Skip to content

on-demand data context allocation - fix for CUDA initialization error in...

Dmitry Emeliyanov requested to merge demelian/athena:master-gpu-hlt-demo-on-demand-cuda-contexts into master

on-demand data context allocation - fix for CUDA initialization error in athenaHLT with multiple processes and threads, athenaMP-compatible initialization uses a separate process to determine the number of GPUs. No change in test results is expected.

Merge request reports