-
Notifications
You must be signed in to change notification settings - Fork 14.3k
[libc][math] Add performance tests for fmul and fmull. #106262
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
@llvm/pr-subscribers-libc Author: Job Henandez Lara (Jobhdez) ChangesFull diff: https://github.com/llvm/llvm-project/pull/106262.diff 3 Files Affected:
diff --git a/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h b/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h
index 63d9768e21899b..5941aa71689a59 100644
--- a/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h
+++ b/libc/test/src/math/performance_testing/BinaryOpSingleOutputPerf.h
@@ -122,6 +122,111 @@ template <typename T> class BinaryOpSingleOutputPerf {
}
};
+template <typename T, typename R> class BinaryOpSingleDifferentTypeOutputPerf {
+ using FPBits = fputil::FPBits<T>;
+ using StorageType = typename FPBits::StorageType;
+ static constexpr StorageType UIntMax =
+ cpp::numeric_limits<StorageType>::max();
+
+public:
+ typedef T Func(R, R);
+
+ static void run_perf_in_range(Func myFunc, Func otherFunc,
+ StorageType startingBit, StorageType endingBit,
+ size_t N, size_t rounds, std::ofstream &log) {
+ if (sizeof(StorageType) <= sizeof(size_t))
+ N = cpp::min(N, static_cast<size_t>(endingBit - startingBit));
+
+ auto runner = [=](Func func) {
+ [[maybe_unused]] volatile T result;
+ if (endingBit < startingBit) {
+ return;
+ }
+
+ StorageType step = (endingBit - startingBit) / N;
+ for (size_t i = 0; i < rounds; i++) {
+ for (StorageType bitsX = startingBit, bitsY = endingBit;;
+ bitsX += step, bitsY -= step) {
+ R x = FPBits(bitsX).get_val();
+ R y = FPBits(bitsY).get_val();
+ result = func(x, y);
+ if (endingBit - bitsX < step) {
+ break;
+ }
+ }
+ }
+ };
+
+ Timer timer;
+ timer.start();
+ runner(myFunc);
+ timer.stop();
+
+ double my_average = static_cast<double>(timer.nanoseconds()) / N / rounds;
+ log << "-- My function --\n";
+ log << " Total time : " << timer.nanoseconds() << " ns \n";
+ log << " Average runtime : " << my_average << " ns/op \n";
+ log << " Ops per second : "
+ << static_cast<uint64_t>(1'000'000'000.0 / my_average) << " op/s \n";
+
+ timer.start();
+ runner(otherFunc);
+ timer.stop();
+
+ double other_average =
+ static_cast<double>(timer.nanoseconds()) / N / rounds;
+ log << "-- Other function --\n";
+ log << " Total time : " << timer.nanoseconds() << " ns \n";
+ log << " Average runtime : " << other_average << " ns/op \n";
+ log << " Ops per second : "
+ << static_cast<uint64_t>(1'000'000'000.0 / other_average) << " op/s \n";
+
+ log << "-- Average runtime ratio --\n";
+ log << " Mine / Other's : " << my_average / other_average << " \n";
+ }
+
+ static void run_perf(Func myFunc, Func otherFunc, int rounds,
+ const char *logFile) {
+ std::ofstream log(logFile);
+ log << " Performance tests with inputs in denormal range:\n";
+ run_perf_in_range(myFunc, otherFunc, /* startingBit= */ StorageType(0),
+ /* endingBit= */ FPBits::max_subnormal().uintval(),
+ 1'000'001, rounds, log);
+ log << "\n Performance tests with inputs in normal range:\n";
+ run_perf_in_range(myFunc, otherFunc,
+ /* startingBit= */ FPBits::min_normal().uintval(),
+ /* endingBit= */ FPBits::max_normal().uintval(),
+ 1'000'001, rounds, log);
+ log << "\n Performance tests with inputs in normal range with exponents "
+ "close to each other:\n";
+ run_perf_in_range(myFunc, otherFunc,
+ /* startingBit= */ FPBits(T(0x1.0p-10)).uintval(),
+ /* endingBit= */ FPBits(T(0x1.0p+10)).uintval(),
+ 1'000'001, rounds, log);
+ }
+
+ static void run_diff(Func myFunc, Func otherFunc, const char *logFile) {
+ uint64_t diffCount = 0;
+ std::ofstream log(logFile);
+ log << " Diff tests with inputs in denormal range:\n";
+ diffCount += run_diff_in_range(
+ myFunc, otherFunc, /* startingBit= */ StorageType(0),
+ /* endingBit= */ FPBits::max_subnormal().uintval(), 1'000'001, log);
+ log << "\n Diff tests with inputs in normal range:\n";
+ diffCount += run_diff_in_range(
+ myFunc, otherFunc,
+ /* startingBit= */ FPBits::min_normal().uintval(),
+ /* endingBit= */ FPBits::max_normal().uintval(), 100'000'001, log);
+ log << "\n Diff tests with inputs in normal range with exponents "
+ "close to each other:\n";
+ diffCount += run_diff_in_range(
+ myFunc, otherFunc, /* startingBit= */ FPBits(T(0x1.0p-10)).uintval(),
+ /* endingBit= */ FPBits(T(0x1.0p+10)).uintval(), 10'000'001, log);
+
+ log << "Total number of differing results: " << diffCount << '\n';
+ }
+};
+
} // namespace testing
} // namespace LIBC_NAMESPACE_DECL
@@ -140,3 +245,13 @@ template <typename T> class BinaryOpSingleOutputPerf {
LIBC_NAMESPACE::testing::BinaryOpSingleOutputPerf<T>::run_perf( \
&myFunc, &otherFunc, rounds, filename); \
}
+
+#define BINARY_OP_SINGLE_DIFFERENT_TYPE_OUTPUT_PERF_EX( \
+ T, R, myFunc, otherFunc, rounds, filename) \
+ \
+ { \
+ LIBC_NAMESPACE::testing::BinaryOpSingleDifferentTypeOutputPerf< \
+ T, R>::run_perf(&myFunc, &otherFunc, rounds, filename); \
+ LIBC_NAMESPACE::testing::BinaryOpSingleDifferentTypeOutputPerf< \
+ T, R>::run_perf(&myFunc, &otherFunc, rounds, filename); \
+ }
diff --git a/libc/test/src/math/performance_testing/CMakeLists.txt b/libc/test/src/math/performance_testing/CMakeLists.txt
index 8e529ca09ed797..207a463fe185f8 100644
--- a/libc/test/src/math/performance_testing/CMakeLists.txt
+++ b/libc/test/src/math/performance_testing/CMakeLists.txt
@@ -476,3 +476,15 @@ add_perf_binary(
COMPILE_OPTIONS
-fno-builtin
)
+
+add_perf_binary(
+ fmul_perf
+ SRCS
+ fmul_perf.cpp
+ DEPENDS
+ .binary_op_single_output_diff
+ libc.src.math.fmul
+ libc.src.math.fmull
+ COMPILE_OPTIONS
+ -fno-builtin
+)
diff --git a/libc/test/src/math/performance_testing/fmul_perf.cpp b/libc/test/src/math/performance_testing/fmul_perf.cpp
new file mode 100644
index 00000000000000..9e692b1848602e
--- /dev/null
+++ b/libc/test/src/math/performance_testing/fmul_perf.cpp
@@ -0,0 +1,33 @@
+//===-- Performance test for maximum and minimum functions ----------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "BinaryOpSingleOutputPerf.h"
+#include "src/math/fmul.h"
+#include "src/math/fmull.h"
+#include <math.h>
+
+static constexpr size_t DOUBLE_ROUNDS = 40;
+static constexpr size_t LONG_DOUBLE_ROUNDS = 40;
+
+float fmul_placeholder_binary(double x, double y) {
+ return static_cast<float>(x * y);
+}
+
+float fmull_placeholder_binary(long double x, long double y) {
+ return static_cast<float>(x * y);
+}
+
+int main() {
+ BINARY_OP_SINGLE_DIFFERENT_TYPE_OUTPUT_PERF_EX(
+ float, double, LIBC_NAMESPACE::fmul, fmul_placeholder_binary,
+ DOUBLE_ROUNDS, "fmul_perf.log")
+ BINARY_OP_SINGLE_DIFFERENT_TYPE_OUTPUT_PERF_EX(
+ float, long double, LIBC_NAMESPACE::fmull, fmull_placeholder_binary,
+ LONG_DOUBLE_ROUNDS, "fmull_perf.log")
+ return 0;
+}
|
lntue
reviewed
Aug 28, 2024
lntue
reviewed
Aug 28, 2024
lntue
reviewed
Aug 28, 2024
lntue
reviewed
Aug 28, 2024
lntue
approved these changes
Aug 29, 2024
qiaojbao
pushed a commit
to GPUOpen-Drivers/llvm-project
that referenced
this pull request
Sep 30, 2024
…c02e5ed6f Local branch amd-gfx a9dc02e Merged main:e05c22484efb5c767115525adfa4273e48b1ae26 into amd-gfx:4744f1225122 Remote branch main 1ace91f [libc][math] Add performance tests for fmul and fmull. (llvm#106262)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.