diff options
| -rw-r--r-- | unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h b/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h index 3b87b114d..e03735611 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h @@ -87,13 +87,13 @@ struct ThreadPoolDevice { const size_t kMinBlockSize = 32768; typedef TensorCostModel<ThreadPoolDevice> CostModel; const size_t num_threads = CostModel::numThreads(n, TensorOpCost(1.0, 1.0, 0), 4); - if (n <= kMinBlockSize || num_threads == 1) { + if (n <= kMinBlockSize || num_threads < 2) { ::memcpy(dst, src, n); } else { const char* src_ptr = static_cast<const char*>(src); char* dst_ptr = static_cast<char*>(dst); const size_t blocksize = (n + (num_threads - 1)) / num_threads; - Barrier barrier(num_threads - 1); + Barrier barrier(static_cast<int>(num_threads - 1)); // Launch the last 3 blocks on worker threads. for (size_t i = 1; i < num_threads; ++i) { enqueue_with_barrier(&barrier, [n, i, src_ptr, dst_ptr, blocksize] { |