Fix for triangular_solve BEF executable unit test failure on ROCm.

[rsanthanam-amd]

/cc @chsigg @hanbinyoon

[hanbinyoon]

Nit: Could we please initialize a_num_elements to n * n? Something like:

int32_t a_num_elements = n * n;
if ((platform == wrapper::Platform::CUDA && side_mode == CUBLAS_SIDE_LEFT) ||
(platform == wrapper::Platform::ROCm && side_mode == rocblas_side_left))
a_num_elements = m * m;

Otherwise, looks good! :)

[rsanthanam-amd]

Nit: Could we please initialize a_num_elements to n * n? Something like:

int32_t a_num_elements = n * n; if ((platform == wrapper::Platform::CUDA && side_mode == CUBLAS_SIDE_LEFT) || (platform == wrapper::Platform::ROCm && side_mode == rocblas_side_left)) a_num_elements = m * m;

Otherwise, looks good! :)

Done.