Adding synchronisation to convolution kernel for sycl backend.

This commit is contained in:
Mehdi Goli 2017-03-13 09:18:37 +00:00
parent 1b32a10053
commit f499fe9496
2 changed files with 6 additions and 0 deletions

View File

@ -539,6 +539,11 @@ for (int iter = 0; iter < 10; ++iter) {
if (Eigen::internal::is_same<Device, Eigen::GpuDevice>::value) { if (Eigen::internal::is_same<Device, Eigen::GpuDevice>::value) {
device_.synchronize(); device_.synchronize();
} }
#elif defined(EIGEN_USE_SYCL)
if (Eigen::internal::is_same<Device, Eigen::SyclDevice>::value) {
device_.synchronize();
}
#endif #endif
StopBenchmarkTiming(); StopBenchmarkTiming();
SetBenchmarkFlopsProcessed(num_items); SetBenchmarkFlopsProcessed(num_items);

View File

@ -425,6 +425,7 @@ struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelAr
} }
} }
}); });
m_device.asynchronousExec();
} }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE CoeffReturnType coeff(Index index) const EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE CoeffReturnType coeff(Index index) const