EIC code displayed by LXR |
|
|
|
|||
File indexing completed on 2026-05-10 08:48:11
0001 //===- LoopGeneratorsKMP.h - IR helper to create loops ----------*- C++ -*-===// 0002 // 0003 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 0004 // See https://llvm.org/LICENSE.txt for license information. 0005 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 0006 // 0007 //===----------------------------------------------------------------------===// 0008 // 0009 // This file contains functions to create scalar and OpenMP parallel loops 0010 // as LLVM-IR. 0011 // 0012 //===----------------------------------------------------------------------===// 0013 #ifndef POLLY_LOOP_GENERATORS_KMP_H 0014 #define POLLY_LOOP_GENERATORS_KMP_H 0015 0016 #include "polly/CodeGen/IRBuilder.h" 0017 #include "polly/CodeGen/LoopGenerators.h" 0018 #include "polly/Support/ScopHelper.h" 0019 #include "llvm/ADT/SetVector.h" 0020 0021 namespace polly { 0022 using llvm::GlobalValue; 0023 using llvm::GlobalVariable; 0024 0025 /// This ParallelLoopGenerator subclass handles the generation of parallelized 0026 /// code, utilizing the LLVM OpenMP library. 0027 class ParallelLoopGeneratorKMP final : public ParallelLoopGenerator { 0028 public: 0029 /// Create a parallel loop generator for the current function. 0030 ParallelLoopGeneratorKMP(PollyIRBuilder &Builder, const DataLayout &DL) 0031 : ParallelLoopGenerator(Builder, DL) { 0032 SourceLocationInfo = createSourceLocation(); 0033 } 0034 0035 protected: 0036 /// The source location struct of this loop. 0037 /// ident_t = type { i32, i32, i32, i32, i8* } 0038 GlobalValue *SourceLocationInfo; 0039 0040 /// Convert the combination of given chunk size and scheduling type (which 0041 /// might have been set via the command line) into the corresponding 0042 /// scheduling type. This may result (e.g.) in a 'change' from 0043 /// "static chunked" scheduling to "static non-chunked" (regarding the 0044 /// provided and returned scheduling types). 0045 /// 0046 /// @param ChunkSize The chunk size, set via command line or its default. 0047 /// @param Scheduling The scheduling, set via command line or its default. 0048 /// 0049 /// @return The corresponding OMPGeneralSchedulingType. 0050 OMPGeneralSchedulingType 0051 getSchedType(int ChunkSize, OMPGeneralSchedulingType Scheduling) const; 0052 0053 /// Returns True if 'LongType' is 64bit wide, otherwise: False. 0054 bool is64BitArch(); 0055 0056 public: 0057 // The functions below may be used if one does not want to generate a 0058 // specific OpenMP parallel loop, but generate individual parts of it 0059 // (e.g. the subfunction definition). 0060 0061 /// Create a runtime library call to spawn the worker threads. 0062 /// 0063 /// @param SubFn The subfunction which holds the loop body. 0064 /// @param SubFnParam The parameter for the subfunction (basically the struct 0065 /// filled with the outside values). 0066 /// @param LB The lower bound for the loop we parallelize. 0067 /// @param UB The upper bound for the loop we parallelize. 0068 /// @param Stride The stride of the loop we parallelize. 0069 void createCallSpawnThreads(Value *SubFn, Value *SubFnParam, Value *LB, 0070 Value *UB, Value *Stride); 0071 0072 void deployParallelExecution(Function *SubFn, Value *SubFnParam, Value *LB, 0073 Value *UB, Value *Stride) override; 0074 0075 Function *prepareSubFnDefinition(Function *F) const override; 0076 0077 std::tuple<Value *, Function *> createSubFn(Value *Stride, AllocaInst *Struct, 0078 SetVector<Value *> UsedValues, 0079 ValueMapT &VMap) override; 0080 0081 /// Create a runtime library call to get the current global thread number. 0082 /// 0083 /// @return A Value ref which holds the current global thread number. 0084 Value *createCallGlobalThreadNum(); 0085 0086 /// Create a runtime library call to request a number of threads. 0087 /// Which will be used in the next OpenMP section (by the next fork). 0088 /// 0089 /// @param GlobalThreadID The global thread ID. 0090 /// @param NumThreads The number of threads to use. 0091 void createCallPushNumThreads(Value *GlobalThreadID, Value *NumThreads); 0092 0093 /// Create a runtime library call to prepare the OpenMP runtime. 0094 /// For dynamically scheduled loops, saving the loop arguments. 0095 /// 0096 /// @param GlobalThreadID The global thread ID. 0097 /// @param LB The loop's lower bound. 0098 /// @param UB The loop's upper bound. 0099 /// @param Inc The loop increment. 0100 /// @param ChunkSize The chunk size of the parallel loop. 0101 void createCallDispatchInit(Value *GlobalThreadID, Value *LB, Value *UB, 0102 Value *Inc, Value *ChunkSize); 0103 0104 /// Create a runtime library call to retrieve the next (dynamically) 0105 /// allocated chunk of work for this thread. 0106 /// 0107 /// @param GlobalThreadID The global thread ID. 0108 /// @param IsLastPtr Pointer to a flag, which is set to 1 if this is 0109 /// the last chunk of work, or 0 otherwise. 0110 /// @param LBPtr Pointer to the lower bound for the next chunk. 0111 /// @param UBPtr Pointer to the upper bound for the next chunk. 0112 /// @param StridePtr Pointer to the stride for the next chunk. 0113 /// 0114 /// @return A Value which holds 1 if there is work to be done, 0 otherwise. 0115 Value *createCallDispatchNext(Value *GlobalThreadID, Value *IsLastPtr, 0116 Value *LBPtr, Value *UBPtr, Value *StridePtr); 0117 0118 /// Create a runtime library call to prepare the OpenMP runtime. 0119 /// For statically scheduled loops, saving the loop arguments. 0120 /// 0121 /// @param GlobalThreadID The global thread ID. 0122 /// @param IsLastPtr Pointer to a flag, which is set to 1 if this is 0123 /// the last chunk of work, or 0 otherwise. 0124 /// @param LBPtr Pointer to the lower bound for the next chunk. 0125 /// @param UBPtr Pointer to the upper bound for the next chunk. 0126 /// @param StridePtr Pointer to the stride for the next chunk. 0127 /// @param ChunkSize The chunk size of the parallel loop. 0128 void createCallStaticInit(Value *GlobalThreadID, Value *IsLastPtr, 0129 Value *LBPtr, Value *UBPtr, Value *StridePtr, 0130 Value *ChunkSize); 0131 0132 /// Create a runtime library call to mark the end of 0133 /// a statically scheduled loop. 0134 /// 0135 /// @param GlobalThreadID The global thread ID. 0136 void createCallStaticFini(Value *GlobalThreadID); 0137 0138 /// Create the current source location. 0139 /// 0140 /// TODO: Generates only(!) dummy values. 0141 GlobalVariable *createSourceLocation(); 0142 }; 0143 } // end namespace polly 0144 #endif
| [ Source navigation ] | [ Diff markup ] | [ Identifier search ] | [ general search ] |
|
This page was automatically generated by the 2.3.7 LXR engine. The LXR team |
|