1 //===--------------------- ResourcePressureView.h ---------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 /// \file 10 /// 11 /// This file define class ResourcePressureView. 12 /// Class ResourcePressureView observes hardware events generated by 13 /// the Pipeline object and collects statistics related to resource usage at 14 /// instruction granularity. 15 /// Resource pressure information is then printed out to a stream in the 16 /// form of a table like the one from the example below: 17 /// 18 /// Resources: 19 /// [0] - JALU0 20 /// [1] - JALU1 21 /// [2] - JDiv 22 /// [3] - JFPM 23 /// [4] - JFPU0 24 /// [5] - JFPU1 25 /// [6] - JLAGU 26 /// [7] - JSAGU 27 /// [8] - JSTC 28 /// [9] - JVIMUL 29 /// 30 /// Resource pressure per iteration: 31 /// [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] 32 /// 0.00 0.00 0.00 0.00 2.00 2.00 0.00 0.00 0.00 0.00 33 /// 34 /// Resource pressure by instruction: 35 /// [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] Instructions: 36 /// - - - - - 1.00 - - - - vpermilpd $1, %xmm0, 37 /// %xmm1 38 /// - - - - 1.00 - - - - - vaddps %xmm0, %xmm1, 39 /// %xmm2 40 /// - - - - - 1.00 - - - - vmovshdup %xmm2, %xmm3 41 /// - - - - 1.00 - - - - - vaddss %xmm2, %xmm3, 42 /// %xmm4 43 /// 44 /// In this example, we have AVX code executed on AMD Jaguar (btver2). 45 /// Both shuffles and vector floating point add operations on XMM registers have 46 /// a reciprocal throughput of 1cy. 47 /// Each add is issued to pipeline JFPU0, while each shuffle is issued to 48 /// pipeline JFPU1. The overall pressure per iteration is reported by two 49 /// tables: the first smaller table is the resource pressure per iteration; 50 /// the second table reports resource pressure per instruction. Values are the 51 /// average resource cycles consumed by an instruction. 52 /// Every vector add from the example uses resource JFPU0 for an average of 1cy 53 /// per iteration. Consequently, the resource pressure on JFPU0 is of 2cy per 54 /// iteration. 55 /// 56 //===----------------------------------------------------------------------===// 57 58 #ifndef LLVM_TOOLS_LLVM_MCA_RESOURCEPRESSUREVIEW_H 59 #define LLVM_TOOLS_LLVM_MCA_RESOURCEPRESSUREVIEW_H 60 61 #include "SourceMgr.h" 62 #include "View.h" 63 #include "llvm/ADT/DenseMap.h" 64 #include "llvm/MC/MCInstPrinter.h" 65 #include "llvm/MC/MCSubtargetInfo.h" 66 #include <map> 67 68 namespace mca { 69 70 /// This class collects resource pressure statistics and it is able to print 71 /// out all the collected information as a table to an output stream. 72 class ResourcePressureView : public View { 73 const llvm::MCSubtargetInfo &STI; 74 llvm::MCInstPrinter &MCIP; 75 const SourceMgr &Source; 76 77 // Map to quickly obtain the ResourceUsage column index from a processor 78 // resource ID. 79 llvm::DenseMap<unsigned, unsigned> Resource2VecIndex; 80 81 // Table of resources used by instructions. 82 std::vector<double> ResourceUsage; 83 unsigned NumResourceUnits; 84 85 const llvm::MCInst &GetMCInstFromIndex(unsigned Index) const; 86 void printResourcePressurePerIteration(llvm::raw_ostream &OS, 87 unsigned Executions) const; 88 void printResourcePressurePerInstruction(llvm::raw_ostream &OS, 89 unsigned Executions) const; 90 void initialize(); 91 92 public: ResourcePressureView(const llvm::MCSubtargetInfo & sti,llvm::MCInstPrinter & Printer,const SourceMgr & SM)93 ResourcePressureView(const llvm::MCSubtargetInfo &sti, 94 llvm::MCInstPrinter &Printer, const SourceMgr &SM) 95 : STI(sti), MCIP(Printer), Source(SM) { 96 initialize(); 97 } 98 99 void onEvent(const HWInstructionEvent &Event) override; 100 printView(llvm::raw_ostream & OS)101 void printView(llvm::raw_ostream &OS) const override { 102 unsigned Executions = Source.getNumIterations(); 103 printResourcePressurePerIteration(OS, Executions); 104 printResourcePressurePerInstruction(OS, Executions); 105 } 106 }; 107 } // namespace mca 108 109 #endif 110