1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 02:33:06 +01:00
llvm-mirror/lib/Target/AMDGPU/AMDGPUPerfHintAnalysis.h
Stanislav Mekhanoshin 5b3e6630e5 [AMDGPU] Tune perfhint analysis to account access width
A function with less memory instructions but wider access
is the same as a function with more but narrower accesses
in terms of memory boundness. In fact the pass would give
different answers before and after vectorization without
this change.

Differential Revision: https://reviews.llvm.org/D105651
2021-07-21 12:46:10 -07:00

55 lines
1.6 KiB
C++

//===- AMDGPUPerfHintAnalysis.h ---- analysis of memory traffic -*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
/// \file
/// \brief Analyzes if a function potentially memory bound and if a kernel
/// kernel may benefit from limiting number of waves to reduce cache thrashing.
///
//===----------------------------------------------------------------------===//
#ifndef LLVM_LIB_TARGET_AMDGPU_MDGPUPERFHINTANALYSIS_H
#define LLVM_LIB_TARGET_AMDGPU_MDGPUPERFHINTANALYSIS_H
#include "llvm/Analysis/CallGraphSCCPass.h"
#include "llvm/IR/ValueMap.h"
namespace llvm {
struct AMDGPUPerfHintAnalysis : public CallGraphSCCPass {
static char ID;
public:
AMDGPUPerfHintAnalysis() : CallGraphSCCPass(ID) {}
bool runOnSCC(CallGraphSCC &SCC) override;
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.setPreservesAll();
}
bool isMemoryBound(const Function *F) const;
bool needsWaveLimiter(const Function *F) const;
struct FuncInfo {
unsigned MemInstCost;
unsigned InstCost;
unsigned IAMInstCost; // Indirect access memory instruction count
unsigned LSMInstCost; // Large stride memory instruction count
FuncInfo() : MemInstCost(0), InstCost(0), IAMInstCost(0), LSMInstCost(0) {}
};
typedef ValueMap<const Function*, FuncInfo> FuncInfoMap;
private:
FuncInfoMap FIM;
};
} // namespace llvm
#endif // LLVM_LIB_TARGET_AMDGPU_MDGPUPERFHINTANALYSIS_H