New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Possible changes for benchmark #15

Open

2 of 5 tasks

mondus opened this issue Jan 26, 2022 · 1 comment

Member

mondus commented Jan 26, 2022 •

edited by ptheywood

Loading

Drift guarding Drift variable guarding #14
Pop size doesn't give full utilisation on v100 (80 SMs, 2048 max threads per SM, but lower for message iteration kernels)
FP64 use is massive there is a 3.14 someone rather than 3.14f
Cube root is (maybe) in double. Is there a f version.
Broadcast access may be better as offset access. Really an issue for the main repo.

Member

ptheywood commented Apr 4, 2022

#18 addresses:

the FP64 3.141 vs 3.141f issue (confirmed by profiling that the absence of the f literal does lead to FP64 pip utilisation)
cbrt was in fp64, but it should also be sqrtf anyway for correct behaviour. This impacts model behaviour rather than being pure perf.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment