Mixed precision dot product example#3191
Conversation
a10acff to
ae9553b
Compare
jan-wassenberg
left a comment
There was a problem hiding this comment.
Thanks for adding - a few minor suggestions :)
|
There is a function |
|
Thanks for updating :) For copying to aligned, I'd recommend just hwy::CopyBytes, no need to write a new function. But it's likely unnecessary: as long as pointers are aligned, LoadU will be as efficient as Load. Load just ensures the binary will crash if there is misalignment. So no need for a separate function, we can just remove the extra aligned path. |
|
Removed the aligned path. |
|
Removed some unnecessary includes and updated the description in the README |
jan-wassenberg
left a comment
There was a problem hiding this comment.
Typo on promtion but let's land as-is before the meeting :)
No description provided.