Fix a segfaul introduced in 76e6109. sub_allocator is owned by gpu_bfc_allocator. deleting it early doesn't any real benefit. So lets the normal clean up do its work.

nouiz · nouiz · commit 07e30d3408a0 · 2021-06-08T13:16:43.000-07:00
diff --git a/tensorflow/core/common_runtime/gpu/gpu_process_state.cc b/tensorflow/core/common_runtime/gpu/gpu_process_state.cc
@@ -206,10 +206,6 @@ Allocator* GPUProcessState::GetGPUAllocator(
       // If true, passes all allocation requests through to cudaMalloc
       // useful for doing memory debugging with tools like cuda-memcheck
       // **WARNING** probably will not work in a multi-gpu scenario
-      delete gpu_bfc_allocator;
-      delete sub_allocator;
-      gpu_bfc_allocator = nullptr;
-      sub_allocator = nullptr;
       gpu_allocator = new GPUcudaMallocAllocator(platform_device_id);
     } else if (UseCudaMallocAsyncAllocator() ||
                options.experimental().use_cuda_malloc_async()) {
@@ -219,10 +215,6 @@ Allocator* GPUProcessState::GetGPUAllocator(
       // TODO: useful for doing memory debugging with tools like
       // compute-sanitizer.
       // TODO: **WARNING** probably will not work in a multi-gpu scenario
-      delete gpu_bfc_allocator;
-      delete sub_allocator;
-      gpu_bfc_allocator = nullptr;
-      sub_allocator = nullptr;
       gpu_allocator =
           new GpuCudaMallocAsyncAllocator(platform_device_id, total_bytes);
     }
