diff options
| author |  TensorFlower Gardener <gardener@tensorflow.org> | 2018-10-01 21:18:17 -0700 |
|---|---|---|
| committer | TensorFlower Gardener <gardener@tensorflow.org> | 2018-10-01 21:18:17 -0700 |
| commit | 6161d8cc4d66b87ba198cb6a16d83ce317c77eff (patch) | |
| tree | 47b45fcd51a5bf0001afd30295fc6e236d563a4e /tensorflow/compiler | |
| parent | 350388fca9cb9509962ff393a9d21fb2879c9179 (diff) | |
| parent | 6a5090b086bc9d665eb9e65f05eb94cdb58baaa2 (diff) | |
Merge pull request #21958 from MattConley:CudaOccupancy
PiperOrigin-RevId: 215331087
Diffstat (limited to 'tensorflow/compiler')
| -rw-r--r-- | tensorflow/compiler/xla/service/gpu/partition_assignment.cc | 9 |
1 files changed, 2 insertions, 7 deletions
diff --git a/tensorflow/compiler/xla/service/gpu/partition_assignment.cc b/tensorflow/compiler/xla/service/gpu/partition_assignment.cc index cf9f102d31..375f68a159 100644 --- a/tensorflow/compiler/xla/service/gpu/partition_assignment.cc +++ b/tensorflow/compiler/xla/service/gpu/partition_assignment.cc @@ -62,13 +62,8 @@ LaunchDimensions CalculateLaunchDimensions( // // <num threads per block> * <max blocks per core> = <max threads per core> - auto threads_per_core = device_desc.threads_per_core_limit(); - auto blocks_per_core = device_desc.blocks_per_core_limit(); - int64 threads_per_block; - if (threads_per_core != 0 && blocks_per_core != 0) { - threads_per_block = device_desc.threads_per_core_limit() / - device_desc.blocks_per_core_limit(); - } else { + int64 threads_per_block = device_desc.threads_per_block_limit(); + if (threads_per_block == 0) { static std::atomic<int64> log_count{0}; if (log_count.fetch_add(1) < 8) { LOG(WARNING) << "Attempting to calculate launch dimensions for GPU " |