minor code cleanup

[aartbik]

No description provided.

[cliffburdick]

/build

[cliffburdick]

Unit test is failing:

[----------] 3 tests from MatmulSparseTestsAll/1, where TypeParam = cuda::std::__4::tuple<matx::matxHalf<__nv_bfloat16>, matx::cudaExecutor>
[ RUN      ] MatmulSparseTestsAll/1.MatmulCOO
 ** On entry to cusparseSpMM_bufferSize() parameter number 6 (matB) had an illegal value: CUDA_R_16BF (__nv_bfloat16) is not supported on GPUs with compute capability < 8.0 (current: 7.5)

[aartbik]

That's strange, and certainly not related to this PR

But do we have a mixture of testing machines?

capability < 8.0 (current: 7.5)

Does MatX have a minimum capability requirement? If so, we need more guards I think

[cliffburdick]

/build

[cliffburdick]

That's strange, and certainly not related to this PR

But do we have a mixture of testing machines?

capability < 8.0 (current: 7.5)

Does MatX have a minimum capability requirement? If so, we need more guards I think

We should work on >= 7.0. Would be good to eventually see why 7.5 is not working for this.

[aartbik]

Agreed. I will follow up to avoid this:

CUDA_R_16BF (__nv_bfloat16) is not supported on GPUs with compute capability < 8.0 (current: 7.5)