Skip to content

Implement generic j-reduction in nbnxm SYCL kernels

Szilárd Páll requested to merge sz_SYCL-nbnxm-local-mem-reduction into master

This commit implements the generic j-reduction identical to the OpenCL version of the same.

Also added a subGroupBarrier() helper which is needed for correctness on the CUDA backend when targetting NVIDIA architectures.

Refs #3934

Edited by Szilárd Páll

Merge request reports