RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 20:51:52 +01:00
Code Issues Projects Releases Wiki Activity

[llvm-exegesis] Finish plumbing the Config field.

Browse Source 
Summary:
Right now there are no snippet generators that emit the `Config` Field,
but I plan to add it to investigate LEA operands for PR32326.

What was broken was:
 - `Config` Was not propagated up until the BenchmarkResult::Key.
 - Clustering should really consider different configs as measuring
 different things, so we should stabilize on (Opcode, Config) instead of
 just Opcode.

Reviewers: gchatelet

Subscribers: tschuett, llvm-commits, lebedev.ri

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68629

llvm-svn: 374031
This commit is contained in:
Clement Courbet 2019-10-08 09:06:48 +00:00
parent 15e28d2fc9
commit 260955c1df
9 changed files with 49 additions and 54 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
test/tools/llvm-exegesis/X86/analysis-cluster-stabilization-config.test
View File

@ -4,8 +4,7 @@
# have different configs, so they should not be placed in the same cluster by
# stabilization.
# CHECK-UNSTABLE: SQRTSSr
# CHECK-UNSTABLE: SQRTSSr
# CHECK-UNSTABLE-NOT: SQRTSSr
---
mode: latency

9
tools/llvm-exegesis/lib/BenchmarkCode.h
View File

@ -9,7 +9,7 @@
#ifndef LLVM_TOOLS_LLVM_EXEGESIS_BENCHMARKCODE_H
#define LLVM_TOOLS_LLVM_EXEGESIS_BENCHMARKCODE_H
#include "RegisterValue.h"
#include "BenchmarkResult.h"
#include "llvm/MC/MCInst.h"
#include <string>
#include <vector>
@ -19,12 +19,7 @@ namespace exegesis {
// A collection of instructions that are to be assembled, executed and measured.
struct BenchmarkCode {
// The sequence of instructions that are to be repeated.
std::vector<llvm::MCInst> Instructions;
// Before the code is executed some instructions are added to setup the
// registers initial values.
std::vector<RegisterValue> RegisterInitialValues;
InstructionBenchmarkKey Key;
// We also need to provide the registers that are live on entry for the
// assembler to generate proper prologue/epilogue.

2
tools/llvm-exegesis/lib/BenchmarkResult.h
View File

@ -15,8 +15,8 @@
#ifndef LLVM_TOOLS_LLVM_EXEGESIS_BENCHMARKRESULT_H
#define LLVM_TOOLS_LLVM_EXEGESIS_BENCHMARKRESULT_H
#include "BenchmarkCode.h"
#include "LlvmState.h"
#include "RegisterValue.h"
#include "llvm/ADT/StringMap.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/MC/MCInst.h"

22
tools/llvm-exegesis/lib/BenchmarkRunner.cpp
View File

@ -31,7 +31,6 @@ BenchmarkRunner::BenchmarkRunner(const LLVMState &State,
BenchmarkRunner::~BenchmarkRunner() = default;
namespace {
class FunctionExecutorImpl : public BenchmarkRunner::FunctionExecutor {
public:
@ -92,10 +91,9 @@ InstructionBenchmark BenchmarkRunner::runConfiguration(
InstrBenchmark.NumRepetitions = NumRepetitions;
InstrBenchmark.Info = BC.Info;
const std::vector<llvm::MCInst> &Instructions = BC.Instructions;
const std::vector<llvm::MCInst> &Instructions = BC.Key.Instructions;
InstrBenchmark.Key.Instructions = Instructions;
InstrBenchmark.Key.RegisterInitialValues = BC.RegisterInitialValues;
InstrBenchmark.Key = BC.Key;
// Assemble at least kMinInstructionsForSnippet instructions by repeating the
// snippet for debug/analysis. This is so that the user clearly understands
@ -104,10 +102,10 @@ InstructionBenchmark BenchmarkRunner::runConfiguration(
{
llvm::SmallString<0> Buffer;
llvm::raw_svector_ostream OS(Buffer);
assembleToStream(
State.getExegesisTarget(), State.createTargetMachine(), BC.LiveIns,
BC.RegisterInitialValues,
Repetitor.Repeat(BC.Instructions, kMinInstructionsForSnippet), OS);
assembleToStream(State.getExegesisTarget(), State.createTargetMachine(),
BC.LiveIns, BC.Key.RegisterInitialValues,
Repetitor.Repeat(Instructions, kMinInstructionsForSnippet),
OS);
const ExecutableFunction EF(State.createTargetMachine(),
getObjectFromBuffer(OS.str()));
const auto FnBytes = EF.getFunctionBytes();
@ -117,7 +115,7 @@ InstructionBenchmark BenchmarkRunner::runConfiguration(
// Assemble NumRepetitions instructions repetitions of the snippet for
// measurements.
const auto Filler =
Repetitor.Repeat(BC.Instructions, InstrBenchmark.NumRepetitions);
Repetitor.Repeat(Instructions, InstrBenchmark.NumRepetitions);
llvm::object::OwningBinary<llvm::object::ObjectFile> ObjectFile;
if (DumpObjectToDisk) {
@ -133,7 +131,7 @@ InstructionBenchmark BenchmarkRunner::runConfiguration(
llvm::SmallString<0> Buffer;
llvm::raw_svector_ostream OS(Buffer);
assembleToStream(State.getExegesisTarget(), State.createTargetMachine(),
BC.LiveIns, BC.RegisterInitialValues, Filler, OS);
BC.LiveIns, BC.Key.RegisterInitialValues, Filler, OS);
ObjectFile = getObjectFromBuffer(OS.str());
}
@ -150,7 +148,7 @@ InstructionBenchmark BenchmarkRunner::runConfiguration(
// Scale the measurements by instruction.
BM.PerInstructionValue /= InstrBenchmark.NumRepetitions;
// Scale the measurements by snippet.
BM.PerSnippetValue *= static_cast<double>(BC.Instructions.size()) /
BM.PerSnippetValue *= static_cast<double>(Instructions.size()) /
InstrBenchmark.NumRepetitions;
}
@ -167,7 +165,7 @@ BenchmarkRunner::writeObjectFile(const BenchmarkCode &BC,
return std::move(E);
llvm::raw_fd_ostream OFS(ResultFD, true /*ShouldClose*/);
assembleToStream(State.getExegesisTarget(), State.createTargetMachine(),
BC.LiveIns, BC.RegisterInitialValues, FillFunction, OFS);
BC.LiveIns, BC.Key.RegisterInitialValues, FillFunction, OFS);
return ResultPath.str();
}

52
tools/llvm-exegesis/lib/Clustering.cpp
View File

@ -237,39 +237,40 @@ void InstructionBenchmarkClustering::clusterizeNaive(unsigned NumOpcodes) {
// We shall find every opcode with benchmarks not in just one cluster, and move
// *all* the benchmarks of said Opcode into one new unstable cluster per Opcode.
void InstructionBenchmarkClustering::stabilize(unsigned NumOpcodes) {
// Given an instruction Opcode, in which clusters do benchmarks of this
// instruction lie? Normally, they all should be in the same cluster.
std::vector<llvm::SmallSet<ClusterId, 1>> OpcodeToClusterIDs;
OpcodeToClusterIDs.resize(NumOpcodes);
// The list of opcodes that have more than one cluster.
llvm::SetVector<size_t> UnstableOpcodes;
// Populate OpcodeToClusterIDs and UnstableOpcodes data structures.
// Given an instruction Opcode and Config, in which clusters do benchmarks of
// this instruction lie? Normally, they all should be in the same cluster.
struct OpcodeAndConfig {
explicit OpcodeAndConfig(const InstructionBenchmark &IB)
: Opcode(IB.keyInstruction().getOpcode()), Config(&IB.Key.Config) {}
unsigned Opcode;
const std::string *Config;
auto Tie() const -> auto { return std::tie(Opcode, *Config); }
bool operator<(const OpcodeAndConfig &O) const { return Tie() < O.Tie(); }
bool operator!=(const OpcodeAndConfig &O) const { return Tie() != O.Tie(); }
};
std::map<OpcodeAndConfig, llvm::SmallSet<ClusterId, 1>>
OpcodeConfigToClusterIDs;
// Populate OpcodeConfigToClusterIDs and UnstableOpcodes data structures.
assert(ClusterIdForPoint_.size() == Points_.size() && "size mismatch");
for (const auto &Point : zip(Points_, ClusterIdForPoint_)) {
const ClusterId &ClusterIdOfPoint = std::get<1>(Point);
if (!ClusterIdOfPoint.isValid())
continue; // Only process fully valid clusters.
const unsigned Opcode = std::get<0>(Point).keyInstruction().getOpcode();
assert(Opcode < NumOpcodes && "NumOpcodes is incorrect (too small)");
const OpcodeAndConfig Key(std::get<0>(Point));
llvm::SmallSet<ClusterId, 1> &ClusterIDsOfOpcode =
OpcodeToClusterIDs[Opcode];
OpcodeConfigToClusterIDs[Key];
ClusterIDsOfOpcode.insert(ClusterIdOfPoint);
// Is there more than one ClusterID for this opcode?.
if (ClusterIDsOfOpcode.size() < 2)
continue; // If not, then at this moment this Opcode is stable.
// Else let's record this unstable opcode for future use.
UnstableOpcodes.insert(Opcode);
}
assert(OpcodeToClusterIDs.size() == NumOpcodes && "sanity check");
// We know with how many [new] clusters we will end up with.
const auto NewTotalClusterCount = Clusters_.size() + UnstableOpcodes.size();
Clusters_.reserve(NewTotalClusterCount);
for (const size_t UnstableOpcode : UnstableOpcodes.getArrayRef()) {
for (const auto &OpcodeConfigToClusterID : OpcodeConfigToClusterIDs) {
const llvm::SmallSet<ClusterId, 1> &ClusterIDs =
OpcodeToClusterIDs[UnstableOpcode];
assert(ClusterIDs.size() > 1 &&
"Should only have Opcodes with more than one cluster.");
OpcodeConfigToClusterID.second;
const OpcodeAndConfig &Key = OpcodeConfigToClusterID.first;
// We only care about unstable instructions.
if (ClusterIDs.size() < 2)
continue;
// Create a new unstable cluster, one per Opcode.
Clusters_.emplace_back(ClusterId::makeValidUnstable(Clusters_.size()));
@ -290,8 +291,8 @@ void InstructionBenchmarkClustering::stabilize(unsigned NumOpcodes) {
// and the rest of the points is for the UnstableOpcode.
const auto it = std::stable_partition(
OldCluster.PointIndices.begin(), OldCluster.PointIndices.end(),
[this, UnstableOpcode](size_t P) {
return Points_[P].keyInstruction().getOpcode() != UnstableOpcode;
[this, &Key](size_t P) {
return OpcodeAndConfig(Points_[P]) != Key;
});
assert(std::distance(it, OldCluster.PointIndices.end()) > 0 &&
"Should have found at least one bad point");
@ -314,7 +315,6 @@ void InstructionBenchmarkClustering::stabilize(unsigned NumOpcodes) {
"New unstable cluster should end up with no less points than there "
"was clusters");
}
assert(Clusters_.size() == NewTotalClusterCount && "sanity check");
}
llvm::Expected<InstructionBenchmarkClustering>

2
tools/llvm-exegesis/lib/CodeTemplate.h
View File

@ -115,6 +115,8 @@ struct CodeTemplate {
CodeTemplate &operator=(const CodeTemplate &) = delete;
ExecutionMode Execution = ExecutionMode::UNKNOWN;
// See InstructionBenchmarkKey.::Config.
std::string Config;
// Some information about how this template has been created.
std::string Info;
// The list of the instructions for this template.

4
tools/llvm-exegesis/lib/SnippetFile.cpp
View File

@ -36,7 +36,7 @@ public:
// instructions.
void EmitInstruction(const MCInst &Instruction,
const MCSubtargetInfo &STI) override {
Result->Instructions.push_back(Instruction);
Result->Key.Instructions.push_back(Instruction);
}
// Implementation of the AsmCommentConsumer.
@ -65,7 +65,7 @@ public:
const StringRef HexValue = Parts[1].trim();
RegVal.Value = APInt(
/* each hex digit is 4 bits */ HexValue.size() * 4, HexValue, 16);
Result->RegisterInitialValues.push_back(std::move(RegVal));
Result->Key.RegisterInitialValues.push_back(std::move(RegVal));
return;
}
if (CommentText.consume_front("LIVEIN")) {

5
tools/llvm-exegesis/lib/SnippetGenerator.cpp
View File

@ -73,12 +73,13 @@ SnippetGenerator::generateConfigurations(
BC.Info = CT.Info;
for (InstructionTemplate &IT : CT.Instructions) {
randomizeUnsetVariables(State.getExegesisTarget(), ForbiddenRegs, IT);
BC.Instructions.push_back(IT.build());
BC.Key.Instructions.push_back(IT.build());
}
if (CT.ScratchSpacePointerInReg)
BC.LiveIns.push_back(CT.ScratchSpacePointerInReg);
BC.RegisterInitialValues =
BC.Key.RegisterInitialValues =
computeRegisterInitialValues(CT.Instructions);
BC.Key.Config = CT.Config;
Output.push_back(std::move(BC));
}
}

4
unittests/tools/llvm-exegesis/X86/SnippetFileTest.cpp
View File

@ -78,8 +78,8 @@ TEST_F(X86SnippetFileTest, Works) {
EXPECT_FALSE((bool)Snippets.takeError());
ASSERT_THAT(*Snippets, SizeIs(1));
const auto &Snippet = (*Snippets)[0];
ASSERT_THAT(Snippet.Instructions, ElementsAre(HasOpcode(X86::INC64r)));
ASSERT_THAT(Snippet.RegisterInitialValues,
ASSERT_THAT(Snippet.Key.Instructions, ElementsAre(HasOpcode(X86::INC64r)));
ASSERT_THAT(Snippet.Key.RegisterInitialValues,
ElementsAre(RegisterInitialValueIs(X86::RAX, 15),
RegisterInitialValueIs(X86::SIL, 0)));
ASSERT_THAT(Snippet.LiveIns, ElementsAre(X86::RDI, X86::DL));