Refactor ProcessState in support of NUMA.

ProcessState is a singleton that anchors per-process resources.
Up until now that meant only GPU-related memory allocators
since CPU allocation was usually done directly from Allocator::cpu_allocator.
Accordingly process_state.h was in common_runtime/gpu and ProcesState
was only used in GPU builds.

With the upcoming introduction of NUMA node specific CPU allocators
it will be important that most of the TF runtime switch to requesting the
proper NUMA-specific CPU allocator.  These allocators will be owned by
and obtained from the ProcessState singleton which will exist in all
builds.  The GPU-specific functions are moved to a new
GPUProcessState, also a singleton.

PoolAllocator is also migrated out of common_rumntime/gpu into common_runtime.

PiperOrigin-RevId: 203002666

1 files changed, 4 insertions, 3 deletions

diff --git a/tensorflow/core/common_runtime/gpu/gpu_device_factory.cc b/tensorflow/core/common_runtime/gpu/gpu_device_factory.cc
index 9a000749c6..e1aaf95df6 100644
--- a/tensorflow/core/common_runtime/gpu/gpu_device_factory.cc
+++ b/tensorflow/core/common_runtime/gpu/gpu_device_factory.cc
@@ -19,7 +19,7 @@ limitations under the License.
 
 #include "tensorflow/core/common_runtime/gpu/gpu_device.h"
 #include "tensorflow/core/common_runtime/gpu/gpu_id.h"
-#include "tensorflow/core/common_runtime/gpu/process_state.h"
+#include "tensorflow/core/common_runtime/gpu/gpu_process_state.h"
 #include "tensorflow/core/common_runtime/threadpool_device.h"
 
 namespace tensorflow {
@@ -40,9 +40,10 @@ class GPUDevice : public BaseGPUDevice {
   }
 
   Allocator* GetAllocator(AllocatorAttributes attr) override {
+    CHECK(cpu_allocator_) << "bad place 1";
     if (attr.on_host()) {
       if (attr.gpu_compatible() || force_gpu_compatible_) {
-        ProcessState* ps = ProcessState::singleton();
+        GPUProcessState* ps = GPUProcessState::singleton();
         return ps->GetCUDAHostAllocator(0);
       } else {
         return cpu_allocator_;
@@ -90,7 +91,7 @@ class GPUCompatibleCPUDevice : public ThreadPoolDevice {
   ~GPUCompatibleCPUDevice() override {}
 
   Allocator* GetAllocator(AllocatorAttributes attr) override {
-    ProcessState* ps = ProcessState::singleton();
+    GPUProcessState* ps = GPUProcessState::singleton();
     if (attr.gpu_compatible() || force_gpu_compatible_) {
       return ps->GetCUDAHostAllocator(0);
     } else {