modules/cuda: completely revamp microarchitecture handling#16002
Open
SoapGentoo wants to merge 1 commit into
Open
modules/cuda: completely revamp microarchitecture handling#16002SoapGentoo wants to merge 1 commit into
SoapGentoo wants to merge 1 commit into
Conversation
dcbaker
requested changes
Jul 13, 2026
dcbaker
left a comment
dcbaker
left a comment
Member
There was a problem hiding this comment.
I'd need to think some more about the what's going on here, but here's a few high level review comments.
Member
Author
|
@dcbaker there's a lot going on I know, but I feel the current interleaving of logic and definitions isn't working anymore, and with the |
Member
|
I don't disagree, I just have to get my head back around the existing code and then around your code as well. |
Replace the per-CUDA-version architecture lists with a declarative
table of support windows (min/max CUDA version per micro-ISA). There's
a clean separation between version bounds and logic, and once a new
arch like Rubin comes out, it's a table entry instead of a change to
complex intertwined logic.
User-visible changes:
- support the specifiers introduced with CUDA 12.8/12.9:
architecture-specific 'X.Ya' ('Hopper(A)', '12.0a', ...) and
family-specific 'X.Yf' ('10.0f', ...), plus the Thor and newer
Blackwell parts
- min_driver_version() table extended through CUDA 13.3
- '+PTX' is now rejected for architecture-specific ('a') targets,
which have no forward compatibility
- warnings distinguish "CUDA too old for arch" from "arch dropped
from CUDA"
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Replace the per-CUDA-version architecture lists with a declarative table of support windows (min/max CUDA version per micro-ISA). There's a clean separation between version bounds and logic, and once a new arch like Rubin comes out, it's a table entry instead of a change to complex intertwined logic.
User-visible changes: