Skip to content
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

Drop per-architecture properties which are actually per-compute-capability #293

Closed
eyalroz opened this issue Jan 14, 2022 · 0 comments
Closed

Drop per-architecture properties which are actually per-compute-capability #293

eyalroz opened this issue Jan 14, 2022 · 0 comments
Assignees
@eyalroz
Labels
resolved-on-development task

Comments

@eyalroz
Copy link
Owner

eyalroz commented Jan 14, 2022

Some properties we offer for the compute_architecture_t type used to really be per-arch, but with Volta-vs-Turing, and now Ampere 8.0 vs Ampere 8.6, have really become per-compute-capability. It is confusing and possibly wrong to let people rely on per-architecture values, e.g. if another GPU is added that differs from the per-arch values.
@eyalroz eyalroz self-assigned this Jan 14, 2022
@eyalroz eyalroz added the task label Jan 14, 2022
eyalroz added a commit that referenced this issue Jan 14, 2022
@eyalroz
Fixes #292, #293:
30d6b22 
* Device properties functions now properly support Ampere GPUs (8.0 and 8.6);
* Corrected some device-properties per-compute-capability values.
* Dropped the not-really-per-architecture values in favor of simply per-compute-capability values.
* No explaining where I got the "max in flight threads per SM" value from.
* Comment tweaks
* Exception description tweak for when an architecture is not known to us.
eyalroz added a commit that referenced this issue Jan 14, 2022
@eyalroz
Fixes #292, #293:
0478bf7 
* Device properties functions now properly support Ampere GPUs (8.0 and 8.6);
* Corrected some device-properties per-compute-capability values.
* Dropped the not-really-per-architecture values in favor of simply per-compute-capability values.
* No explaining where I got the "max in flight threads per SM" value from.
* Comment tweaks
* Exception description tweak for when an architecture is not known to us.
eyalroz added a commit that referenced this issue Jan 14, 2022
@eyalroz
Fixes #292, #293:
a0e70fd 
* Device properties functions now properly support Ampere GPUs (8.0 and 8.6);
* Corrected some device-properties per-compute-capability values.
* Dropped the not-really-per-architecture values in favor of simply per-compute-capability values.
* No explaining where I got the "max in flight threads per SM" value from.
* Comment tweaks
* Exception description tweak for when an architecture is not known to us.
@eyalroz eyalroz closed this as completed Jan 14, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@eyalroz eyalroz

Labels
resolved-on-development task
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@eyalroz