src/gpu/graphite/mtl/MtlComputeCommandEncoder.h

/*
 * Copyright 2022 Google Inc.
 *
 * Use of this source code is governed by a BSD-style license that can be
 * found in the LICENSE file.
 */

#ifndef skgpu_graphite_MtlComputeCommandEncoder_DEFINED
#define skgpu_graphite_MtlComputeCommandEncoder_DEFINED

#include "include/core/SkRefCnt.h"
#include "include/ports/SkCFObject.h"
#include "src/gpu/graphite/ComputeTypes.h"
#include "src/gpu/graphite/Resource.h"

#import <Metal/Metal.h>

namespace skgpu::graphite {

/**
 * Wraps a MTLComputeCommandEncoder object and associated tracked state
 */
class MtlComputeCommandEncoder : public Resource {
public:
    static sk_sp<MtlComputeCommandEncoder> Make(const SharedContext* sharedContext,
                                                id<MTLCommandBuffer> commandBuffer) {
        // Inserting a pool here so the autorelease occurs when we return and the
        // only remaining ref is the retain below.
        @autoreleasepool {
            // Adding a retain here to keep our own ref separate from the autorelease pool
            sk_cfp<id<MTLComputeCommandEncoder>> encoder =
                    sk_ret_cfp([commandBuffer computeCommandEncoder]);

            // TODO(armansito): Support concurrent dispatch of compute passes using
            // MTLDispatchTypeConcurrent on macOS 10.14+ and iOS 12.0+.
            return sk_sp<MtlComputeCommandEncoder>(
                    new MtlComputeCommandEncoder(sharedContext, std::move(encoder)));
        }
    }

    const char* getResourceType() const override { return "Metal Compute Command Encoder"; }

    void setLabel(NSString* label) { [(*fCommandEncoder) setLabel:label]; }

    void pushDebugGroup(NSString* string) { [(*fCommandEncoder) pushDebugGroup:string]; }
    void popDebugGroup() { [(*fCommandEncoder) popDebugGroup]; }
    void insertDebugSignpost(NSString* string) { [(*fCommandEncoder) insertDebugSignpost:string]; }

    void setComputePipelineState(id<MTLComputePipelineState> pso) {
        if (fCurrentComputePipelineState != pso) {
            [(*fCommandEncoder) setComputePipelineState:pso];
            fCurrentComputePipelineState = pso;
        }
    }

    void setBuffer(id<MTLBuffer> buffer, NSUInteger offset, NSUInteger index) {
        SkASSERT(buffer != nil);
        SkASSERT(index < kMaxExpectedBuffers);
        if (@available(macOS 10.11, iOS 8.3, tvOS 9.0, *)) {
            if (fBuffers[index] == buffer) {
                this->setBufferOffset(offset, index);
                return;
            }
        }
        if (fBuffers[index] != buffer || fBufferOffsets[index] != offset) {
            [(*fCommandEncoder) setBuffer:buffer offset:offset atIndex:index];
            fBuffers[index] = buffer;
            fBufferOffsets[index] = offset;
        }
    }

    void setBufferOffset(NSUInteger offset, NSUInteger index)
            SK_API_AVAILABLE(macos(10.11), ios(8.3), tvos(9.0)) {
        SkASSERT(index < kMaxExpectedBuffers);
        if (fBufferOffsets[index] != offset) {
            [(*fCommandEncoder) setBufferOffset:offset atIndex:index];
            fBufferOffsets[index] = offset;
        }
    }

    void setTexture(id<MTLTexture> texture, NSUInteger index) {
        SkASSERT(index < kMaxExpectedTextures);
        if (fTextures[index] != texture) {
            [(*fCommandEncoder) setTexture:texture atIndex:index];
            fTextures[index] = texture;
        }
    }

    void setSamplerState(id<MTLSamplerState> sampler, NSUInteger index) {
        SkASSERT(index < kMaxExpectedTextures);
        if (fSamplers[index] != sampler) {
            [(*fCommandEncoder) setSamplerState:sampler atIndex:index];
            fSamplers[index] = sampler;
        }
    }

    // `length` must be 16-byte aligned
    void setThreadgroupMemoryLength(NSUInteger length, NSUInteger index) {
        SkASSERT(length % 16 == 0);
        [(*fCommandEncoder) setThreadgroupMemoryLength:length atIndex:index];
    }

    void dispatchThreadgroups(const WorkgroupSize& globalSize, const WorkgroupSize& localSize) {
        MTLSize threadgroupCount =
                MTLSizeMake(globalSize.fWidth, globalSize.fHeight, globalSize.fDepth);
        MTLSize threadsPerThreadgroup =
                MTLSizeMake(localSize.fWidth, localSize.fHeight, localSize.fDepth);
        [(*fCommandEncoder) dispatchThreadgroups:threadgroupCount
                           threadsPerThreadgroup:threadsPerThreadgroup];
    }

    void dispatchThreadgroupsWithIndirectBuffer(id<MTLBuffer> indirectBuffer,
                                                NSUInteger offset,
                                                const WorkgroupSize& localSize) {
        MTLSize threadsPerThreadgroup =
                MTLSizeMake(localSize.fWidth, localSize.fHeight, localSize.fDepth);
        [(*fCommandEncoder) dispatchThreadgroupsWithIndirectBuffer:indirectBuffer
                                              indirectBufferOffset:offset
                                             threadsPerThreadgroup:threadsPerThreadgroup];
    }

    void endEncoding() { [(*fCommandEncoder) endEncoding]; }

private:
    inline static constexpr int kMaxExpectedBuffers = 16;
    inline static constexpr int kMaxExpectedTextures = 16;

    MtlComputeCommandEncoder(const SharedContext* sharedContext,
                             sk_cfp<id<MTLComputeCommandEncoder>> encoder)
            : Resource(sharedContext,
                       Ownership::kOwned,
                       /*gpuMemorySize=*/0)
            , fCommandEncoder(std::move(encoder)) {
        for (int i = 0; i < kMaxExpectedBuffers; i++) {
            fBuffers[i] = nil;
        }
        for (int i = 0; i < kMaxExpectedTextures; i++) {
            fTextures[i] = nil;
            fSamplers[i] = nil;
        }
    }

    void freeGpuData() override { fCommandEncoder.reset(); }

    sk_cfp<id<MTLComputeCommandEncoder>> fCommandEncoder;

    id<MTLComputePipelineState> fCurrentComputePipelineState = nil;

    id<MTLBuffer> fBuffers[kMaxExpectedBuffers];
    NSUInteger    fBufferOffsets[kMaxExpectedBuffers];

    id<MTLTexture>      fTextures[kMaxExpectedTextures];
    id<MTLSamplerState> fSamplers[kMaxExpectedTextures];
};

}  // namespace skgpu::graphite

#endif  // skgpu_graphite_MtlComputeCommandEncoder_DEFINED