Skip to content

fp32 kernel main loop improvements - #12

Open
ylh440104 wants to merge 1 commit into
deepseek-ai:mainfrom
ylh440104:fp32-main-loop-opt
Open

ylh440104 wants to merge 1 commit into
deepseek-ai:mainfrom
ylh440104:fp32-main-loop-opt

Conversation

@ylh440104

Copy link
Copy Markdown

hoist loop invariant offsets and swizzle masks out of the loop
add adaptive reconstruct trigger for long rows to match the bf16 scan path
skip reconstruct on the last round

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.

1 participant