composable_kernel icon indicating copy to clipboard operation
composable_kernel copied to clipboard
verified publisher icon ROCmSoftwarePlatform

[Ck tile] Support layernorm one pass

Open rocking5566 opened this issue 1 year ago • 2 comments

  1. Add one pass pipeline, switch one pass and two pass pipeline according to problem size
  2. Fix compile error
  3. Support padding

rocking5566 avatar Sep 13 '24 06:09 rocking5566

Since OnePassLayernorm2dFwd() and TwoPassLayernorm2dFwd() share lots of codes. I think maybe we can merge them into one.

poyenc avatar Sep 19 '24 04:09 poyenc

Since OnePassLayernorm2dFwd() and TwoPassLayernorm2dFwd() share lots of codes. I think maybe we can merge them into one.

1c644ec523223823a6980172399d549d420df8d0

rocking5566 avatar Sep 24 '24 12:09 rocking5566