[SYCL][CUDA][BFLOAT16]Temp oneapi test file adds unary cov.. #889
Conversation
..erage fixes broken verify_logic test. Once a backend agnostic ext_oneapi_bf16 aspect is added we can add the patches from this file to the existing bfloat16 test. Until then the cuda backend requires the Target provided at compile time to use the Target_Builtins.
|
This PR can be used to test intel/llvm#5393. Once intel/llvm#5393 is merged I will move the cuda specific test file into the existing bfloat16_type.cpp file. Having two separate files for now helps me be sure that both test files should pass. |
intel/llvm#5393 must be committed with the test changes, otherwise llvm-test-suite will report failures in pre-commit as you can see in intel/llvm#5393 test results. |
Integrated CUDA specific changes into bfloat16_type.cpp as well as new unary neg unit test case.
|
/verify with intel/llvm#5393 |
I've fixed the formatting and used a single test file. The test passes for me locally for all backends using intel/llvm#5393 (As previously the test is only built for backends other than CUDA, now the test is also executed for the CUDA case). |
The test failed in CI system for all backends using intel/llvm#5393. Intel GPUNVIDIA GPUDoes it make sense to re-run the tests? These were gathered 3 days ago. |
I understand the failures now:
Although I do not see the link for http://pre-commit.llvm.intel.com:8013/file_review?path_name=llvm_test_suite_associate/PR5393 on this page, I can access the link. However it only shows the tests from 31/3 and 1/4. I cannot see the test that was executed today. Interestingly the 31/3 test passed (When I originally used the two files) even though the checks were not successful according to this page. This may mean that the checks executed today were successful? But I don't have access to the logs. |
|
You should add another "/verify with intel/llvm#5393" comment to re-run the tests with intel/llvm#5393. Jenkins systems is using latest available nightly build for testing by default. |
Ah OK thanks. |
|
/verify with intel/llvm#5393 |
|
/verify with intel/llvm#5393 |
Thanks for the tips. The new logs have appeared for me at the bottom and have passed. |
…5393) There is a bug in the verify_logic function in the bfloat16_type.cpp test (C accessor is not written to) - I'm not sure how this did not lead to a failure already. With the bug fixed the test passes for the CUDA backend with this patch. I've added a draft test file that also increases the coverage to test unary minus operator here: intel/llvm-test-suite#889. Note that the unary neg intrinsic added here that is used in unary minus will be pulled down from upstream via e.g. https://reviews.llvm.org/D117887.
…el#889) Fixes broken verify_logic test. Once a backend agnostic ext_oneapi_bf16 aspect is added we can add the patches from this file to the existing bfloat16 test. Until then the cuda backend requires the Target provided at compile time to use the Target_Builtins.
CMPLRTST-16319: skip more double type tests on GEN11/GEN12 GPU
…el/llvm-test-suite#889) Fixes broken verify_logic test. Once a backend agnostic ext_oneapi_bf16 aspect is added we can add the patches from this file to the existing bfloat16 test. Until then the cuda backend requires the Target provided at compile time to use the Target_Builtins.
..erage
fixes broken verify_logic test.
Requires: intel/llvm#5393
Once a backend agnostic ext_oneapi_bf16 aspect is added we can add the patches from this file to the existing bfloat16 test.
Until then the cuda backend requires the Target provided at compile time to use the Target_Builtins.