| /external/tensorflow/tensorflow/stream_executor/cuda/ |
| cuda_platform_id.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_platform_id.h"
20 namespace cuda { namespace in namespace:perftools::gputools
24 } // namespace cuda
|
| cuda_platform_id.h | 23 namespace cuda { namespace in namespace:perftools::gputools
25 // Opaque and unique identifier for the cuda platform.
32 } // namespace cuda
|
| cuda_activation.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_activation.h"
18 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
24 namespace cuda { namespace in namespace:perftools::gputools
42 } // namespace cuda
|
| cuda_activation.h | 16 // This file contains APIs that assume a StreamExecutor is backed by CUDA.
17 // It reaches into the CUDA implementation to activate an underlying CUDA
21 // code does not also have to depend on cuda.h.
33 namespace cuda { namespace in namespace:perftools::gputools
38 // Activates a CUDA context within an enclosing scope.
41 // Form that takes a CUDA executor implementation.
44 // Form that takes a pImpl executor and extracts a CUDA implementation --
45 // fatal failure if it is not CUDA inside.
52 // The cuda.h-using datatype that we wrap
[all...] |
| cuda_event.h | 19 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
20 #include "tensorflow/stream_executor/cuda/cuda_stream.h"
26 namespace cuda { namespace in namespace:perftools::gputools
36 // Populates the CUDA-platform-specific elements of this object.
46 // Polls the CUDA platform for the event's current status.
49 // The underlying CUDA event element.
56 // The underlying CUDA event element.
60 } // namespace cuda
|
| cuda_rng.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_rng.h"
18 #include "tensorflow/stream_executor/cuda/cuda_activation.h"
19 #include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
20 #include "tensorflow/stream_executor/cuda/cuda_helpers.h"
21 #include "tensorflow/stream_executor/cuda/cuda_platform_id.h"
22 #include "tensorflow/stream_executor/cuda/cuda_stream.h"
29 #include "cuda/include/curand.h"
59 namespace cuda { namespace in namespace:perftools::gputools
69 cuda::ScopedActivateExecutorContext sac{parent}; \
273 } // namespace cuda
[all...] |
| cuda_timer.h | 16 // Defines the CUDATimer type - the CUDA-specific implementation of the generic
23 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
24 #include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
28 namespace cuda { namespace in namespace:perftools::gputools
80 } // namespace cuda
|
| cuda_diagnostics.h | 27 namespace cuda { namespace in namespace:perftools::gputools
43 // Logs diagnostic information when CUDA appears to be misconfigured (e.g. is
47 // produce very much log spew beyond saying, "looks like there's no CUDA
95 } // namespace cuda
|
| cuda_platform.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_platform.h"
18 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
19 #include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
20 #include "tensorflow/stream_executor/cuda/cuda_platform_id.h"
29 namespace cuda { namespace in namespace:perftools::gputools
68 : name_("CUDA"), min_numa_node_(0), limit_numa_node_(0) {}
138 if (!cuda::CUDADriver::Init().ok()) {
179 "failed initializing StreamExecutor for CUDA device ordinal %d: %s",
188 LOG(FATAL) << "not yet implemented: register CUDA trace listener";
192 LOG(FATAL) << "not yet implemented: unregister CUDA trace listener"
[all...] |
| cuda_platform.h | 36 namespace cuda { namespace in namespace:perftools::gputools
38 // Opaque and unique identifier for the CUDA platform plugin.
43 // Cuda-specific platform plugin, registered as a singleton value via module
106 } // namespace cuda
|
| cuda_stream.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_stream.h"
18 #include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
24 namespace cuda { namespace in namespace:perftools::gputools
61 } // namespace cuda
|
| cuda_stream.h | 16 // Defines the CUDAStream type - the CUDA-specific implementation of the generic
22 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
28 namespace cuda { namespace in namespace:perftools::gputools
49 // Explicitly initialize the CUDA resources associated with this stream, used
53 // Explicitly destroy the CUDA resources associated with this stream, used by
65 // Returns the CUstream value for passing to the CUDA API.
78 CUstream cuda_stream_; // Wrapped CUDA stream handle.
91 } // namespace cuda
|
| cuda_event.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_event.h"
18 #include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
19 #include "tensorflow/stream_executor/cuda/cuda_stream.h"
24 namespace cuda { namespace in namespace:perftools::gputools
70 } // namespace cuda
|
| cuda_helpers.h | 16 // Common helper functions used for dealing with CUDA API datatypes.
27 #include "cuda/include/cuComplex.h"
28 #include "cuda/include/cuda.h"
36 namespace cuda { namespace in namespace:perftools::gputools
39 // CUDA
47 // pointer in CUDA device memory.
63 // Type traits to get CUDA complex types from std::complex<>.
103 } // namespace cuda
|
| cuda_kernel.h | 16 // The CUDA implementation of the StreamExecutorInterface functionality.
17 // CUDA inclusions are ideally confined to this implementation file.
19 // The notions from the StreamExecutor basically correspond to the CUDA streams
27 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
31 #include "cuda/include/cuda.h"
40 "CUDA runtime being included into CUDA GPU executor; should be driver only."
45 namespace cuda { namespace in namespace:perftools::gputools
57 // As arity cannot be reflected upon using the CUDA API, the arity i
[all...] |
| cuda_rng.h | 34 namespace cuda { namespace in namespace:perftools::gputools
41 // CUDA-platform implementation of the random number generation support
100 } // namespace cuda
|
| cuda_timer.cc | 16 #include "tensorflow/stream_executor/cuda/cuda_timer.h"
18 #include "tensorflow/stream_executor/cuda/cuda_driver.h"
19 #include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
20 #include "tensorflow/stream_executor/cuda/cuda_stream.h"
25 namespace cuda { namespace in namespace:perftools::gputools
66 // CUDA docs say a resolution of about 0.5us
86 } // namespace cuda
|
| /external/tensorflow/tensorflow/contrib/makefile/ |
| build_all_android.sh | 62 if [ ! -d ${JETPACK}/cuda ]; then
63 ln -s $(ls -d ${JETPACK}/cuda-*/|sort -r|head -n1) ${JETPACK}/cuda
65 if [ ! -d ${JETPACK}/cuda ]; then
66 ln -s $(ls -d ${JETPACK}/cuda-*/|sort -r|head -n1) ${JETPACK}/cuda
|
| /external/tensorflow/tensorflow/tools/ci_build/windows/gpu/bazel/ |
| run_cc_test_windows.sh | 59 # bazel test --config=win-cuda -k $slow_compiling_test --test_output=errors
60 bazel test -c opt --config=win-cuda -k $passing_tests --test_output=errors
|
| /external/tensorflow/tensorflow/tools/ci_build/linux/gpu/ |
| run_cc_core.sh | 37 bazel test --config=cuda --test_tag_filters=-no_oss,-oss_serial,-no_gpu,-benchmark-test -k \
|
| run_py3_core.sh | 37 bazel test --config=cuda --test_tag_filters=-no_oss,-oss_serial,-no_gpu,-benchmark-test -k \
|
| /external/tensorflow/tensorflow/tools/ci_build/xla/linux/gpu/ |
| run_py3.sh | 38 bazel test --config=cuda --test_tag_filters=-no_gpu,-benchmark-test -k \
|
| /external/tensorflow/tensorflow/core/common_runtime/gpu/ |
| gpu_cudamalloc_allocator.cc | 17 #include "cuda/include/cuda.h"
18 #include "tensorflow/stream_executor/cuda/cuda_activation.h"
41 gpu::cuda::ScopedActivateExecutorContext scoped_activation{stream_exec_};
|
| /external/tensorflow/tools/ |
| tf_env_collect.sh | 108 echo '== cuda libs ==================================================='
111 find /usr/local -type f -name 'libcudart*' 2>/dev/null | grep cuda | grep -v "\\.cache" >> ${OUTPUT_FILE}
112 find /usr/local -type f -name 'libudnn*' 2>/dev/null | grep cuda | grep -v "\\.cache" >> ${OUTPUT_FILE}
|
| /external/tensorflow/tensorflow/tools/build_info/ |
| gen_build_info.py | 29 "cuda").
39 if build_config == "cuda":
87 help="Either 'cuda' for GPU builds or 'cpu' for CPU builds.")
|
