diff options
| author |  A. Unique TensorFlower <nobody@tensorflow.org> | 2016-05-17 13:42:26 -0800 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2016-05-17 14:52:08 -0700 |
| commit | 4d55cb5c55fcc85009171b6a4657cbd966fd85e5 (patch) | |
| tree | 303d69d5be3054475262d49ceefc9c116ec5da40 /tensorflow/core/util/port.h | |
| parent | 2709b2303b39b848b9c6a86cd89811820d390580 (diff) | |
Enable fp16 support for MatMul via cuBLAS, gated on compilation with
CUDA 7.5 or higher. (If not, the GPU tests will also not be run,
so that the tests as a whole will keep passing on CUDA 7.0.)
Change: 122566230
Diffstat (limited to 'tensorflow/core/util/port.h')
| -rw-r--r-- | tensorflow/core/util/port.h | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/tensorflow/core/util/port.h b/tensorflow/core/util/port.h index 009d8968da..65d628c841 100644 --- a/tensorflow/core/util/port.h +++ b/tensorflow/core/util/port.h @@ -21,6 +21,10 @@ namespace tensorflow { // Returns true if GOOGLE_CUDA is defined. bool IsGoogleCudaEnabled(); +// Returns true if GOOGLE_CUDA is defined, and the given CUDA version supports +// half-precision matrix multiplications and convolution operations. +bool CudaSupportsHalfMatMulAndConv(); + } // end namespace tensorflow #endif // TENSORFLOW_UTIL_PORT_H_ |