dpctl icon indicating copy to clipboard operation
dpctl copied to clipboard

Published 20 hours ago • verified publisher icon IntelPython

sycl::vec overloads for elementwise functions

Open ndgrigorian opened this issue 1 year ago • 10 comments

This PR adds overloads for sycl::vec input to the implementations of dpctl.tensor.abs, dpctl.tensor.cos, dpctl.tensor.expm1, dpctl.tensor.log, dpctl.tensor.log1p, and dpctl.tensor.sqrt.

  • [X] Have you provided a meaningful PR description?
  • [X] Have you added a test, reproducer or referred to an issue with a reproducer?
  • [X] Have you tested your changes locally for CPU and GPU devices?
  • [X] Have you made sure that new changes do not introduce compiler warnings?
  • [ ] Have you checked performance impact of proposed changes?
  • [X] If this PR is a work in progress, are you opening the PR as a draft?

ndgrigorian avatar Jun 01 '23 18:06 ndgrigorian

View rendered docs @ https://intelpython.github.io/dpctl/pulls/1223/index.html

github-actions[bot] avatar Jun 01 '23 18:06 github-actions[bot]

Coverage Status

coverage: 85.942%. remained the same when pulling 67bde5971a172535b0c0ef122a28031addd53a54 on elementwise-func-fixes into 5ec9fd5becdc45849c269dd553f255e5841cba49 on master.

coveralls avatar Jun 01 '23 19:06 coveralls

Array API standard conformance tests for dpctl=0.14.3dev3=py310h7bf5fec_9 ran successfully. Passed: 259 Failed: 741 Skipped: 116

github-actions[bot] avatar Jun 01 '23 19:06 github-actions[bot]

@ndgrigorian Please check that enabling vec brings on performance benefits on Max GPU.

oleksandr-pavlyk avatar Jun 02 '23 09:06 oleksandr-pavlyk

Array API standard conformance tests for dpctl=0.14.3dev3=py310h7bf5fec_20 ran successfully. Passed: 320 Failed: 680 Skipped: 119

github-actions[bot] avatar Jun 08 '23 09:06 github-actions[bot]

Array API standard conformance tests for dpctl=0.14.4=py310h7bf5fec_11 ran successfully. Passed: 388 Failed: 612 Skipped: 119

github-actions[bot] avatar Jun 22 '23 22:06 github-actions[bot]

Array API standard conformance tests for dpctl=0.14.5dev1=py310h7bf5fec_10 ran successfully. Passed: 448 Failed: 552 Skipped: 119

github-actions[bot] avatar Jul 12 '23 14:07 github-actions[bot]

Array API standard conformance tests for dpctl=0.14.6dev0=py310h7bf5fec_6 ran successfully. Passed: 474 Failed: 526 Skipped: 119

github-actions[bot] avatar Jul 18 '23 22:07 github-actions[bot]

Array API standard conformance tests for dpctl=0.15.1dev2=py310h15de555_20 ran successfully. Passed: 876 Failed: 55 Skipped: 59

github-actions[bot] avatar Nov 25 '23 09:11 github-actions[bot]

Testing has been performed and little-to-no significant performance gains were found for unary functions using sycl::vec overloads.

TODO: benchmark with sub-group loading disabled as well.

ndgrigorian avatar Jun 09 '24 04:06 ndgrigorian