Repository navigation
Conversation
Aiden2244
force-pushed
the
archspec/support-gpus
branch
5 times, most recently
from
October 1, 2026 16:23
7206ff0 to
9584da7
Compare
Aiden2244
force-pushed
the
archspec/support-gpus
branch
from
October 6, 2026 15:50
9584da7 to
5e4c97f
Compare
Aiden2244
force-pushed
the
archspec/support-gpus
branch
2 times, most recently
from
October 20, 2026 17:49
f9c77e1 to
553976c
Compare
Adds the archspec.gpu package and its detection JSON, plus the "archspec gpu" CLI subcommand, from the archspec detect-gpus branch. Needed by the upcoming concretizer change that uses GPUs detected on the host as variant defaults. Signed-off-by: Aiden McCormack <aiden.mccormack@kitware.com>
Aiden2244
force-pushed
the
archspec/support-gpus
branch
2 times, most recently
from
October 22, 2026 17:03
895fbf1 to
3fd39a8
Compare
Signed-off-by: Aiden McCormack <aiden.mccormack@kitware.com>
Signed-off-by: Aiden McCormack <aiden.mccormack@kitware.com>
Aiden2244
force-pushed
the
archspec/support-gpus
branch
3 times, most recently
from
October 23, 2026 16:28
7cba474 to
952bb3e
Compare
Signed-off-by: Aiden McCormack <aiden.mccormack@kitware.com>
Bare cuda_arch / amdgpu_target values are compared exactly, by the variant that holds them, as they are on develop. Only a range form will defer to the GPU logic on ArchSpec. Signed-off-by: Aiden McCormack <aiden.mccormack@kitware.com>
To preserve existing behavior, comparisons between GPU-variant specs will match based on exact values of cuda_arch/amdgpu_target unless the value of those variants is explicitly marked with a leading colon (e.g. ``cuda_arch=:86``). This syntax is based on CPU target ranges and indicates that the comparison should be a compatibility check instead of an exact match. Signed-off-by: Aiden McCormack <aiden.mccormack@kitware.com>
Aiden2244
force-pushed
the
archspec/support-gpus
branch
from
October 23, 2026 22:29
952bb3e to
a96e77b
Compare
This branch has not been deployed
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
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.
Add feature to spack which uses archspec to auto-detect GPU microarchitecture identifiers for NVIDIA and AMD GPUs when the
+cudaor+rocmvariants are specified without versions.This is blocked on archspec #247 which adds GPU detection capability for archspec.
TODO: At current, this works by making the
+cudaand+rocmvariants special cases in the solver and overriding the package defaults. This is not desirable, and the relevant logic is to be refactored into the Spack packages API.