DTBSM, triangular band matrix solve with multiple right-hand sides

Hi OpenBLAS community, I received an email from a BLAS/LAPACK user and they were wondering why there exists Level 2 DTBSV, and why there does not exist a Level 3 DTBSM. So a DTBSV with multiple right-hand sides. The user has an application where a Level 3 DTBSM and, if speed-up could be obtained by blocking on the multiple right-hand sides, they would not mind the speed-up. DTBSM does not exist in BLAS, and does not exist in BLAS Technical Forum Standard. Few questions: (1) Do you see any reason(s) to exclude DTBSM from BLAS? [ (drawback = Complexity, software maintenance) vs ( not clear how many users ) vs ( not clear how much speedup ) are certainly good reasons. Wondering if there are some more reasons. ] (2) Do you know of requests for DTBSM from users (or yourself)? (3) Do you have DTBSM in OpenBLAS? (typically it could fit neatly and seamlessly under LAPACK DPBTRS). Cheers, Julien.

This is an open design discussion rather than a scoped implementation task. Start with the existing DTBSV and LAPACK DPBTRS interfaces named in the issue, then review the BLAS Technical Forum Standard to establish the requested DTBSM behavior. Done would require a documented decision and, if accepted, a concrete implementation scope.