework objectivec · at 12:44 ↗ 
MTLCompileOptions* options = [MTLCompileOptions new];
options.optimizationLevel = MTLLibraryOptimizationLevelSize;

NSString* source = @"...";
NSError* error = nil;
id<MTLLibrary> lib = [device newLibraryWithSource:source
                                          options:options
                                            error:&error];

Resources