Add test for SYCL 2020 specialization constants #255
Conversation
AlexeySachkov
changed the title
| // RUN: %GPU_RUN_PLACEHOLDER %t.out | ||
| // FIXME: ACC devices use emulation path, which is not yet supported | ||
| // FIXME: CUDA uses emulation path, which is not yet supported | ||
| // UNSUPPORTED: cuda |
There was a problem hiding this comment.
I don't have ACC_RUN_PLACEHOLDER, so it is not necessary
|
|
||
| auto custom_type_acc = | ||
| custom_type_buffer.get_access<sycl::access::mode::read>(); | ||
| const custom_type custom_type_ref; |
There was a problem hiding this comment.
Nit: the default value of the custom_type_id coincides with default CPP (zeroes). the test could be strengthened by assigning custom_type_id some specific non-zero value to verify new SYCL2020 capabilities work for non-primitive types as well.
There was a problem hiding this comment.
Nit: the default value of the
custom_type_idcoincides with default CPP (zeroes).
It should have some non-zero default values, as I have its fields initialized: https://github.com/intel/llvm-test-suite/pull/255/files#diff-308baade624f8087060f981e03c24cffa77255b25f54deb6f7b7f1f6349a84afR16
| return false; | ||
|
|
||
| return true; | ||
| } |
There was a problem hiding this comment.
For completeness, kernel_bundle-API based should also be tested. I guess these would be dozen extra lines of code. Can be addressed in a separate PR of course.
There was a problem hiding this comment.
For completeness, kernel_bundle-API based should also be tested. I guess these would be dozen extra lines of code.
Sure, I definitively don't want to add those into the same test as it is huge enough already, but a separate file with such tests is needed for sure
|
@vladimirlaz, if I recall correctly, @kbobrovs approved this PR once. Since then only formatting changed - I suggest that we merge the PR and if there would be any additional feedback, I will apply it in a separate PR. |
No description provided.