Performance improvement of transposed [SD]GEMV on A64FX and Neoverse V1.

[iha-taisei]

I provided Pull Request #4803 for SVE-enablement of [SD]GEMV on A64FX, but there is still room for performance improvement.
Therefore, I'd like to propose another patch for such improvement of transposed [SD]GEMV on A64FX and Neoverse V1.

[Mousius]

Hi @iha-taisei,

It's always good to keep adding new optimized kernels 😸

How would this be different from https://github.com/OpenMathLib/OpenBLAS/blob/develop/kernel/arm64/gemv_t_sve.c ?

[iha-taisei]

Hi @Mousius,

As you see above, I did loop-unrolling too.

[martin-frbg]

closing as resolved/implemented by your #4996 - thank you very much