[PATCH v2 1/4] riscv: Add multiarch scaffolding for str*/mem* routines

[Jeff Law]

On 10/9/25 10:46 AM, Adhemerval Zanella Netto wrote:


> 
> Also, I think you might want to coordinate with other parties, it seems that we
> have at least 3 different submissions to add RVV multiarch support with slightly
> different semantics [2] [3] [4].
I can deal with the coordination issues.  Much of my work and I Yao's is 
based on their earlier efforts.

> 
> 
> For the patches, the scaffolding for str*/mem* routines does not make much
> sense [5]. Each new fallback implementation should be submitted along with t
> he new RVV implementation.
I'd tend to agree.  Hence my suggestion we wrap memset as a whole first, 
then proceed to the next routine.  memset is about ready to push over 
the finish line after which everything else becomes a lot easier.


> 
> 
> You need to decide which mechanism to use for iFUNC selection: hwprobe or hwcap.
> Having both is just confusing.
hwprobe is the right way to go here.  This was discussed in the BOF a 
couple weeks ago.

Jeff