aboutsummaryrefslogtreecommitdiffhomepage
path: root/tensorflow/stream_executor/module_spec.h
diff options
context:
space:
mode:
Diffstat (limited to 'tensorflow/stream_executor/module_spec.h')
-rw-r--r--tensorflow/stream_executor/module_spec.h65
1 files changed, 65 insertions, 0 deletions
diff --git a/tensorflow/stream_executor/module_spec.h b/tensorflow/stream_executor/module_spec.h
new file mode 100644
index 0000000000..212ae7ba9c
--- /dev/null
+++ b/tensorflow/stream_executor/module_spec.h
@@ -0,0 +1,65 @@
+/* Copyright 2018 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#ifndef TENSORFLOW_STREAM_EXECUTOR_MODULE_SPEC_H_
+#define TENSORFLOW_STREAM_EXECUTOR_MODULE_SPEC_H_
+
+#include "tensorflow/stream_executor/lib/array_slice.h"
+#include "tensorflow/stream_executor/lib/stringpiece.h"
+#include "tensorflow/stream_executor/platform/logging.h"
+#include "tensorflow/stream_executor/platform/port.h"
+
+namespace stream_executor {
+
+// Describes how to load a module on a target platform.
+//
+// The exact meaning of a "module" may differ from platform to platform but
+// loosely speaking a module a collection of kernels and global variables. It
+// corresponds to CUmodule when running on CUDA.
+class MultiModuleLoaderSpec {
+ public:
+ bool has_cuda_cubin_in_memory() const { return has_cuda_cubin_in_memory_; }
+ port::ArraySlice<const uint8> cuda_cubin_in_memory() const {
+ CHECK(has_cuda_cubin_in_memory());
+ return {cuda_cubin_in_memory_.data(), cuda_cubin_in_memory_.size()};
+ }
+
+ bool has_cuda_ptx_in_memory() const { return has_cuda_ptx_in_memory_; }
+ const char* cuda_ptx_in_memory() const {
+ CHECK(has_cuda_ptx_in_memory());
+ return cuda_ptx_in_memory_;
+ }
+
+ void AddCudaCubinInMemory(port::ArraySlice<const uint8> cubin_bytes) {
+ has_cuda_cubin_in_memory_ = true;
+ cuda_cubin_in_memory_ = cubin_bytes;
+ }
+
+ void AddCudaPtxInMemory(const char* ptx) {
+ has_cuda_ptx_in_memory_ = true;
+ // The CUDA driver does not like getting an empty string as PTX.
+ cuda_ptx_in_memory_ = *ptx ? ptx : nullptr;
+ }
+
+ private:
+ port::ArraySlice<const uint8> cuda_cubin_in_memory_;
+ bool has_cuda_cubin_in_memory_ = false;
+ const char* cuda_ptx_in_memory_;
+ bool has_cuda_ptx_in_memory_ = false;
+};
+
+} // namespace stream_executor
+
+#endif // TENSORFLOW_STREAM_EXECUTOR_MODULE_SPEC_H_