[SYCL][E2E] Change test markup to enable build-only mode on AMD
#17003
Conversation
|
Some notes:
|
| @@ -1,5 +1,5 @@ | |||
| // REQUIRES: aspect-ext_intel_legacy_image | |||
| // UNSUPPORTED: cuda || hip | |||
| // UNSUPPORTED: cuda || target-amd | |||
There was a problem hiding this comment.
This is a bit surprising that it's a runtime condition for CUDA and compile-time for AMD. tagging @intel/llvm-reviewers-cuda for any potential follow-up to this.
| @@ -3,6 +3,9 @@ | |||
| // XFAIL-TRACKER: https://github.com/intel/llvm/issues/14387 | |||
| // RUN: %{build} -o %t.out | |||
| // RUN: %{run} %t.out | FileCheck %s | |||
| // | |||
| // UNSUPPORTED: target-amd | |||
| // UNSUPPORTED-INTENDED: legacy image not supported on AMD | |||
There was a problem hiding this comment.
why isnt this failing on amd today?
There was a problem hiding this comment.
I think what is going on is that the required feature never appears on any AMD devices, so historically it was an implicit UNSUPPORTED: hip. Though I think it would be good to confirm this with people that are more familiar with these particular aspects.
| @@ -1,4 +1,6 @@ | |||
| // REQUIRES: aspect-ext_oneapi_image_array | |||
| // UNSUPPORTED: target-amd | |||
There was a problem hiding this comment.
same q on all of these on why they arent failing now
There was a problem hiding this comment.
The aspect ext_oneapi_image_array returns false when using the hip backend so this test does not run on AMD hardware. Same for L0 but that is planned to change soon.
|
@aelovikov-intel @sarnex FIY, since your approvals I added 3 new file changes, based on this CI run https://github.com/intel/llvm/actions/runs/13331013654/job/37234912237?pr=16953.
|
|
@intel/bindless-images-reviewers @intel/syclcompat-lib-reviewers @intel/sycl-graphs-reviewers @intel/unified-runtime-reviewers Friendly ping. The changes in this pr are required for enabling #16953 which is part of the ongoing activity to improve CI performance for everybody |
|
Both failures are the same as #17008 |
|
@intel/sycl-graphs-reviewers @intel/unified-runtime-reviewers Hello, would appreciate a review here. This pr is needed to continue progress with #16953 which is ready to go apart from the approval/merging of this pr. |
|
@intel/sycl-graphs-reviewers , this is unacceptable. @EwanC , @reble , @fabiomestre , @Bensuo , @konradkusiak97 |
No description provided.