[CIR][CUDA] Add attribute for CUDA fat binary name #1377
Merged
Conversation
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
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
AdUhTkJm
force-pushed
the
fatbinhandle
branch
from
de7b9cc to
0dde030
Compare
bcardosolopes
approved these changes
Feb 20, 2025
Member
bcardosolopes
left a comment
bcardosolopes
left a comment
There was a problem hiding this comment.
LGTM after fixing minor nits
AdUhTkJm
force-pushed
the
fatbinhandle
branch
from
0dde030 to
f9247af
Compare
AdUhTkJm
force-pushed
the
fatbinhandle
branch
from
f9247af to
d24050e
Compare
lanza
pushed a commit
that referenced
this pull request
Mar 18, 2025
This is a preparation of generating registration functions in LoweringPrepare. CUDA compilation works as follows (irrelevant arguments omitted): ```sh # First compile for device, generating PTX assembly clang++ test.cu -fcuda-is-device -o device.s # Convert that into a binary file ptxas device.s --output-file device.o fatbin --create device.fatbin --image=profile=sm_52,file=device.o # Pass that file as an argument to host clang++ test.cu -fcuda-include-gpubinary device.fatbin -cuid="some unique id" ``` And from the name of GPU binary, we can obtain a handle for registration. So we add an attribute to ModuleOp, recording that name. If that `-fcuda-include-gpubinary` is not specified (like in the test `simple.cu`), OG will not generate any registration function. We do the same here by not generating the attribute.
terapines-osc-cir
pushed a commit
to Terapines/clangir
that referenced
this pull request
Sep 2, 2025
This is a preparation of generating registration functions in LoweringPrepare. CUDA compilation works as follows (irrelevant arguments omitted): ```sh # First compile for device, generating PTX assembly clang++ test.cu -fcuda-is-device -o device.s # Convert that into a binary file ptxas device.s --output-file device.o fatbin --create device.fatbin --image=profile=sm_52,file=device.o # Pass that file as an argument to host clang++ test.cu -fcuda-include-gpubinary device.fatbin -cuid="some unique id" ``` And from the name of GPU binary, we can obtain a handle for registration. So we add an attribute to ModuleOp, recording that name. If that `-fcuda-include-gpubinary` is not specified (like in the test `simple.cu`), OG will not generate any registration function. We do the same here by not generating the attribute.
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.
This is a preparation of generating registration functions in LoweringPrepare.
CUDA compilation works as follows (irrelevant arguments omitted):
And from the name of GPU binary, we can obtain a handle for registration. So we add an attribute to ModuleOp, recording that name.
If that
-fcuda-include-gpubinaryis not specified (like in the testsimple.cu), OG will not generate any registration function. We do the same here by not generating the attribute.