Skip to content

Conversation

@samnordmann
Copy link
Contributor

Add support for new datatypes in ucc. (openucx/ucc@8afd34a)

New datatypes: float128, float32_complex, float64_complex, float128_complex.

  • float32_complex and float64_complex are supported on HOST and CUDA. float128 and float128_complex are supported only on HOST.
  • The new datatypes are supported by all collectives. float128 supported with the reduction ops: min, max, sum, prod, avg. float32(64,128)_complex supported with the reduction ops: sum, prod, avg.
  • new datatypes not supported by nccl nor sharp TLs

@ompiteam-bot
Copy link

Can one of the admins verify this patch?

@jsquyres
Copy link
Member

ok to test

Copy link
Member

@jsquyres jsquyres left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@samnordmann Please add a signed-off-by line in your commit message to indicate that you agree to Open MPI's Contributor's Declaration: https://github.com/open-mpi/ompi/blob/main/.github/CONTRIBUTING.md#contributors-declaration

@samnordmann samnordmann requested a review from jsquyres June 22, 2022 16:01
@jsquyres jsquyres dismissed their stale review June 22, 2022 18:29

Commit is now signed off properly; thanks.

@janjust janjust merged commit a43c46f into open-mpi:main Jun 27, 2022
@janjust
Copy link
Contributor

janjust commented Jun 27, 2022

@samnordmann please open up the v5.0.x PR

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants