Introduce new key-value runtime option 'pxtrmm_for_generalized' and...
Introduce new key-value runtime option 'pxtrmm_for_generalized' and alternative codepath for GPU that uses PxTRAN+hermitian_multiply instead
Introduce new key-value runtime option 'pxtrmm_for_generalized' and alternative codepath for GPU that uses PxTRAN+hermitian_multiply instead