Remove guards from gpu_lt_matmul_thunk target (NFC).

Those should not be necessary. PiperOrigin-RevId: 649579418
Intel-tensorflow · Jul 5, 2024 · fbee6d7 · fbee6d7
1 parent 35e0457
commit fbee6d7
Show file tree

Hide file tree

Showing 3 changed files with 12 additions and 19 deletions.
diff --git a/third_party/xla/xla/service/gpu/BUILD b/third_party/xla/xla/service/gpu/BUILD
@@ -355,6 +355,7 @@ cc_library(
         "//xla/service/gpu/runtime:fft_thunk",
         "//xla/service/gpu/runtime:fused_mha_thunk",
         "//xla/service/gpu/runtime:gemm_thunk",
+        "//xla/service/gpu/runtime:gpublas_lt_matmul_thunk",
         "//xla/service/gpu/runtime:infeed_thunk",
         "//xla/service/gpu/runtime:kernel_thunk",
         "//xla/service/gpu/runtime:nccl_all_gather_thunk",
@@ -420,7 +421,6 @@ cc_library(
         "//xla/service/gpu/fusions",
         "//xla/service/gpu/runtime:cholesky_thunk",
         "//xla/service/gpu/runtime:cub_sort_thunk",
-        "//xla/service/gpu/runtime:gpublas_lt_matmul_thunk",
         "//xla/service/gpu/runtime:triangular_solve_thunk",
     ]) + if_rocm_is_configured([
         "@local_config_rocm//rocm:rocm_headers",

diff --git a/third_party/xla/xla/service/gpu/ir_emitter_unnested.cc b/third_party/xla/xla/service/gpu/ir_emitter_unnested.cc
@@ -123,6 +123,7 @@ limitations under the License.
 #include "xla/service/gpu/runtime/fft_thunk.h"
 #include "xla/service/gpu/runtime/fused_mha_thunk.h"
 #include "xla/service/gpu/runtime/gemm_thunk.h"
+#include "xla/service/gpu/runtime/gpublas_lt_matmul_thunk.h"
 #include "xla/service/gpu/runtime/infeed_thunk.h"
 #include "xla/service/gpu/runtime/kernel_thunk.h"
 #include "xla/service/gpu/runtime/nccl_all_gather_thunk.h"
@@ -167,10 +168,6 @@ limitations under the License.
 #include "tsl/protobuf/dnn.pb.h"
 #include "triton/Dialect/Triton/IR/Dialect.h"
 
-#if GOOGLE_CUDA || TF_HIPBLASLT
-#include "xla/service/gpu/runtime/gpublas_lt_matmul_thunk.h"
-#endif  // GOOGLE_CUDA || TF_HIPBLASLT
-
 #if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
 #include "xla/service/gpu/ir_emitter_triton.h"
 #include "xla/service/gpu/runtime/cholesky_thunk.h"

diff --git a/third_party/xla/xla/service/gpu/runtime/BUILD b/third_party/xla/xla/service/gpu/runtime/BUILD
@@ -686,26 +686,22 @@ cc_library(
 
 cc_library(
     name = "gpublas_lt_matmul_thunk",
-    srcs = if_gpu_is_configured(["gpublas_lt_matmul_thunk.cc"]),
-    hdrs = if_gpu_is_configured(["gpublas_lt_matmul_thunk.h"]),
-    local_defines = if_cuda_is_configured(["GOOGLE_CUDA=1"]) + if_rocm_is_configured([
-        "TENSORFLOW_USE_ROCM=1",
-    ]),
-    deps = if_gpu_is_configured([
-        "@com_google_absl//absl/base:core_headers",
-        "@com_google_absl//absl/container:flat_hash_map",
-        "@com_google_absl//absl/synchronization",
+    srcs = ["gpublas_lt_matmul_thunk.cc"],
+    hdrs = ["gpublas_lt_matmul_thunk.h"],
+    deps = [
+        "//xla:status_macros",
         "//xla/service:buffer_assignment",
+        "//xla/service/gpu:buffer_allocations",
         "//xla/service/gpu:matmul_utils",
         "//xla/service/gpu/runtime:thunk",
-        "//xla/stream_executor:device_memory",
         "//xla/stream_executor",
+        "//xla/stream_executor:device_memory",
         "//xla/stream_executor/gpu:gpu_blas_lt",
-        "@local_tsl//tsl/platform:logging",
-    ]) + [
-        "//xla:status_macros",
-        "//xla/service/gpu:buffer_allocations",
+        "@com_google_absl//absl/base:core_headers",
+        "@com_google_absl//absl/container:flat_hash_map",
         "@com_google_absl//absl/status",
+        "@com_google_absl//absl/synchronization",
+        "@local_tsl//tsl/platform:logging",
         "@local_tsl//tsl/platform:statusor",
     ],
 )