Skip to content
GitLab

Added support for multi-operator WLs in the CUDA backend

Merged Sven-Bodo Scholz requested to merge sbs/sac2c:cuda-support-mowls into develop

This mainly required 3 upgrades:

  1. allow mo-WLs to be marked CUDARIZABLE
  2. make sure all results from cudarised WLs are transferred out (device2host)
  3. adjust the CUDA kernel generation to properly deal with shared wlidxes if these are shared