Skip to content

Conversation

@amitchawla1
Copy link

@amitchawla1 amitchawla1 commented Oct 30, 2025

Todo: Handle void ElementD
Check if we need to handle assumption that accumulator is same as compute.
Test: Compiles for all generated kernels with void as Element type

@amitchawla1 amitchawla1 force-pushed the dev/achawla/fix_void_c_in_epilogue branch from 3d09e67 to 21d2651 Compare October 30, 2025 08:12
Copy link

@jiyang1011 jiyang1011 left a comment

Choose a reason for hiding this comment

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

LGTM

amitchawla1 and others added 8 commits November 3, 2025 02:27
Todo: Handle void ElementD
Check if we need to handle assumption that accumulator is same as compute.

Signed-off-by: Chawla, Amit K <[email protected]>
1. ElementC is 'void', ElementCompute and ElementOutput
different in LinearCombination

2. ElementAccumulator and ElementC have different types
D=Ax B + C;  => BF16=BF16xBF16+BF16 <=>BF16=FP32+BF16
@amitchawla1 amitchawla1 force-pushed the dev/achawla/fix_void_c_in_epilogue branch 2 times, most recently from 4363474 to 4d9276d Compare November 3, 2025 11:14
Signed-off-by: Chawla, Amit K <[email protected]>
@amitchawla1 amitchawla1 force-pushed the dev/achawla/fix_void_c_in_epilogue branch from 4d9276d to db56f06 Compare November 3, 2025 12:44
@Antonyvance Antonyvance merged commit d17d407 into intel:main Nov 4, 2025
6 checks passed
@amitchawla1 amitchawla1 deleted the dev/achawla/fix_void_c_in_epilogue branch November 5, 2025 04:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants