[MKL][ROCBLAS] support the omatcopy and omatadd functions in row and column majors

[jinz2014]

Thank you for your review.

[muhammad-tanvir-1211]

Hi @jinz2014
Similar to #333, do you mind attaching the test log and confirm if clang-format has indeed been applied to this patch? Thanks

[jinz2014]

I will need to update this PR. Thanks for your reminder.

[jinz2014]

@muhammad-tanvir-1211
I tried to update the files. The build is successful, but there are failures when running the ctests. Could you please run the ctests and let me know if they work ? Thank you.

[muhammad-tanvir-1211]

Hi @jinz2014, apologies for the delay. I was able to recreate the test failures locally. I have suggested the fix for this in the code. Thanks.

[jinz2014]

@muhammad-tanvir-1211 The codes were updated with your suggestions. Thanks.

[muhammad-tanvir-1211]

Great, could you run the tests locally to see if they are passing on your end as well, and attach the log here?

[jinz2014]

I have a question. Did you run the tests successfully on your end with your suggested changes ?

[muhammad-tanvir-1211]

Yes, I did.

[jinz2014]

There are segfaults when running 'ctest'.

Note: Google Test filter = OmataddTestSuite/OmataddTests.RealSinglePrecision/Column_Major_AMD_Instinct_MI100
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from OmataddTestSuite/OmataddTests
[ RUN      ] OmataddTestSuite/OmataddTests.RealSinglePrecision/Column_Major_AMD_Instinct_MI100
[New Thread 0x7fffea9ff640 (LWP 368626)]
[       OK ] OmataddTestSuite/OmataddTests.RealSinglePrecision/Column_Major_AMD_Instinct_MI100 (563 ms)
[----------] 1 test from OmataddTestSuite/OmataddTests (563 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (563 ms total)
[  PASSED  ] 1 test.

Thread 4 "test_main_blas_" received signal SIGSEGV, Segmentation fault.
[Switching to thread 4 (Thread 0x7fffea9ff640 (LWP 368626))]
0x00007ffff2c30bfd in ?? () from /opt/rocm/hip/lib/libamdhip64.so.5
(gdb) bt
#0  0x00007ffff2c30bfd in ?? () from /opt/rocm/hip/lib/libamdhip64.so.5
#1  0x00007ffff2c3119a in ?? () from /opt/rocm/hip/lib/libamdhip64.so.5
#2  0x00007ffff2a7bd6f in ?? () from /opt/rocm/hip/lib/libamdhip64.so.5
#3  0x00007ffff2a81eb7 in hipFree () from /opt/rocm/hip/lib/libamdhip64.so.5
#4  0x00007ffdd2cf0320 in _rocblas_handle::~_rocblas_handle() () from /opt/rocm-5.4.0/lib/librocblas.so.0
#5  0x00007ffdd2cf1f96 in rocblas_destroy_handle () from /opt/rocm-5.4.0/lib/librocblas.so.0
#6  0x00007fffebd413a8 in oneapi::mkl::blas::rocblas::rocblas_handle_container<_pi_context*>::~rocblas_handle_container() ()
   from /noback/5zj/codeplay/oneMKL/build-rocblas2/lib/libonemkl_blas_rocblas.so
#7  0x00007ffff4595d9f in __GI___call_tls_dtors () at ./stdlib/cxa_thread_atexit_impl.c:159
#8  0x00007ffff45e49c5 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:450
#9  0x00007ffff4676a00 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

Could you please paste the test log ?

[muhammad-tanvir-1211]

Hi @jinz2014,
Please find attached the changes I made to rocblas_extensions file and the relevant test logs. Thanks.

rocblas_changes.log
omatcopy_omatadd.log

[jinz2014]

A few tests are skipped in the log. I am not sure which tests should be skipped. Can you please have another reviewer evaluate the PR ? Thanks.

[muhammad-tanvir-1211]

Maybe @andrewtbarker can help?

[andrewtbarker]

/intelci: run

[andrewtbarker]

It looks like it is skipping the batch tests, but the added omatcopy and omatadd tests are passing. That's fine, we can leave the batch versions unimplemented for now and put them in a future PR if necessary.

[jinz2014]

Thank you for your comments

[andrewtbarker]

The CI doesn't like the way the edits are formatted. Can you check clang-format again and make sure you're using the "correct" version 9.0.0 (yes, that's an old version, but it's the standard for this repo).

[andrewtbarker]

Thanks for the additions!

[jinz2014]

Thank you for your suggestion.