Using `fetch_min` / `fetch_max` in `__parallel_find_or` and in it's stuff

oneapi-src / oneDPL

oneAPI DPC++ Library (oneDPL) https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/dpc-library.html

Apache License 2.0

715 stars 112 forks source link

Closed SergeyKopienko closed 1 week ago

SergeyKopienko commented 2 weeks ago

In this PR we using fetch_min / fetch_max in __parallel_find_or and it's stuff. This approach give us additional performance boost:

now we shouldn't try to setup atomic states inside for-based loops with additional checks.

This idea was proposed by @danhoeflinger.

Also in this PR we using __dpl_ceiling_div function where it's applicable in the code.