Skip to content
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.

Sign up for GitHub

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

[XLA:GPU] Turn --xla_gpu_triton_gemm_any on by default. #22515

Merged
merged 1 commit into from
Feb 10, 2025
Merged

[XLA:GPU] Turn --xla_gpu_triton_gemm_any on by default. #22515

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 0 additions & 4 deletions xla/backends/gpu/codegen/triton/fusion_emitter_int4_device_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,6 @@ limitations under the License.
#include <vector>

#include <gtest/gtest.h>
#include "absl/log/log.h"
#include "absl/strings/str_cat.h"
#include "absl/strings/str_replace.h"
#include "absl/strings/str_split.h"
Expand Down Expand Up @@ -720,7 +719,6 @@ INSTANTIATE_TEST_SUITE_P(ParametrizedTritonTest, ParametrizedTritonTest,
I4TestParams::ToString);

TEST_F(TritonTest, NonstandardLayoutWithManyNonContractingDims) {
// We cannot do triton_gemm and we use cuBLAS instead.
constexpr absl::string_view kHloText = R"(
HloModule NonstandardLayoutWithManyNonContractingDims

Expand All @@ -734,7 +732,6 @@ TEST_F(TritonTest, NonstandardLayoutWithManyNonContractingDims) {
)";

TF_ASSERT_OK_AND_ASSIGN(auto module, GetOptimizedModule(kHloText));
EXPECT_TRUE(*RunFileCheck(module->ToString(), R"(CHECK: "__cublas$gemm")"));
EXPECT_TRUE(RunAndCompare(kHloText, ErrorSpec{/*aabs=*/1e-3, /*arel=*/1e-2}));
}

Expand All @@ -753,7 +750,6 @@ TEST_F(TritonTest, NonstandardLayoutWithManyNonContractingDimsReversedLayout) {
)";

TF_ASSERT_OK_AND_ASSIGN(auto module, GetOptimizedModule(kHloText));
EXPECT_TRUE(*RunFileCheck(module->ToString(), R"(CHECK: "__cublas$gemm")"));
EXPECT_TRUE(RunAndCompare(kHloText, ErrorSpec{/*aabs=*/1e-3, /*arel=*/1e-3}));
}

Expand Down
2 changes: 1 addition & 1 deletion xla/debug_options_flags.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -209,7 +209,7 @@ DebugOptions DefaultDebugOptionsIgnoringFlags() {
opts.set_xla_gpu_enable_triton_gemm(true);
opts.set_xla_gpu_unsupported_enable_generic_triton_emitter_for_gemms(false);
opts.set_xla_gpu_enable_cudnn_int8x32_convolution_reordering(true);
opts.set_xla_gpu_triton_gemm_any(false);
opts.set_xla_gpu_triton_gemm_any(true);
opts.set_xla_gpu_verify_triton_fusion_numerics(false);

// Moving reduce-scatter out of while loops can increase memory footprint, so
Expand Down
Loading