-
Notifications
You must be signed in to change notification settings - Fork 787
[SYCL][ROCm] Setup lit tests for ROCm plugin #4163
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.
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
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM. Thanks!
|
||
if(SYCL_BUILD_PI_ROCM) | ||
add_custom_target(check-sycl-rocm) | ||
if("${SYCL_BUILD_PI_ROCM_PLATFORM}" STREQUAL "NVIDIA") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Very minor comment.
See the discussion in #4069. In case we need to support other spellings for SYCL_BUILD_PI_ROCM_PLATFORM
, we might need to make sure it's all capital.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah that's probably a good idea, however there's already a bunch of other places that do this comparison. So I'd suggest to keep it like that in this patch and then once it's merged I can put up a new PR updating all of the uses of this to be case-insensitive.
I've marked the remaining two failing tests with NVIDIA as XFAIL, so |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, although there is some contradiction between this and another your patch - #4239. I think they suggest using different options for setting target architecture.
Yes there is, because #4239 changes the option that should be used. The PRs are independent at the moment so either one can go in first but I'll have to update the other to resolve conflict and/or update the new changes. |
@DenisBakhvalov, @kbobrovs, @kychendev, ping. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Changes in sycl/test/esimd/odr.cpp
look good.
LGTM |
We are not doing ROCm plug-in validation on NVIDIA cards, so I'm not sure if it make sense to block the merge of this pull request by #4123. |
Yes, the tests addressed in #4123 are marked as XFAIL here, so this MR should pass I do need to change the |
If merging this patch is not urgent, let's make this change here. There should be no difference in where such change is reviewed - within this PR or separately. |
4fcb1e7
Rebased and updated the uses of |
@v-klochkov, it looks like failure in reduction test on Windows is not related to the patch. Please, take a look. |
Fix for timeout on ACC: |
…ut (#5032) A follow up to #4163 (comment) `buildbot` script only allows `NVIDIA` capitalise the variable in case user spelled it differently.
This patch wires up the lit tests for the ROCm plugin adding
check-sycl-rocm
.check-sycl-rocm
includescheck-sycl-rocm-on-device
andcheck-sycl-rocm-ptx
to run the compiler tests with the NVIDIA triple if using the NVIDIA platform for ROCm orcheck-sycl-rocm-gcn
if using the AMD platform.This PR is marked as draft is it requires a bit more work to get
check-sycl-rocm
to work properly.On AMD it should pass after these two patches are merged/fixed:
On NVIDIA it requires a bit more test filtering and/or debugging because of issues in the HIP APIs, #4123 is one of them but there is a few other bugs that need to be addressed or filtered before this PR can be merged.