1 //===- TensorToSPIRV.cpp - Tensor to SPIR-V Patterns ----------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This file implements patterns to convert Tensor dialect to SPIR-V dialect. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "mlir/Conversion/TensorToSPIRV/TensorToSPIRV.h" 14 #include "../SPIRVCommon/Pattern.h" 15 #include "mlir/Dialect/SPIRV/IR/SPIRVDialect.h" 16 #include "mlir/Dialect/SPIRV/IR/SPIRVOps.h" 17 #include "mlir/Dialect/SPIRV/Transforms/SPIRVConversion.h" 18 #include "mlir/Dialect/SPIRV/Utils/LayoutUtils.h" 19 #include "mlir/Dialect/Tensor/IR/Tensor.h" 20 #include "mlir/IR/AffineMap.h" 21 #include "mlir/Support/LogicalResult.h" 22 #include "llvm/ADT/SetVector.h" 23 #include "llvm/Support/Debug.h" 24 25 #define DEBUG_TYPE "tensor-to-spirv-pattern" 26 27 using namespace mlir; 28 29 //===----------------------------------------------------------------------===// 30 // Operation conversion 31 //===----------------------------------------------------------------------===// 32 33 namespace { 34 35 /// Converts tensor.extract into loading using access chains from SPIR-V local 36 /// variables. 37 class TensorExtractPattern final 38 : public OpConversionPattern<tensor::ExtractOp> { 39 public: 40 TensorExtractPattern(TypeConverter &typeConverter, MLIRContext *context, 41 int64_t threshold, PatternBenefit benefit = 1) 42 : OpConversionPattern(typeConverter, context, benefit), 43 byteCountThreshold(threshold) {} 44 45 LogicalResult 46 matchAndRewrite(tensor::ExtractOp extractOp, OpAdaptor adaptor, 47 ConversionPatternRewriter &rewriter) const override { 48 TensorType tensorType = extractOp.tensor().getType().cast<TensorType>(); 49 50 if (!tensorType.hasStaticShape()) 51 return rewriter.notifyMatchFailure(extractOp, "non-static tensor"); 52 53 if (tensorType.getNumElements() * tensorType.getElementTypeBitWidth() > 54 byteCountThreshold * 8) 55 return rewriter.notifyMatchFailure(extractOp, 56 "exceeding byte count threshold"); 57 58 Location loc = extractOp.getLoc(); 59 60 int64_t rank = tensorType.getRank(); 61 SmallVector<int64_t, 4> strides(rank, 1); 62 for (int i = rank - 2; i >= 0; --i) { 63 strides[i] = strides[i + 1] * tensorType.getDimSize(i + 1); 64 } 65 66 Type varType = spirv::PointerType::get(adaptor.tensor().getType(), 67 spirv::StorageClass::Function); 68 69 spirv::VariableOp varOp; 70 if (adaptor.tensor().getDefiningOp<spirv::ConstantOp>()) { 71 varOp = rewriter.create<spirv::VariableOp>( 72 loc, varType, spirv::StorageClass::Function, 73 /*initializer=*/adaptor.tensor()); 74 } else { 75 // Need to store the value to the local variable. It's questionable 76 // whether we want to support such case though. 77 return failure(); 78 } 79 80 auto &typeConverter = *getTypeConverter<SPIRVTypeConverter>(); 81 auto indexType = typeConverter.getIndexType(); 82 83 Value index = spirv::linearizeIndex(adaptor.indices(), strides, 84 /*offset=*/0, indexType, loc, rewriter); 85 auto acOp = rewriter.create<spirv::AccessChainOp>(loc, varOp, index); 86 87 rewriter.replaceOpWithNewOp<spirv::LoadOp>(extractOp, acOp); 88 89 return success(); 90 } 91 92 private: 93 int64_t byteCountThreshold; 94 }; 95 96 } // namespace 97 98 //===----------------------------------------------------------------------===// 99 // Pattern population 100 //===----------------------------------------------------------------------===// 101 102 void mlir::populateTensorToSPIRVPatterns(SPIRVTypeConverter &typeConverter, 103 int64_t byteCountThreshold, 104 RewritePatternSet &patterns) { 105 patterns.add<TensorExtractPattern>(typeConverter, patterns.getContext(), 106 byteCountThreshold); 107 } 108