[GitHub] [tvm] AndrewZhaoLuo opened a new pull request #9252: [WIP] [Codegen] [CUDA] enable fast math flags

[GitBox <gi...@apache.org>]

AndrewZhaoLuo opened a new pull request #9252:
URL: https://github.com/apache/tvm/pull/9252


   WIP until I profile and think of a better way to pass these strings in. E.g. through target strings or something seperate. 
   
   Link #9223 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] AndrewZhaoLuo commented on pull request #9252: [WIP] [Codegen] [CUDA] enable fast math flags

[GitBox <gi...@apache.org>]

AndrewZhaoLuo commented on pull request #9252:
URL: https://github.com/apache/tvm/pull/9252#issuecomment-942740031


   Performance is extremely bad, with pretty much every model I have having final inference regressions in the range of 60%. I'm going to close this for now, but might investigate this later.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] AndrewZhaoLuo closed pull request #9252: [WIP] [Codegen] [CUDA] enable fast math flags

[GitBox <gi...@apache.org>]

AndrewZhaoLuo closed pull request #9252:
URL: https://github.com/apache/tvm/pull/9252


   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org