[SW] Add kernels: gemv, spmv, conjugate gradient

[husterZC]

Description of PR that completes issue here...

Added

• spmv
• gemv
• conjugate gradient

[mp-17]

Thanks a lot @husterZC, I will check it asap!

[husterZC]

Hi Matteo, one thing to notice:

In the conjugate_gradient/shared_kernels. these are all soft link files, but I don't know why we I push them, they loss the link. Could you please fix them yourself, either make a hard copy or make a link file. And for some workloads using other kernels, do you suggest to make a link or a hard copy?

[mp-17]

I would prefer symlinks to avoid duplicated code. I created a PR to your main: https://github.com/husterZC/ara/pull/1

[mp-17]

Thanks a lot for the contribution!
Can you please pass everything through clang-format and adjust the vertical alignment in apps/common/default_args.mk?
Then, if the CI checks pass, LGTM!

[mp-17]

@husterZC Ping me if you need support! :-)

[husterZC]

To be honest, I'm busy on other stuff. I plan to fix this last in this week

[husterZC]

Hey Matteo,

I've added the dependence of sklearn python package for conjugate gradient kernel.
We need this one to generate the symmetric and positive-defined (SPD) dense/sparse matrix for conjugate gradient
If you find it is a bit troublesome to add new dependency, there are other solutions:

1. I can bring a pre-build SPD matrix and delete the function to generate SPD matrix at given size and sparsity. Well that would be very boring, and not very meaningful.
2. I can look into how to generate SPD matrix at given size and sparsity, drive myself to be a mathematician and take somedays of working.

[mp-17]

https://github.com/pulp-platform/ara/actions/runs/3563364133/jobs/5989785427#step:8:289

[husterZC]

Hi Matteo, I wonder how to add the python package "sklearn" in the CI environment, otherwise it will always sticking here ;)

[mp-17]

I will check everything and merge here -> #267
Thanks again Chi!