[NFC][SYCL][MATRIX] Add std::ignore for unused parameters to fix warnings.

[arnamoy10]

No description provided.

[dkhaldi]

LGTM

[bader]

@intel/dpcpp-esimd-reviewers, @arnamoy10, it seems like ESIMD tests are broken in pre-commit. It doesn't look like this patch can cause such failures.

********************
Timed Out Tests (8):
  SYCL :: lsc/lsc_usm_block_load_u8_u16.cpp
  SYCL :: lsc/lsc_usm_block_load_u8_u16_64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u32.cpp
  SYCL :: lsc/lsc_usm_prefetch_u32_64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u32_scalar_off.cpp
  SYCL :: lsc/lsc_usm_prefetch_u64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u64_64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u64_scalar_off.cpp

********************
Failed Tests (25):
  SYCL :: lsc/lsc_surf_load_u16u32.cpp
  SYCL :: lsc/lsc_surf_load_u32.cpp
  SYCL :: lsc/lsc_surf_load_u32_stateless.cpp
  SYCL :: lsc/lsc_surf_load_u64.cpp
  SYCL :: lsc/lsc_surf_load_u64_stateless.cpp
  SYCL :: lsc/lsc_surf_load_u8_u16.cpp
  SYCL :: lsc/lsc_surf_load_u8_u16_stateless.cpp
  SYCL :: lsc/lsc_surf_load_u8u32.cpp
  SYCL :: lsc/lsc_surf_prefetch_u16u32.cpp
  SYCL :: lsc/lsc_surf_prefetch_u32.cpp
  SYCL :: lsc/lsc_surf_prefetch_u64.cpp
  SYCL :: lsc/lsc_surf_prefetch_u8u32.cpp
  SYCL :: lsc/lsc_surf_store_u16u32.cpp
  SYCL :: lsc/lsc_surf_store_u32.cpp
  SYCL :: lsc/lsc_surf_store_u32_stateless.cpp
  SYCL :: lsc/lsc_surf_store_u64.cpp
  SYCL :: lsc/lsc_surf_store_u64_stateless.cpp
  SYCL :: lsc/lsc_surf_store_u8_u16.cpp
  SYCL :: lsc/lsc_surf_store_u8u32.cpp
  SYCL :: lsc/lsc_usm_block_load_u32.cpp
  SYCL :: lsc/lsc_usm_block_load_u32_64.cpp
  SYCL :: lsc/lsc_usm_block_load_u32_scalar_off.cpp
  SYCL :: lsc/lsc_usm_block_load_u64.cpp
  SYCL :: lsc/lsc_usm_block_load_u64_64.cpp
  SYCL :: lsc/lsc_usm_block_load_u64_scalar_off.cpp

[v-klochkov]

@intel/dpcpp-esimd-reviewers, @arnamoy10, it seems like ESIMD tests are broken in pre-commit. It doesn't look like this patch can cause such failures.

********************
Timed Out Tests (8):
  SYCL :: lsc/lsc_usm_block_load_u8_u16.cpp
  SYCL :: lsc/lsc_usm_block_load_u8_u16_64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u32.cpp
  SYCL :: lsc/lsc_usm_prefetch_u32_64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u32_scalar_off.cpp
  SYCL :: lsc/lsc_usm_prefetch_u64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u64_64.cpp
  SYCL :: lsc/lsc_usm_prefetch_u64_scalar_off.cpp

********************
Failed Tests (25):
  SYCL :: lsc/lsc_surf_load_u16u32.cpp
  SYCL :: lsc/lsc_surf_load_u32.cpp
  SYCL :: lsc/lsc_surf_load_u32_stateless.cpp
  SYCL :: lsc/lsc_surf_load_u64.cpp
  SYCL :: lsc/lsc_surf_load_u64_stateless.cpp
  SYCL :: lsc/lsc_surf_load_u8_u16.cpp
  SYCL :: lsc/lsc_surf_load_u8_u16_stateless.cpp
  SYCL :: lsc/lsc_surf_load_u8u32.cpp
  SYCL :: lsc/lsc_surf_prefetch_u16u32.cpp
  SYCL :: lsc/lsc_surf_prefetch_u32.cpp
  SYCL :: lsc/lsc_surf_prefetch_u64.cpp
  SYCL :: lsc/lsc_surf_prefetch_u8u32.cpp
  SYCL :: lsc/lsc_surf_store_u16u32.cpp
  SYCL :: lsc/lsc_surf_store_u32.cpp
  SYCL :: lsc/lsc_surf_store_u32_stateless.cpp
  SYCL :: lsc/lsc_surf_store_u64.cpp
  SYCL :: lsc/lsc_surf_store_u64_stateless.cpp
  SYCL :: lsc/lsc_surf_store_u8_u16.cpp
  SYCL :: lsc/lsc_surf_store_u8u32.cpp
  SYCL :: lsc/lsc_usm_block_load_u32.cpp
  SYCL :: lsc/lsc_usm_block_load_u32_64.cpp
  SYCL :: lsc/lsc_usm_block_load_u32_scalar_off.cpp
  SYCL :: lsc/lsc_usm_block_load_u64.cpp
  SYCL :: lsc/lsc_usm_block_load_u64_64.cpp
  SYCL :: lsc/lsc_usm_block_load_u64_scalar_off.cpp

Most likely that is the typical scenario with Old(morning) compiler build used by CI and NEW tests.
We recently merged 2 commits:
PR: #8316
Test PR: #8316

PR (8316) did not catch the morning compiler build time.

@fineg74 - can you please check/confirm if that is new regression caused by your fix, or just bad timing for the merge?

[bader]

@v-klochkov, I checked before commenting that @arnamoy10's branch is based on HEAD of sycl branch, so all compiler commits are there. AFAIK, CI takes HEAD of intel branch in llvm-test-suite repository. You put wrong link for the test PR. Could you tell of test PR was committed less than hour ago?

[fineg74]

I am not sure it is related to the change but likely some emulator issue:
The test was merged together with compiler and bad timing would lead to compilation errors rather than execution errors
All the tests fail with following error:
EMU: *** Error --------------------------------------------------------------------------
Received signal 6 SIGABRT. Terminating.

command stderr:

terminate called after throwing an instance of 'sycl::_V1::memory_allocation_error'
what():
EMU: *** Error --------------------------------------------------------------------------
Received signal 6 SIGABRT. Terminating

The only change that could be related to allocation is use of aligned memory allocation i.e. sycl::aligned_alloc_shared and sycl::usm_allocator<T, sycl::usm::alloc::host, Flags::template alignment<__ESIMD_DNS::__raw_t>> in the tests. The code patch doesn't do any allocation.
I am not sure that there is a difference between regular allocation vs aligned allocation from emulator perspective.
Not every test affected by the change has failed here and different CI runs produce different set of failing sets (https://github.com/intel/llvm/actions/runs/4439865937/jobs/7794571728 for example has 3 timeouts and 34 failures vs 8 timeouts and 25 failures here)

[steffenlarsen]

I am merging this as it is very clearly a NFC and it would be best to get post-commit building again. Feel free to continue the discussion.

[bader]

I am not sure it is related to the change but likely some emulator issue: The test was merged together with compiler and bad timing would lead to compilation errors rather than execution errors All the tests fail with following error: EMU: *** Error -------------------------------------------------------------------------- Received signal 6 SIGABRT. Terminating.

command stderr:

terminate called after throwing an instance of 'sycl::_V1::memory_allocation_error' what(): EMU: *** Error -------------------------------------------------------------------------- Received signal 6 SIGABRT. Terminating

The only change that could be related to allocation is use of aligned memory allocation i.e. sycl::aligned_alloc_shared and sycl::usm_allocator<T, sycl::usm::alloc::host, Flags::template alignment<__ESIMD_DNS::__raw_t>> in the tests. The code patch doesn't do any allocation. I am not sure that there is a difference between regular allocation vs aligned allocation from emulator perspective. Not every test affected by the change has failed here and different CI runs produce different set of failing sets (https://github.com/intel/llvm/actions/runs/4439865937/jobs/7794571728 for example has 3 timeouts and 34 failures vs 8 timeouts and 25 failures here)

@fineg74, @intel/dpcpp-esimd-reviewers, please, investigate and fix this issues as soon as possible. This issue impact pre-commit validation.
See https://github.com/intel/llvm/actions/runs/4447776170/jobs/7810525573?pr=8688 - pre-commit results for #8688. I need time for analysis, I suggest disabling tests temporarily.

[bader]

Failed for my NFC patch too - https://github.com/intel/llvm/actions/runs/4451701743/jobs/7818993494#step:22:2268