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

gpu_blas_lt_gemm_runner #2766

Merged
merged 4 commits into from
Jan 2, 2025
Merged

gpu_blas_lt_gemm_runner #2766

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
e538a1c
gpu_blas_lt_gemm_runner
ScXfjiang Oct 25, 2024
b7d31bd
change the location of gemm runner for Batched GEMM
ScXfjiang Nov 21, 2024
7630bad
use xla flags to enable hipblaslt instead of env vars
ScXfjiang Nov 22, 2024
8ee6c6d
non-xla hipblaslt for ThenBlasGemmStridedBatched
ScXfjiang Nov 22, 2024
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
2 changes: 1 addition & 1 deletion tensorflow/compiler/xla/service/computation_placer.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -163,7 +163,7 @@ StatusOr<DeviceAssignment> ComputationPlacer::AssignDevices(
ComputationPlacerCreationFunction creation_function) {
absl::MutexLock lock(&ComputationPlacer::platform_computation_placer_mutex_);
auto* computation_placers = GetPlatformComputationPlacers();
CHECK(computation_placers->find(platform_id) == computation_placers->end());
// CHECK(computation_placers->find(platform_id) == computation_placers->end());
i-chaochen marked this conversation as resolved.
Show resolved Hide resolved
(*computation_placers)[platform_id].creation_function = creation_function;
}

Expand Down
1 change: 1 addition & 0 deletions tensorflow/compiler/xla/stream_executor/BUILD
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -450,6 +450,7 @@ tsl_gpu_library(
":temporary_memory_manager",
":timer",
"//tensorflow/compiler/xla/stream_executor/platform",
"//tensorflow/compiler/xla/stream_executor/gpu:gpu_blas_lt_gemm_runner",
"//tensorflow/tsl/platform:env",
"//tensorflow/tsl/platform:errors",
"//tensorflow/tsl/platform:logging",
Expand Down
16 changes: 9 additions & 7 deletions tensorflow/compiler/xla/stream_executor/gpu/BUILD
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,6 +78,7 @@ cc_library(
#"//tensorflow/core/platform:env",
"//tensorflow/tsl/util:env_var",
"@com_google_absl//absl/types:any",
"//tensorflow/compiler/xla:debug_options_flags",
]),
)

Expand All @@ -87,13 +88,14 @@ cc_library(
srcs = if_gpu_is_configured(["gpu_blas_lt_gemm_runner.cc"]),
hdrs = if_gpu_is_configured(["gpu_blas_lt_gemm_runner.h"]),
deps = if_gpu_is_configured([
"//tensorflow/core:autotuning_proto_cc",
"//tensorflow/core:autotune_results_proto_cc",
"//tensorflow/compiler/xla:xla_proto",
"//tensorflow/compiler/xla/stream_executor:scratch_allocator",
"//tensorflow/compiler/xla/service/gpu:autotuner_util",
"//tensorflow/compiler/xla:debug_options_flags",
":gpu_blas_lt",
"//tensorflow/core/protobuf:autotuning_proto_cc",
"//tensorflow/compiler/xla:autotune_results_proto_cc",
"//tensorflow/compiler/xla:xla_proto_cc",
"//tensorflow/compiler/xla:xla_data_proto_cc",
"//tensorflow/compiler/xla/stream_executor:scratch_allocator",
"//tensorflow/compiler/xla/service/gpu:autotuner_util",
"//tensorflow/compiler/xla:debug_options_flags",
":gpu_blas_lt",
]),
)

Expand Down
8 changes: 8 additions & 0 deletions tensorflow/compiler/xla/stream_executor/gpu/gpu_blas_lt.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@ limitations under the License.
#include "tensorflow/compiler/xla/stream_executor/stream_executor.h"
#include "tensorflow/compiler/xla/util.h"
#include "tensorflow/tsl/util/env_var.h"
#include "tensorflow/compiler/xla/debug_options_flags.h"

namespace stream_executor {

Expand All @@ -31,6 +32,13 @@ using blas::ComputationType;
using blas::DataType;
using xla::PrimitiveType;

bool GpuBlasLtEnabled() {
ScXfjiang marked this conversation as resolved.
Show resolved Hide resolved
static std::atomic_bool result{[] {
return xla::GetDebugOptionsFromFlags().xla_gpu_enable_cublaslt();
}()};
return result;
}

namespace {

bool TF32_Enabled() {
Expand Down
2 changes: 2 additions & 0 deletions tensorflow/compiler/xla/stream_executor/gpu/gpu_blas_lt.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,8 @@ namespace stream_executor {

namespace gpu {

bool GpuBlasLtEnabled();

xla::StatusOr<blas::DataType> AsBlasDataType(xla::PrimitiveType dtype);

xla::StatusOr<blas::ComputationType> GetBlasComputationType(
Expand Down
Loading
Loading