[BLAS:: SYCL-BLAS backend] Enable sycl blas routines

[pgorlani]

Description

This PR enables the following routines in the SYCL-BLAS backend: sbmv, tbmv, spmv, tpmv, tbsv, trsv and rotmg.

This patch depends on #262 and #264.

Checklist

All Submissions

• [*] Do all unit tests pass locally? Logs: test_results.txt
  Failing ones are due to requiring fp64 support: Intel(R) UHD Graphics 770 [0x4680] is not supported.

• [*] Have you formatted the code using clang-format?

New features

• [*] Have you provided motivation for adding a new feature?
• [*] Have you added relevant tests?

[mkrainiuk]

Looks good to me, thank you!

[muhammad-tanvir-1211]

Thanks @mkrainiuk.
@andrewtbarker could you please take a look at this PR here and let us know if we need to make any changes?

[andrewtbarker]

Sorry for the late review. This looks fine to me, can you test the newly supported APIs on nvidia hardware to make sure that configuration is working before we merge?

[pgorlani]

Sorry for the late review. This looks fine to me, can you test the newly supported APIs on nvidia hardware to make sure that configuration is working before we merge?

No problem! Here are the test logs for the NVIDIA gpu, NVIDIA_GPU_test_results.txt.

Thanks for your reviews, @mkrainiuk and @andrewtbarker.

We are planning to enable more SYCL-BLAS routines for the future. This will involve other PRs very similar to this, i.e., each changed line will add a routine. In this regard, we are wondering what could be an ideal PR size that could ease your reviewing tasks.

Do you prefer fewer PRs containing more routines, or more PRs containing fewer routines?

[andrewtbarker]

Do you prefer fewer PRs containing more routines, or more PRs containing fewer routines?

Speaking only for myself, for a PR like this I think it makes sense to include many routines in few PRs. I like to run the tests locally for confirmation and that will be easier to do this way.

If you were changing logic / algorithms it would be a different story, then smaller PRs would likely be better.

[mkrainiuk]

Do you prefer fewer PRs containing more routines, or more PRs containing fewer routines?

Speaking only for myself, for a PR like this I think it makes sense to include many routines in few PRs. I like to run the tests locally for confirmation and that will be easier to do this way.

If you were changing logic / algorithms it would be a different story, then smaller PRs would likely be better.

Same here, thanks!

[pgorlani]

Do you prefer fewer PRs containing more routines, or more PRs containing fewer routines?

Speaking only for myself, for a PR like this I think it makes sense to include many routines in few PRs. I like to run the tests locally for confirmation and that will be easier to do this way.
If you were changing logic / algorithms it would be a different story, then smaller PRs would likely be better.

Same here, thanks!

I was thinking the same. Thanks for your feedbacks, @mkrainiuk and @andrewtbarker.