Skip to content

Disable scalar BF16 operators on current GCC versions#2733

Merged
copybara-service[bot] merged 1 commit into
google:masterfrom
vekterli:vekterli/no-scalar-bf16-operators-on-gcc
Sep 26, 2025
Merged

Disable scalar BF16 operators on current GCC versions#2733
copybara-service[bot] merged 1 commit into
google:masterfrom
vekterli:vekterli/no-scalar-bf16-operators-on-gcc

Conversation

@vekterli
Copy link
Copy Markdown
Contributor

GCC >= 13 will insert a function call to the __extendbfsf2 helper function for scalar conversions from __bf16 to float. This is prohibitively expensive, so refrain from using scalar BF16 operators on these compiler versions.

See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121853

Ref: #2703 and #2699

GCC >= 13 will insert a function call to the `__extendbfsf2`
helper function for scalar conversions from `__bf16` to `float`.
This is prohibitively expensive, so refrain from using scalar
BF16 operators on these compiler versions.

See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121853
Copy link
Copy Markdown
Member

@jan-wassenberg jan-wassenberg left a comment

Choose a reason for hiding this comment

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

Nice comment, thanks for adding.

@copybara-service copybara-service Bot merged commit c20907b into google:master Sep 26, 2025
14 of 36 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants