[CIR][CUDA] multi-arch support for the CIR offload-merge pipeline#9
Open
RiverDave wants to merge 1 commit into
Conversation
Owner
Author
This stack of pull requests is managed by Graphite. Learn more about stacking. |
RiverDave
changed the title
RiverDave
changed the title
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Depends on: #7
This adds CUDA multi-arch support to the CIR offload-merge pipeline.
The stock CUDA driver compiles each GPU arch separately, then collapses the per-arch device outputs into one CUDA fatbinary before passing it to the host via
-fcuda-include-gpubinary. The CIR merge path now mirrors that: after splitting the combined CIR container, each device module resumes backend/assemble for its own arch, and the resulting outputs are packaged into one fatbin for the resumed host compile.This also makes
OffloadActionpreserve the bound arch per device dependence. The split action is a shared node, so the arch cannot live only on the action itself; each dependence needs to carry the arch it is selecting from the split.