Skip collective tests for ROCm

hsharsha · hsharsha · commit 9482582cc4a1 · 2025-02-10T17:56:32.000Z
diff --git a/third_party/xla/xla/stream_executor/gpu/gpu_executor_test.cc b/third_party/xla/xla/stream_executor/gpu/gpu_executor_test.cc
@@ -56,6 +56,8 @@ TEST_F(GetPointerMemorySpaceTest, Device) {
 }
 
 TEST_F(GetPointerMemorySpaceTest, Collective) {
+  // TODO (rocm) Remove this once https://github.com/openxla/xla/pull/22532 is merged
+  GTEST_SKIP() << "Collectives are not enabled for ROCm";
   StreamExecutor* executor = GetPlatform()->ExecutorForDevice(0).value();
   auto mem =
       executor->Allocate(64, static_cast<int64_t>(MemoryType::kCollective));

Original file line number	Diff line number	Diff line change
`@@ -56,6 +56,8 @@ TEST_F(GetPointerMemorySpaceTest, Device) {`
`56`	`56`	`}`
`57`	`57`
`58`	`58`	`TEST_F(GetPointerMemorySpaceTest, Collective) {`
	`59`	`+ // TODO (rocm) Remove this once https://github.com/openxla/xla/pull/22532 is merged`
	`60`	`+ GTEST_SKIP() << "Collectives are not enabled for ROCm";`
`59`	`61`	`StreamExecutor* executor = GetPlatform()->ExecutorForDevice(0).value();`
`60`	`62`	`auto mem =`
`61`	`63`	`executor->Allocate(64, static_cast<int64_t>(MemoryType::kCollective));`