caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[llvm-mca] Add command line option -call-latency (#92958)

Browse Source 
Currently we assume a constant latency of 100 cycles for call
instructions. This commit allows the user to specify a custom value for
the same as a command line argument. Default latency is set to 100.
This commit is contained in:
Chinmay Deshpande
2024-05-22 13:51:55 -07:00
committed by GitHub
parent 25b65be43d
commit 848bef5d85
6 changed files with 80 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
llvm/include/llvm/MCA/InstrBuilder.h
View File

@@ -78,6 +78,7 @@ class InstrBuilder {
bool FirstCallInst;
bool FirstReturnInst;
unsigned CallLatency;
using InstRecycleCallback = std::function<Instruction *(const InstrDesc &)>;
InstRecycleCallback InstRecycleCB;
@@ -98,7 +99,7 @@ class InstrBuilder {
public:
InstrBuilder(const MCSubtargetInfo &STI, const MCInstrInfo &MCII,
const MCRegisterInfo &RI, const MCInstrAnalysis *IA,
const InstrumentManager &IM);
const InstrumentManager &IM, unsigned CallLatency);
void clear() {
Descriptors.clear();

20
llvm/lib/MCA/InstrBuilder.cpp
View File

@@ -31,9 +31,9 @@ InstrBuilder::InstrBuilder(const llvm::MCSubtargetInfo &sti,
const llvm::MCInstrInfo &mcii,
const llvm::MCRegisterInfo &mri,
const llvm::MCInstrAnalysis *mcia,
const mca::InstrumentManager &im)
const mca::InstrumentManager &im, unsigned cl)
: STI(sti), MCII(mcii), MRI(mri), MCIA(mcia), IM(im), FirstCallInst(true),
FirstReturnInst(true) {
FirstReturnInst(true), CallLatency(cl) {
const MCSchedModel &SM = STI.getSchedModel();
ProcResourceMasks.resize(SM.getNumProcResourceKinds());
computeProcResourceMasks(STI.getSchedModel(), ProcResourceMasks);
@@ -220,17 +220,19 @@ static void initializeUsedResources(InstrDesc &ID,
static void computeMaxLatency(InstrDesc &ID, const MCInstrDesc &MCDesc,
const MCSchedClassDesc &SCDesc,
const MCSubtargetInfo &STI) {
const MCSubtargetInfo &STI,
unsigned CallLatency) {
if (MCDesc.isCall()) {
// We cannot estimate how long this call will take.
// Artificially set an arbitrarily high latency (100cy).
ID.MaxLatency = 100U;
// Artificially set an arbitrarily high latency.
ID.MaxLatency = CallLatency;
return;
}
int Latency = MCSchedModel::computeInstrLatency(STI, SCDesc);
// If latency is unknown, then conservatively assume a MaxLatency of 100cy.
ID.MaxLatency = Latency < 0 ? 100U : static_cast<unsigned>(Latency);
// If latency is unknown, then conservatively assume the MaxLatency set for
// calls.
ID.MaxLatency = Latency < 0 ? CallLatency : static_cast<unsigned>(Latency);
}
static Error verifyOperands(const MCInstrDesc &MCDesc, const MCInst &MCI) {
@@ -568,7 +570,7 @@ InstrBuilder::createInstrDescImpl(const MCInst &MCI,
// We don't correctly model calls.
WithColor::warning() << "found a call in the input assembly sequence.\n";
WithColor::note() << "call instructions are not correctly modeled. "
<< "Assume a latency of 100cy.\n";
<< "Assume a latency of " << CallLatency << "cy.\n";
FirstCallInst = false;
}
@@ -580,7 +582,7 @@ InstrBuilder::createInstrDescImpl(const MCInst &MCI,
}
initializeUsedResources(*ID, SCDesc, STI, ProcResourceMasks);
computeMaxLatency(*ID, MCDesc, SCDesc, STI);
computeMaxLatency(*ID, MCDesc, SCDesc, STI, CallLatency);
if (Error Err = verifyOperands(MCDesc, MCI))
return std::move(Err);

58
llvm/test/tools/llvm-mca/X86/call-latency.s Normal file
View File

@@ -0,0 +1,58 @@
# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
# RUN: llvm-mca -mtriple=x86_64-unknown-unknown -mcpu=btver2 -iterations=1 %s | FileCheck --check-prefixes=ALL,DEFAULT %s
# RUN: llvm-mca -mtriple=x86_64-unknown-unknown -mcpu=btver2 -call-latency=50 -iterations=1 %s | FileCheck --check-prefixes=ALL,CUSTOM %s
callq printf
# ALL: Iterations: 1
# ALL-NEXT: Instructions: 1
# CUSTOM-NEXT: Total Cycles: 53
# DEFAULT-NEXT: Total Cycles: 103
# ALL-NEXT: Total uOps: 1
# ALL: Dispatch Width: 2
# CUSTOM-NEXT: uOps Per Cycle: 0.02
# CUSTOM-NEXT: IPC: 0.02
# DEFAULT-NEXT: uOps Per Cycle: 0.01
# DEFAULT-NEXT: IPC: 0.01
# ALL-NEXT: Block RThroughput: 0.5
# ALL: Instruction Info:
# ALL-NEXT: [1]: #uOps
# ALL-NEXT: [2]: Latency
# ALL-NEXT: [3]: RThroughput
# ALL-NEXT: [4]: MayLoad
# ALL-NEXT: [5]: MayStore
# ALL-NEXT: [6]: HasSideEffects (U)
# ALL: [1] [2] [3] [4] [5] [6] Instructions:
# ALL-NEXT: 1 1 0.50 callq printf
# ALL: Resources:
# ALL-NEXT: [0] - JALU0
# ALL-NEXT: [1] - JALU1
# ALL-NEXT: [2] - JDiv
# ALL-NEXT: [3] - JFPA
# ALL-NEXT: [4] - JFPM
# ALL-NEXT: [5] - JFPU0
# ALL-NEXT: [6] - JFPU1
# ALL-NEXT: [7] - JLAGU
# ALL-NEXT: [8] - JMul
# ALL-NEXT: [9] - JSAGU
# ALL-NEXT: [10] - JSTC
# ALL-NEXT: [11] - JVALU0
# ALL-NEXT: [12] - JVALU1
# ALL-NEXT: [13] - JVIMUL
# ALL: Resource pressure per iteration:
# ALL-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]
# ALL-NEXT: - 1.00 - - - - - - - - - - - -
# ALL: Resource pressure by instruction:
# ALL-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] Instructions:
# ALL-NEXT: - 1.00 - - - - - - - - - - - - callq printf

7
llvm/tools/llvm-mca/llvm-mca.cpp
View File

@@ -135,6 +135,11 @@ static cl::opt<unsigned>
"(instructions per cycle)"),
cl::cat(ToolOptions), cl::init(0));
static cl::opt<unsigned>
CallLatency("call-latency", cl::Hidden,
cl::desc("Number of cycles to assume for a call instruction"),
cl::cat(ToolOptions), cl::init(100U));
enum class SkipType { NONE, LACK_SCHED, PARSE_FAILURE, ANY_FAILURE };
static cl::opt<enum SkipType> SkipUnsupportedInstructions(
@@ -568,7 +573,7 @@ int main(int argc, char **argv) {
}
// Create an instruction builder.
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM, CallLatency);
// Create a context to control ownership of the pipeline hardware.
mca::Context MCA(*MRI, *STI);

2
llvm/unittests/tools/llvm-mca/MCATestBase.cpp
View File

@@ -66,7 +66,7 @@ Error MCATestBase::runBaselineMCA(json::Object &Result, ArrayRef<MCInst> Insts,
// Default InstrumentManager
auto IM = std::make_unique<mca::InstrumentManager>(*STI, *MCII);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM, /*CallLatency=*/100);
const SmallVector<mca::Instrument *> Instruments;
SmallVector<std::unique_ptr<mca::Instruction>> LoweredInsts;

4
llvm/unittests/tools/llvm-mca/X86/TestIncrementalMCA.cpp
View File

@@ -33,7 +33,7 @@ TEST_F(X86TestBase, TestResumablePipeline) {
P->addEventListener(SV.get());
auto IM = std::make_unique<mca::InstrumentManager>(*STI, *MCII);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM, /*CallLatency=*/100);
const SmallVector<mca::Instrument *> Instruments;
// Tile size = 7
@@ -124,7 +124,7 @@ TEST_F(X86TestBase, TestInstructionRecycling) {
// Default InstrumentManager
auto IM = std::make_unique<mca::InstrumentManager>(*STI, *MCII);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM);
mca::InstrBuilder IB(*STI, *MCII, *MRI, MCIA.get(), *IM, /*CallLatency=*/100);
IB.setInstRecycleCallback(GetRecycledInst);
const SmallVector<mca::Instrument *> Instruments;