[llvm] f8f8598 - [Exegesis] Add the ability to dry-run the measurement phase (#121991)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Jan 9 09:25:55 PST 2025
Author: Min-Yih Hsu
Date: 2025-01-09T09:25:51-08:00
New Revision: f8f8598fd886cddfd374fa43eb6d7d37d301b576
URL: https://github.com/llvm/llvm-project/commit/f8f8598fd886cddfd374fa43eb6d7d37d301b576
DIFF: https://github.com/llvm/llvm-project/commit/f8f8598fd886cddfd374fa43eb6d7d37d301b576.diff
LOG: [Exegesis] Add the ability to dry-run the measurement phase (#121991)
With the new benchmark phase, `dry-run-measurement`, llvm-exegesis can
run everything except the actual snippet execution. It is useful when we
want to test some parts of the code between the `assemble-measured-code`
and `measure` phase without actually running on native platforms.
Added:
llvm/test/tools/llvm-exegesis/dry-run-measurement.test
Modified:
llvm/docs/CommandGuide/llvm-exegesis.rst
llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
llvm/tools/llvm-exegesis/lib/Target.cpp
llvm/tools/llvm-exegesis/llvm-exegesis.cpp
Removed:
################################################################################
diff --git a/llvm/docs/CommandGuide/llvm-exegesis.rst b/llvm/docs/CommandGuide/llvm-exegesis.rst
index 8266d891a5e6b1..d357c2ceea4189 100644
--- a/llvm/docs/CommandGuide/llvm-exegesis.rst
+++ b/llvm/docs/CommandGuide/llvm-exegesis.rst
@@ -301,6 +301,7 @@ OPTIONS
* ``prepare-and-assemble-snippet``: Same as ``prepare-snippet``, but also dumps an excerpt of the sequence (hex encoded).
* ``assemble-measured-code``: Same as ``prepare-and-assemble-snippet``. but also creates the full sequence that can be dumped to a file using ``--dump-object-to-disk``.
* ``measure``: Same as ``assemble-measured-code``, but also runs the measurement.
+ * ``dry-run-measurement``: Same as measure, but does not actually execute the snippet.
.. option:: --x86-lbr-sample-period=<nBranches/sample>
diff --git a/llvm/test/tools/llvm-exegesis/dry-run-measurement.test b/llvm/test/tools/llvm-exegesis/dry-run-measurement.test
new file mode 100644
index 00000000000000..e4449d7df3d826
--- /dev/null
+++ b/llvm/test/tools/llvm-exegesis/dry-run-measurement.test
@@ -0,0 +1,11 @@
+# RUN: llvm-exegesis --mtriple=riscv64 --mcpu=sifive-p470 --mode=latency --opcode-name=ADD --use-dummy-perf-counters --benchmark-phase=dry-run-measurement | FileCheck %s
+# REQUIRES: riscv-registered-target
+
+# This test makes sure that llvm-exegesis doesn't execute "cross-compiled" snippets in the presence of
+# --dry-run-measurement. RISC-V was chosen simply because most of the time we run tests on X86 machines.
+
+# Should not contain misleading results.
+# CHECK: measurements: []
+
+# Should not contain error messages like "snippet crashed while running: Segmentation fault".
+# CHECK: error: ''
diff --git a/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h b/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
index 3c09a8380146e5..5480d856168784 100644
--- a/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
+++ b/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
@@ -38,6 +38,7 @@ enum class BenchmarkPhaseSelectorE {
PrepareAndAssembleSnippet,
AssembleMeasuredCode,
Measure,
+ DryRunMeasure,
};
enum class BenchmarkFilter { All, RegOnly, WithMem };
diff --git a/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp b/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
index a7771b99e97b1a..cc46f7feb6cf7f 100644
--- a/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
+++ b/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
@@ -99,7 +99,7 @@ class InProcessFunctionExecutorImpl : public BenchmarkRunner::FunctionExecutor {
static Expected<std::unique_ptr<InProcessFunctionExecutorImpl>>
create(const LLVMState &State, object::OwningBinary<object::ObjectFile> Obj,
BenchmarkRunner::ScratchSpace *Scratch,
- std::optional<int> BenchmarkProcessCPU) {
+ std::optional<int> BenchmarkProcessCPU, bool DryRun) {
Expected<ExecutableFunction> EF =
ExecutableFunction::create(State.createTargetMachine(), std::move(Obj));
@@ -107,14 +107,17 @@ class InProcessFunctionExecutorImpl : public BenchmarkRunner::FunctionExecutor {
return EF.takeError();
return std::unique_ptr<InProcessFunctionExecutorImpl>(
- new InProcessFunctionExecutorImpl(State, std::move(*EF), Scratch));
+ new InProcessFunctionExecutorImpl(State, std::move(*EF), Scratch,
+ DryRun));
}
private:
InProcessFunctionExecutorImpl(const LLVMState &State,
ExecutableFunction Function,
- BenchmarkRunner::ScratchSpace *Scratch)
- : State(State), Function(std::move(Function)), Scratch(Scratch) {}
+ BenchmarkRunner::ScratchSpace *Scratch,
+ bool DryRun)
+ : State(State), Function(std::move(Function)), Scratch(Scratch),
+ DryRun(DryRun) {}
static void accumulateCounterValues(const SmallVector<int64_t, 4> &NewValues,
SmallVector<int64_t, 4> *Result) {
@@ -143,9 +146,14 @@ class InProcessFunctionExecutorImpl : public BenchmarkRunner::FunctionExecutor {
CrashRecoveryContext CRC;
CrashRecoveryContext::Enable();
const bool Crashed = !CRC.RunSafely([this, Counter, ScratchPtr]() {
- Counter->start();
- this->Function(ScratchPtr);
- Counter->stop();
+ if (DryRun) {
+ Counter->start();
+ Counter->stop();
+ } else {
+ Counter->start();
+ this->Function(ScratchPtr);
+ Counter->stop();
+ }
});
CrashRecoveryContext::Disable();
PS.reset();
@@ -177,6 +185,7 @@ class InProcessFunctionExecutorImpl : public BenchmarkRunner::FunctionExecutor {
const LLVMState &State;
const ExecutableFunction Function;
BenchmarkRunner::ScratchSpace *const Scratch;
+ bool DryRun = false;
};
#ifdef __linux__
@@ -664,6 +673,9 @@ Expected<std::unique_ptr<BenchmarkRunner::FunctionExecutor>>
BenchmarkRunner::createFunctionExecutor(
object::OwningBinary<object::ObjectFile> ObjectFile,
const BenchmarkKey &Key, std::optional<int> BenchmarkProcessCPU) const {
+ bool DryRun =
+ BenchmarkPhaseSelector == BenchmarkPhaseSelectorE::DryRunMeasure;
+
switch (ExecutionMode) {
case ExecutionModeE::InProcess: {
if (BenchmarkProcessCPU.has_value())
@@ -671,7 +683,8 @@ BenchmarkRunner::createFunctionExecutor(
"support benchmark core pinning.");
auto InProcessExecutorOrErr = InProcessFunctionExecutorImpl::create(
- State, std::move(ObjectFile), Scratch.get(), BenchmarkProcessCPU);
+ State, std::move(ObjectFile), Scratch.get(), BenchmarkProcessCPU,
+ DryRun);
if (!InProcessExecutorOrErr)
return InProcessExecutorOrErr.takeError();
@@ -679,6 +692,10 @@ BenchmarkRunner::createFunctionExecutor(
}
case ExecutionModeE::SubProcess: {
#ifdef __linux__
+ if (DryRun)
+ return make_error<Failure>("The subprocess execution mode cannot "
+ "dry-run measurement at this moment.");
+
auto SubProcessExecutorOrErr = SubProcessFunctionExecutorImpl::create(
State, std::move(ObjectFile), Key, BenchmarkProcessCPU);
if (!SubProcessExecutorOrErr)
diff --git a/llvm/tools/llvm-exegesis/lib/Target.cpp b/llvm/tools/llvm-exegesis/lib/Target.cpp
index 29e58692f0e92b..e2251ff978888b 100644
--- a/llvm/tools/llvm-exegesis/lib/Target.cpp
+++ b/llvm/tools/llvm-exegesis/lib/Target.cpp
@@ -98,7 +98,7 @@ ExegesisTarget::createBenchmarkRunner(
return nullptr;
case Benchmark::Latency:
case Benchmark::InverseThroughput:
- if (BenchmarkPhaseSelector == BenchmarkPhaseSelectorE::Measure &&
+ if (BenchmarkPhaseSelector >= BenchmarkPhaseSelectorE::Measure &&
!PfmCounters.CycleCounter) {
const char *ModeName = Mode == Benchmark::Latency
? "latency"
@@ -116,7 +116,7 @@ ExegesisTarget::createBenchmarkRunner(
State, Mode, BenchmarkPhaseSelector, ResultAggMode, ExecutionMode,
ValidationCounters, BenchmarkRepeatCount);
case Benchmark::Uops:
- if (BenchmarkPhaseSelector == BenchmarkPhaseSelectorE::Measure &&
+ if (BenchmarkPhaseSelector >= BenchmarkPhaseSelectorE::Measure &&
!PfmCounters.UopsCounter && !PfmCounters.IssueCounters)
return make_error<Failure>(
"can't run 'uops' mode, sched model does not define uops or issue "
diff --git a/llvm/tools/llvm-exegesis/llvm-exegesis.cpp b/llvm/tools/llvm-exegesis/llvm-exegesis.cpp
index fa37e05956be8c..07bd44ee64f1f2 100644
--- a/llvm/tools/llvm-exegesis/llvm-exegesis.cpp
+++ b/llvm/tools/llvm-exegesis/llvm-exegesis.cpp
@@ -132,7 +132,10 @@ static cl::opt<BenchmarkPhaseSelectorE> BenchmarkPhaseSelector(
clEnumValN(
BenchmarkPhaseSelectorE::Measure, "measure",
"Same as prepare-measured-code, but also runs the measurement "
- "(default)")),
+ "(default)"),
+ clEnumValN(
+ BenchmarkPhaseSelectorE::DryRunMeasure, "dry-run-measurement",
+ "Same as measure, but does not actually execute the snippet")),
cl::init(BenchmarkPhaseSelectorE::Measure));
static cl::opt<bool>
@@ -476,7 +479,7 @@ static void runBenchmarkConfigurations(
}
void benchmarkMain() {
- if (BenchmarkPhaseSelector == BenchmarkPhaseSelectorE::Measure &&
+ if (BenchmarkPhaseSelector >= BenchmarkPhaseSelectorE::Measure &&
!UseDummyPerfCounters) {
#ifndef HAVE_LIBPFM
ExitWithError(
@@ -501,7 +504,7 @@ void benchmarkMain() {
// Preliminary check to ensure features needed for requested
// benchmark mode are present on target CPU and/or OS.
- if (BenchmarkPhaseSelector == BenchmarkPhaseSelectorE::Measure)
+ if (BenchmarkPhaseSelector >= BenchmarkPhaseSelectorE::Measure)
ExitOnErr(State.getExegesisTarget().checkFeatureSupport());
if (ExecutionMode == BenchmarkRunner::ExecutionModeE::SubProcess &&
More information about the llvm-commits
mailing list