experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Copying cuBLAS and cuDNN headers into separate directories.

Browse Source 
Make {CUDA|cuDNN} headers available under third_party/gpus/{cuda|cudnn}/ instead of cuda/

PiperOrigin-RevId: 247177891
This commit is contained in:
A. Unique TensorFlower 2019-05-08 02:36:07 -07:00 committed by TensorFlower Gardener
parent 5309596211
commit d58b53e19c
60 changed files with 203 additions and 270 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorflow/compiler/tf2tensorrt/convert/convert_graph.cc
View File

@ -57,7 +57,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorrt/include/NvInfer.h"
namespace tensorflow {
namespace tensorrt {

4
tensorflow/compiler/tf2tensorrt/convert/convert_nodes_test.cc
View File

@ -50,8 +50,8 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda.h"
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorrt/include/NvInfer.h"
namespace tensorflow {

2
tensorflow/compiler/tf2tensorrt/kernels/trt_engine_op.cc
View File

@ -41,7 +41,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorrt/include/NvInfer.h"
namespace tensorflow {

2
tensorflow/compiler/tf2tensorrt/kernels/trt_engine_op_test.cc
View File

@ -32,7 +32,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
namespace tensorflow {
namespace tensorrt {

4
tensorflow/compiler/tf2tensorrt/tensorrt_test.cc
View File

@ -20,8 +20,8 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda.h"
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorrt/include/NvInfer.h"
namespace tensorflow {

2
tensorflow/compiler/tf2tensorrt/utils/trt_allocator.cc
View File

@ -19,7 +19,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#endif // GOOGLE_TENSORRT
#endif // GOOGLE_CUDA

2
tensorflow/compiler/tf2tensorrt/utils/trt_int8_calibrator.cc
View File

@ -22,7 +22,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
namespace tensorflow {
namespace tensorrt {

2
tensorflow/compiler/tf2tensorrt/utils/trt_int8_calibrator.h
View File

@ -25,7 +25,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorrt/include/NvInfer.h"
namespace tensorflow {

4
tensorflow/compiler/xla/service/gpu/cusolver_context.h
View File

@ -18,8 +18,8 @@ limitations under the License.
#include <complex>
#include "cuda/include/cublas_v2.h"
#include "cuda/include/cusolverDn.h"
#include "third_party/gpus/cuda/include/cublas_v2.h"
#include "third_party/gpus/cuda/include/cusolverDn.h"
#include "tensorflow/compiler/xla/statusor.h"
#include "tensorflow/compiler/xla/types.h"
#include "tensorflow/compiler/xla/util.h"

2
tensorflow/contrib/fused_conv/kernels/fused_conv2d_bias_activation_op.cc
View File

@ -42,7 +42,7 @@ limitations under the License.
#if GOOGLE_CUDA
#include "google/protobuf/duration.pb.h"
#include "absl/time/time.h"
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/core/kernels/conv_ops_gpu.h"
#include "tensorflow/core/platform/logger.h"
#include "tensorflow/core/platform/stream_executor.h"

2
tensorflow/contrib/tensorrt/custom_plugin_examples/inc_op_kernel.cu.cc
View File

@ -21,7 +21,7 @@ limitations under the License.
#include <vector>
#define EIGEN_USE_GPU
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/platform/stream_executor.h"
#include "tensorflow/core/util/gpu_launch_config.h"

2
tensorflow/contrib/tensorrt/custom_plugin_examples/inc_op_kernel.h
View File

@ -18,7 +18,7 @@ limitations under the License.
#if GOOGLE_CUDA
#if GOOGLE_TENSORRT
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
namespace tensorflow {
namespace tensorrt {

4
tensorflow/core/common_runtime/direct_session_test.cc
View File

@ -50,8 +50,8 @@ limitations under the License.
#include "tensorflow/core/util/device_name_utils.h"
#ifdef GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#endif // GOOGLE_CUDA
namespace tensorflow {

2
tensorflow/core/common_runtime/gpu/gpu_cudamalloc_allocator.cc
View File

@ -14,7 +14,7 @@ limitations under the License.
==============================================================================*/
#ifdef GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/stream_executor/cuda/cuda_activation.h"
#endif // GOOGLE_CUDA

2
tensorflow/core/common_runtime/gpu/gpu_device.cc
View File

@ -78,7 +78,7 @@ limitations under the License.
#if !defined(PLATFORM_GOOGLE)
#if GOOGLE_CUDA
#include "cuda/cuda_config.h"
#include "third_party/gpus/cuda/cuda_config.h"
#endif
#endif

2
tensorflow/core/common_runtime/gpu/gpu_managed_allocator.cc
View File

@ -14,7 +14,7 @@ limitations under the License.
==============================================================================*/
#ifdef GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#define EIGEN_USE_GPU
#endif

4
tensorflow/core/common_runtime/process_function_library_runtime_test.cc
View File

@ -34,8 +34,8 @@ limitations under the License.
#include "tensorflow/core/public/version.h"
#ifdef GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#endif // GOOGLE_CUDA
namespace tensorflow {

6
tensorflow/core/grappler/clusters/utils.cc
View File

@ -18,9 +18,9 @@ limitations under the License.
#include "third_party/eigen3/Eigen/Core"
#if GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "cuda/include/cuda_runtime_api.h"
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cudnn/cudnn.h"
#endif
#if TENSORFLOW_USE_ROCM

2
tensorflow/core/grappler/optimizers/auto_mixed_precision_lists.h
View File

@ -22,7 +22,7 @@ limitations under the License.
#if GOOGLE_CUDA
// Needed for CUDA_VERSION macro.
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#endif // GOOGLE_CUDA
namespace tensorflow {

2
tensorflow/core/kernels/batch_matmul_op_real.cc
View File

@ -16,7 +16,7 @@ limitations under the License.
#include "tensorflow/core/kernels/batch_matmul_op_impl.h"
#if GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#endif // GOOGLE_CUDA
namespace tensorflow {

2
tensorflow/core/kernels/conv_2d_gpu.h
View File

@ -25,7 +25,7 @@ limitations under the License.
#include <limits>
#include <utility>
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/core/framework/register_types.h"
#include "tensorflow/core/kernels/conv_2d.h"
#include "tensorflow/core/lib/math/math_util.h"

2
tensorflow/core/kernels/conv_ops_fused_impl.h
View File

@ -57,7 +57,7 @@ limitations under the License.
#include "tensorflow/core/util/use_cudnn.h"
#if GOOGLE_CUDA
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/core/kernels/conv_ops_gpu.h"
#include "tensorflow/core/platform/stream_executor.h"
#include "tensorflow/core/util/proto/proto_utils.h"

4
tensorflow/core/kernels/cuda_solvers.cc
View File

@ -21,8 +21,8 @@
#include <unordered_map>
#include <vector>
#include "cuda/include/cublas_v2.h"
#include "cuda/include/cusolverDn.h"
#include "third_party/gpus/cuda/include/cublas_v2.h"
#include "third_party/gpus/cuda/include/cusolverDn.h"
#include "tensorflow/core/common_runtime/gpu/gpu_event_mgr.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/types.h"

4
tensorflow/core/kernels/cuda_solvers.h
View File

@ -26,8 +26,8 @@ limitations under the License.
#include <functional>
#include <vector>
#include "cuda/include/cublas_v2.h"
#include "cuda/include/cusolverDn.h"
#include "third_party/gpus/cuda/include/cublas_v2.h"
#include "third_party/gpus/cuda/include/cusolverDn.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/lib/core/status.h"

2
tensorflow/core/kernels/cuda_sparse.cc
View File

@ -22,7 +22,7 @@
#include <utility>
#include <vector>
#include "cuda/include/cusparse.h"
#include "third_party/gpus/cuda/include/cusparse.h"
#include "tensorflow/core/common_runtime/gpu/gpu_event_mgr.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/types.h"

2
tensorflow/core/kernels/cuda_sparse.h
View File

@ -25,7 +25,7 @@ limitations under the License.
#include <functional>
#include <vector>
#include "cuda/include/cusparse.h"
#include "third_party/gpus/cuda/include/cusparse.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/tensor_types.h"

2
tensorflow/core/kernels/depthwise_conv_grad_op.cc
View File

@ -38,7 +38,7 @@ limitations under the License.
#include "tensorflow/core/util/work_sharder.h"
#if GOOGLE_CUDA
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/core/platform/stream_executor.h"
#endif // GOOGLE_CUDA

2
tensorflow/core/kernels/depthwise_conv_op.cc
View File

@ -39,7 +39,7 @@ limitations under the License.
#include "tensorflow/core/util/work_sharder.h"
#if GOOGLE_CUDA
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/core/platform/stream_executor.h"
#endif // GOOGLE_CUDA

2
tensorflow/core/kernels/fused_batch_norm_op.cu.cc
View File

@ -15,7 +15,7 @@ limitations under the License.
#if GOOGLE_CUDA
#define EIGEN_USE_GPU
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/core/kernels/fused_batch_norm_op.h"
#include "tensorflow/core/util/gpu_kernel_helper.h"

2
tensorflow/core/kernels/lrn_op.cc
View File

@ -35,7 +35,7 @@ limitations under the License.
#endif
#if GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/core/platform/stream_executor.h"
#include "tensorflow/core/util/stream_executor_util.h"
#endif // GOOGLE_CUDA

2
tensorflow/core/kernels/matmul_op.cc
View File

@ -25,7 +25,7 @@ limitations under the License.
#include "tensorflow/core/kernels/fill_functor.h"
#include "tensorflow/core/util/matmul_autotune.h"
#if GOOGLE_CUDA
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/core/kernels/gpu_utils.h"
#include "tensorflow/core/platform/stream_executor.h"
#endif // GOOGLE_CUDA

2
tensorflow/core/kernels/maxpooling_op.cc
View File

@ -41,7 +41,7 @@ limitations under the License.
#include "tensorflow/core/util/use_cudnn.h"
#if GOOGLE_CUDA
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/core/kernels/maxpooling_op_gpu.h"
#include "tensorflow/core/kernels/pooling_ops_common_gpu.h"
#include "tensorflow/core/platform/stream_executor.h"

2
tensorflow/core/kernels/pooling_ops_common.cc
View File

@ -21,7 +21,7 @@ limitations under the License.
#include "tensorflow/core/framework/tensor.h"
#if GOOGLE_CUDA
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/core/kernels/conv_2d.h"
#include "tensorflow/core/kernels/pooling_ops_common_gpu.h"
#include "tensorflow/core/platform/stream_executor.h"

2
tensorflow/core/kernels/reduction_gpu_kernels.cu.h
View File

@ -28,7 +28,7 @@ limitations under the License.
#include "third_party/cub/iterator/counting_input_iterator.cuh"
#include "third_party/cub/iterator/transform_input_iterator.cuh"
#include "third_party/cub/warp/warp_reduce.cuh"
#include "cuda/include/cuComplex.h"
#include "third_party/gpus/cuda/include/cuComplex.h"
#include "tensorflow/core/kernels/reduction_ops.h"
#include "tensorflow/core/lib/core/bits.h"
#include "tensorflow/core/util/gpu_device_functions.h"

2
tensorflow/core/kernels/scan_ops_gpu.h
View File

@ -29,7 +29,7 @@ limitations under the License.
#include "third_party/cub/block/block_store.cuh"
#include "third_party/cub/iterator/counting_input_iterator.cuh"
#include "third_party/cub/iterator/transform_input_iterator.cuh"
#include "cuda/include/cuComplex.h"
#include "third_party/gpus/cuda/include/cuComplex.h"
#include "tensorflow/core/framework/numeric_types.h"
#include "tensorflow/core/framework/register_types.h"
#include "tensorflow/core/kernels/scan_ops.h"

2
tensorflow/core/platform/default/cuda_libdevice_path.cc
View File

@ -19,7 +19,7 @@ limitations under the License.
#include <vector>
#if !defined(PLATFORM_GOOGLE)
#include "cuda/cuda_config.h"
#include "third_party/gpus/cuda/cuda_config.h"
#endif
#include "tensorflow/core/platform/logging.h"

2
tensorflow/core/platform/default/device_tracer.cc
View File

@ -25,7 +25,7 @@ limitations under the License.
#include "absl/strings/ascii.h"
#include "absl/strings/str_cat.h"
#include "absl/strings/str_format.h"
#include "cuda/extras/CUPTI/include/cupti.h"
#include "third_party/gpus/cuda/extras/CUPTI/include/cupti.h"
#include "tensorflow/core/common_runtime/step_stats_collector.h"
#include "tensorflow/core/framework/step_stats.pb.h"
#include "tensorflow/core/lib/core/errors.h"

4
tensorflow/core/util/gpu_device_functions.h
View File

@ -31,8 +31,8 @@ limitations under the License.
#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
#if GOOGLE_CUDA
#include "cuda/include/cuComplex.h"
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuComplex.h"
#include "third_party/gpus/cuda/include/cuda.h"
#endif
#include "tensorflow/core/platform/types.h"

2
tensorflow/core/util/gpu_kernel_helper.h
View File

@ -19,7 +19,7 @@ limitations under the License.
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
#if GOOGLE_CUDA
#include "cuda/include/cuda_fp16.h"
#include "third_party/gpus/cuda/include/cuda_fp16.h"
#endif
#include "tensorflow/core/util/gpu_device_functions.h"
#include "tensorflow/core/util/gpu_launch_config.h"

4
tensorflow/stream_executor/cuda/cublas_stub.cc
View File

@ -12,8 +12,8 @@ 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.
==============================================================================*/
#include "cuda/include/cublas.h"
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cublas.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

6
tensorflow/stream_executor/cuda/cuda_blas.cc
View File

@ -13,8 +13,8 @@ See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include "cuda/include/cublas_v2.h"
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cublas_v2.h"
#include "third_party/gpus/cuda/include/cuda.h"
#define SE_CUDA_DATA_HALF CUDA_R_16F
@ -40,7 +40,7 @@ limitations under the License.
// TODO(b/73793421): Remove the following code block to switch to the second
// approach when the issue is fixed.
#if CUDA_VERSION < 9000
#include "cuda/include/cuda_fp16.h"
#include "third_party/gpus/cuda/include/cuda_fp16.h"
#define EIGEN_HAS_CUDA_FP16
#endif

2
tensorflow/stream_executor/cuda/cuda_dnn.cc
View File

@ -43,7 +43,7 @@ limitations under the License.
#include "tensorflow/stream_executor/stream.h"
#include "tensorflow/stream_executor/stream_executor_pimpl.h"
// clang-format off
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "absl/strings/string_view.h"
// clang-format on

2
tensorflow/stream_executor/cuda/cuda_driver.cc
View File

@ -31,7 +31,7 @@ limitations under the License.
#include "absl/strings/str_format.h"
#include "absl/synchronization/mutex.h"
#include "absl/synchronization/notification.h"
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorflow/stream_executor/cuda/cuda_diagnostics.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/lib/error.h"

2
tensorflow/stream_executor/cuda/cuda_fft.h
View File

@ -20,7 +20,7 @@ limitations under the License.
#ifndef TENSORFLOW_STREAM_EXECUTOR_CUDA_CUDA_FFT_H_
#define TENSORFLOW_STREAM_EXECUTOR_CUDA_CUDA_FFT_H_
#include "cuda/include/cufft.h"
#include "third_party/gpus/cuda/include/cufft.h"
#include "tensorflow/stream_executor/fft.h"
#include "tensorflow/stream_executor/platform/port.h"
#include "tensorflow/stream_executor/plugin_registry.h"

2
tensorflow/stream_executor/cuda/cuda_rng.cc
View File

@ -27,7 +27,7 @@ limitations under the License.
#include "tensorflow/stream_executor/platform/logging.h"
#include "tensorflow/stream_executor/rng.h"
// clang-format off
#include "cuda/include/curand.h"
#include "third_party/gpus/cuda/include/curand.h"
// clang-format on
// Formats curandStatus_t to output prettified values into a log stream.

2
tensorflow/stream_executor/cuda/cuda_stub.cc
View File

@ -12,7 +12,7 @@ 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.
==============================================================================*/
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/cudart_stub.cc
View File

@ -16,7 +16,7 @@ limitations under the License.
// This file wraps cuda runtime calls with dso loader so that we don't need to
// have explicit linking to libcuda.
#include "cuda/include/cuda_runtime_api.h"
#include "third_party/gpus/cuda/include/cuda_runtime_api.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/cudnn_stub.cc
View File

@ -12,7 +12,7 @@ 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.
==============================================================================*/
#include "cuda/include/cudnn.h"
#include "third_party/gpus/cudnn/cudnn.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/cufft_stub.cc
View File

@ -12,7 +12,7 @@ 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.
==============================================================================*/
#include "cuda/include/cufft.h"
#include "third_party/gpus/cuda/include/cufft.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/cupti_stub.cc
View File

@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include "cuda/extras/CUPTI/include/cupti.h"
#include "third_party/gpus/cuda/extras/CUPTI/include/cupti.h"
// IWYU pragma: no_include "perftools/gputools/executor/stream_executor.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/curand_stub.cc
View File

@ -12,7 +12,7 @@ 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.
==============================================================================*/
#include "cuda/include/curand.h"
#include "third_party/gpus/cuda/include/curand.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/cusolver_stub.cc
View File

@ -12,7 +12,7 @@ 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.
==============================================================================*/
#include "cuda/include/cusolverDn.h"
#include "third_party/gpus/cuda/include/cusolverDn.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/cuda/cusparse_stub.cc
View File

@ -12,7 +12,7 @@ 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.
==============================================================================*/
#include "cuda/include/cusparse.h"
#include "third_party/gpus/cuda/include/cusparse.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/platform/dso_loader.h"

2
tensorflow/stream_executor/gpu/gpu_driver.h
View File

@ -21,7 +21,7 @@ limitations under the License.
#include <stddef.h>
#include "tensorflow/stream_executor/platform/port.h"
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuda.h"
#include "tensorflow/stream_executor/device_options.h"
#include "tensorflow/stream_executor/lib/status.h"
#include "tensorflow/stream_executor/lib/statusor.h"

4
tensorflow/stream_executor/gpu/gpu_types.h
View File

@ -28,8 +28,8 @@ limitations under the License.
#else // CUDA
#include "cuda/include/cuComplex.h"
#include "cuda/include/cuda.h"
#include "third_party/gpus/cuda/include/cuComplex.h"
#include "third_party/gpus/cuda/include/cuda.h"
// cannot include curand.h here
// because it triggers the #error in cuda/cuda_gpu_executor.cc

2
tensorflow/stream_executor/platform/default/dso_loader.cc
View File

@ -18,7 +18,7 @@ limitations under the License.
#include "absl/strings/str_cat.h"
#include "absl/strings/string_view.h"
#include "cuda/cuda_config.h"
#include "third_party/gpus/cuda/cuda_config.h"
#include "tensorflow/stream_executor/lib/env.h"
#include "tensorflow/stream_executor/lib/error.h"
#include "tensorflow/stream_executor/lib/path.h"

122
third_party/gpus/cuda/BUILD.tpl vendored
View File

@ -28,36 +28,37 @@ config_setting(
config_setting(
name = "darwin",
values = {"cpu": "darwin"},
visibility = ["//visibility:public"],
)
config_setting(
name = "freebsd",
values = {"cpu": "freebsd"},
visibility = ["//visibility:public"],
)
# Provides CUDA headers for '#include "third_party/gpus/cuda/include/cuda.h"'
# All clients including TensorFlow should use these directives.
cc_library(
name = "cuda_headers",
name = "cuda_virtual_headers",
hdrs = [
"cuda/cuda_config.h",
%{cuda_headers}
":cuda-include"
],
includes = [
".",
"cuda/include",
"cuda/include/crt",
],
visibility = ["//visibility:public"],
include_prefix = "third_party/gpus",
visibility = ["//visibility:private"],
)
# Provides CUDA headers for '#include <cuda.h>'.
# CUDA itself as well as Eigen use these directives.
cc_library(
name = "cuda_headers",
textual_hdrs = [":cuda-include"],
includes = ["cuda/include"],
deps = [":cuda_virtual_headers"],
)
cc_library(
name = "cudart_static",
srcs = ["cuda/lib/%{cudart_static_lib}"],
includes = [
".",
"cuda/include",
],
linkopts = select({
":freebsd": [],
"//conditions:default": ["-ldl"],
@ -65,104 +66,82 @@ cc_library(
"-lpthread",
%{cudart_static_linkopt}
],
visibility = ["//visibility:public"],
)
cc_library(
name = "cuda_driver",
srcs = ["cuda/lib/%{cuda_driver_lib}"],
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
)
cc_library(
name = "cudart",
srcs = ["cuda/lib/%{cudart_lib}"],
data = ["cuda/lib/%{cudart_lib}"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cublas_virtual_headers",
hdrs = [":cublas-include"],
include_prefix = "third_party/gpus/cuda/include",
strip_include_prefix = "cublas/include",
visibility = ["//visibility:private"],
deps = [":cuda_headers"],
)
cc_library(
name = "cublas_headers",
textual_hdrs = [":cublas-include"],
includes = ["cublas/include"],
deps = [":cublas_virtual_headers"],
)
cc_library(
name = "cublas",
srcs = ["cuda/lib/%{cublas_lib}"],
data = ["cuda/lib/%{cublas_lib}"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cusolver",
srcs = ["cuda/lib/%{cusolver_lib}"],
data = ["cuda/lib/%{cusolver_lib}"],
includes = [
".",
"cuda/include",
],
linkopts = ["-lgomp"],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cudnn",
srcs = ["cuda/lib/%{cudnn_lib}"],
data = ["cuda/lib/%{cudnn_lib}"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cudnn_header",
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
hdrs = [":cudnn-include"],
include_prefix = "third_party/gpus/cudnn",
strip_include_prefix = "cudnn/include",
deps = [":cuda_headers"],
)
cc_library(
name = "cufft",
srcs = ["cuda/lib/%{cufft_lib}"],
data = ["cuda/lib/%{cufft_lib}"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "curand",
srcs = ["cuda/lib/%{curand_lib}"],
data = ["cuda/lib/%{curand_lib}"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cuda",
visibility = ["//visibility:public"],
deps = [
":cublas",
":cuda_headers",
@ -173,46 +152,37 @@ cc_library(
],
)
cc_library(
name = "cupti_virtual_headers",
hdrs = [":cuda-extras"],
include_prefix="third_party/gpus",
visibility = ["//visibility:private"],
deps = [":cuda_headers"],
)
cc_library(
name = "cupti_headers",
hdrs = [
"cuda/cuda_config.h",
":cuda-extras",
],
includes = [
".",
"cuda/extras/CUPTI/include/",
],
visibility = ["//visibility:public"],
textual_hdrs = [":cuda-extras"],
includes = ["cuda/extras/CUPTI/include/"],
deps = [":cupti_virtual_headers"],
)
cc_library(
name = "cupti_dsos",
data = ["cuda/lib/%{cupti_lib}"],
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
)
cc_library(
name = "cusparse",
srcs = ["cuda/lib/%{cusparse_lib}"],
data = ["cuda/lib/%{cusparse_lib}"],
includes = [
".",
"cuda/include",
],
linkopts = ["-lgomp"],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "libdevice_root",
data = [":cuda-nvvm"],
visibility = ["//visibility:public"],
)
%{copy_rules}

60
third_party/gpus/cuda_configure.bzl vendored
View File

@ -784,8 +784,11 @@ def _create_dummy_repository(repository_ctx):
"%{curand_lib}": lib_name("curand", cpu_value),
"%{cupti_lib}": lib_name("cupti", cpu_value),
"%{cusparse_lib}": lib_name("cusparse", cpu_value),
"%{copy_rules}": "",
"%{cuda_headers}": "",
"%{copy_rules}": """
filegroup(name="cuda-include")
filegroup(name="cublas-include")
filegroup(name="cudnn-include")
""",
},
)
@ -983,25 +986,21 @@ def _create_local_cuda_repository(repository_ctx):
out_dir = "cuda/extras/CUPTI/include",
),
]
included_files = _read_dir(repository_ctx, cuda_include_path)
if not any([file.endswith("cublas.h") for file in included_files]):
copy_rules.append(make_copy_files_rule(
repository_ctx,
name = "cublas-include",
srcs = [
cublas_include_path + "/cublas.h",
cublas_include_path + "/cublas_v2.h",
cublas_include_path + "/cublas_api.h",
],
outs = [
"cuda/include/cublas.h",
"cuda/include/cublas_v2.h",
"cuda/include/cublas_api.h",
],
))
else:
copy_rules.append("filegroup(name = 'cublas-include')\n")
copy_rules.append(make_copy_files_rule(
repository_ctx,
name = "cublas-include",
srcs = [
cublas_include_path + "/cublas.h",
cublas_include_path + "/cublas_v2.h",
cublas_include_path + "/cublas_api.h",
],
outs = [
"cublas/include/cublas.h",
"cublas/include/cublas_v2.h",
"cublas/include/cublas_api.h",
],
))
cuda_libs = _find_libs(repository_ctx, cuda_config)
cuda_lib_srcs = []
@ -1023,16 +1022,12 @@ def _create_local_cuda_repository(repository_ctx):
out_dir = "cuda/bin",
))
# Copy cudnn.h if cuDNN was not installed to CUDA_TOOLKIT_PATH.
if not any([file.endswith("cudnn.h") for file in included_files]):
copy_rules.append(make_copy_files_rule(
repository_ctx,
name = "cudnn-include",
srcs = [cudnn_header_dir + "/cudnn.h"],
outs = ["cuda/include/cudnn.h"],
))
else:
copy_rules.append("filegroup(name = 'cudnn-include')\n")
copy_rules.append(make_copy_files_rule(
repository_ctx,
name = "cudnn-include",
srcs = [cudnn_header_dir + "/cudnn.h"],
outs = ["cudnn/include/cudnn.h"],
))
# Set up BUILD file for cuda/
_tpl(
@ -1062,11 +1057,6 @@ def _create_local_cuda_repository(repository_ctx):
"%{cupti_lib}": cuda_libs["cupti"].basename,
"%{cusparse_lib}": cuda_libs["cusparse"].basename,
"%{copy_rules}": "\n".join(copy_rules),
"%{cuda_headers}": (
'":cuda-include",\n' +
' ":cublas-include",' +
' ":cudnn-include",'
),
},
"cuda/BUILD",
)

137
third_party/toolchains/preconfig/ubuntu14.04/cuda10.0-cudnn7/cuda/BUILD vendored
View File

@ -28,37 +28,37 @@ config_setting(
config_setting(
name = "darwin",
values = {"cpu": "darwin"},
visibility = ["//visibility:public"],
)
config_setting(
name = "freebsd",
values = {"cpu": "freebsd"},
visibility = ["//visibility:public"],
)
# Provides CUDA headers for '#include "third_party/gpus/cuda/include/cuda.h"'
# All clients including TensorFlow should use these directives.
cc_library(
name = "cuda_headers",
name = "cuda_virtual_headers",
hdrs = [
"cuda/cuda_config.h",
":cuda-include",
":cudnn-include",
],
includes = [
".",
"cuda/include",
"cuda/include/crt",
],
visibility = ["//visibility:public"],
include_prefix = "third_party/gpus",
visibility = ["//visibility:private"],
)
# Provides CUDA headers for '#include <cuda.h>'.
# CUDA itself as well as Eigen use these directives.
cc_library(
name = "cuda_headers",
includes = ["cuda/include"],
textual_hdrs = [":cuda-include"],
deps = [":cuda_virtual_headers"],
)
cc_library(
name = "cudart_static",
srcs = ["cuda/lib/libcudart_static.a"],
includes = [
".",
"cuda/include",
],
linkopts = select({
":freebsd": [],
"//conditions:default": ["-ldl"],
@ -66,104 +66,82 @@ cc_library(
"-lpthread",
"-lrt",
],
visibility = ["//visibility:public"],
)
cc_library(
name = "cuda_driver",
srcs = ["cuda/lib/libcuda.so"],
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
)
cc_library(
name = "cudart",
srcs = ["cuda/lib/libcudart.so.10.0"],
data = ["cuda/lib/libcudart.so.10.0"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cublas_virtual_headers",
hdrs = [":cublas-include"],
include_prefix = "third_party/gpus/cuda/include",
strip_include_prefix = "cublas/include",
visibility = ["//visibility:private"],
deps = [":cuda_headers"],
)
cc_library(
name = "cublas_headers",
includes = ["cublas/include"],
textual_hdrs = [":cublas-include"],
deps = [":cublas_virtual_headers"],
)
cc_library(
name = "cublas",
srcs = ["cuda/lib/libcublas.so.10.0"],
data = ["cuda/lib/libcublas.so.10.0"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cusolver",
srcs = ["cuda/lib/libcusolver.so.10.0"],
data = ["cuda/lib/libcusolver.so.10.0"],
includes = [
".",
"cuda/include",
],
linkopts = ["-lgomp"],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cudnn",
srcs = ["cuda/lib/libcudnn.so.7"],
data = ["cuda/lib/libcudnn.so.7"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cudnn_header",
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
hdrs = [":cudnn-include"],
include_prefix = "third_party/gpus/cudnn",
strip_include_prefix = "cudnn/include",
deps = [":cuda_headers"],
)
cc_library(
name = "cufft",
srcs = ["cuda/lib/libcufft.so.10.0"],
data = ["cuda/lib/libcufft.so.10.0"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "curand",
srcs = ["cuda/lib/libcurand.so.10.0"],
data = ["cuda/lib/libcurand.so.10.0"],
includes = [
".",
"cuda/include",
],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "cuda",
visibility = ["//visibility:public"],
deps = [
":cublas",
":cuda_headers",
@ -174,46 +152,37 @@ cc_library(
],
)
cc_library(
name = "cupti_virtual_headers",
hdrs = [":cuda-extras"],
include_prefix = "third_party/gpus",
visibility = ["//visibility:private"],
deps = [":cuda_headers"],
)
cc_library(
name = "cupti_headers",
hdrs = [
"cuda/cuda_config.h",
":cuda-extras",
],
includes = [
".",
"cuda/extras/CUPTI/include/",
],
visibility = ["//visibility:public"],
includes = ["cuda/extras/CUPTI/include/"],
textual_hdrs = [":cuda-extras"],
deps = [":cupti_virtual_headers"],
)
cc_library(
name = "cupti_dsos",
data = ["cuda/lib/libcupti.so.10.0"],
includes = [
".",
"cuda/include",
],
visibility = ["//visibility:public"],
)
cc_library(
name = "cusparse",
srcs = ["cuda/lib/libcusparse.so.10.0"],
data = ["cuda/lib/libcusparse.so.10.0"],
includes = [
".",
"cuda/include",
],
linkopts = ["-lgomp"],
linkstatic = 1,
visibility = ["//visibility:public"],
)
cc_library(
name = "libdevice_root",
data = [":cuda-nvvm"],
visibility = ["//visibility:public"],
)
genrule(
@ -1253,6 +1222,16 @@ genrule(
cmd = """cp -rLf "/usr/local/cuda-10.0/extras/CUPTI/include/." "$(@D)/cuda/extras/CUPTI/include/" """,
)
genrule(
name = "cublas-include",
outs = [
"cublas/include/cublas.h",
"cublas/include/cublas_v2.h",
"cublas/include/cublas_api.h",
],
cmd = """cp -f "/usr/local/cuda-10.0/include/cublas.h" $(location cublas/include/cublas.h) && cp -f "/usr/local/cuda-10.0/include/cublas_v2.h" $(location cublas/include/cublas_v2.h) && cp -f "/usr/local/cuda-10.0/include/cublas_api.h" $(location cublas/include/cublas_api.h) """,
)
genrule(
name = "cuda-lib",
outs = [
@ -1267,7 +1246,7 @@ genrule(
"cuda/lib/libcupti.so.10.0",
"cuda/lib/libcusparse.so.10.0",
],
cmd = """cp -f "/usr/local/cuda-10.0/lib64/stubs/libcuda.so" $(location cuda/lib/libcuda.so) && cp -f "/usr/local/cuda-10.0/lib64/libcudart.so.10.0" $(location cuda/lib/libcudart.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcudart_static.a" $(location cuda/lib/libcudart_static.a) && cp -f "/usr/local/cuda-10.0/lib64/libcublas.so.10.0" $(location cuda/lib/libcublas.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcusolver.so.10.0" $(location cuda/lib/libcusolver.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcusparse.so.10.0" $(location cuda/lib/libcusparse.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcurand.so.10.0" $(location cuda/lib/libcurand.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcufft.so.10.0" $(location cuda/lib/libcufft.so.10.0) && cp -f "/usr/lib/x86_64-linux-gnu/libcudnn.so.7" $(location cuda/lib/libcudnn.so.7) && cp -f "/usr/local/cuda-10.0/extras/CUPTI/lib64/libcupti.so.10.0" $(location cuda/lib/libcupti.so.10.0) """,
cmd = """cp -f "/usr/local/cuda-10.0/lib64/stubs/libcuda.so" $(location cuda/lib/libcuda.so) && cp -f "/usr/local/cuda-10.0/lib64/libcudart.so.10.0" $(location cuda/lib/libcudart.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcudart_static.a" $(location cuda/lib/libcudart_static.a) && cp -f "/usr/local/cuda-10.0/lib64/libcublas.so.10.0" $(location cuda/lib/libcublas.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcusolver.so.10.0" $(location cuda/lib/libcusolver.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcurand.so.10.0" $(location cuda/lib/libcurand.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcufft.so.10.0" $(location cuda/lib/libcufft.so.10.0) && cp -f "/usr/lib/x86_64-linux-gnu/libcudnn.so.7" $(location cuda/lib/libcudnn.so.7) && cp -f "/usr/local/cuda-10.0/extras/CUPTI/lib64/libcupti.so.10.0" $(location cuda/lib/libcupti.so.10.0) && cp -f "/usr/local/cuda-10.0/lib64/libcusparse.so.10.0" $(location cuda/lib/libcusparse.so.10.0) """,
)
genrule(
@ -1297,7 +1276,7 @@ genrule(
genrule(
name = "cudnn-include",
outs = [
"cuda/include/cudnn.h",
"cudnn/include/cudnn.h",
],
cmd = """cp -f "/usr/include/cudnn.h" $(location cuda/include/cudnn.h) """,
cmd = """cp -f "/usr/include/cudnn.h" $(location cudnn/include/cudnn.h) """,
)

14
third_party/toolchains/preconfig/ubuntu14.04/tensorrt5/BUILD vendored
View File

@ -3,12 +3,12 @@
licenses(["notice"])
exports_files(["LICENSE"])
load("@local_config_cuda//cuda:build_defs.bzl", "cuda_default_copts")
package(default_visibility = ["//visibility:public"])
exports_files(["LICENSE"])
cc_library(
name = "tensorrt_headers",
hdrs = [":tensorrt_include"],
@ -18,15 +18,9 @@ cc_library(
cc_library(
name = "tensorrt",
srcs = [
"tensorrt/lib/libnvinfer.so.5",
"tensorrt/lib/libnvinfer_plugin.so.5",
],
srcs = [":tensorrt_lib"],
copts = cuda_default_copts(),
data = [
"tensorrt/lib/libnvinfer.so.5",
"tensorrt/lib/libnvinfer_plugin.so.5",
],
data = [":tensorrt_lib"],
include_prefix = "",
linkstatic = 1,
visibility = ["//visibility:public"],