On Jun 29, 2026, "Robin Dapp" <[email protected]> wrote:

> Rather than adding a third option, I'd prefer regularizing the existing once 
> to 
> match what others are doing, i.e.
>  riscv_v -> can compiler vector (with or without added options, add options 
> if 
>  necessary)
>  riscv_v_hw -> can execute vector

There are lots of riscv*_ok patterns that follow the riscv_v_ok
exceptional pattern.  Adapting all those tests was far too much work;
not so much because of s/_ok$//, but because those that don't have _ok,
such as riscv_v, don't add-options riscv_v, which they'd need to do once
the behavior of riscv_v_ok is folded into riscv_v.

So I figured an incremental approach was the only viable path to address
this.  I figured that meant adding a third choice, starting its
adoption, and progressively phase out existing uses of riscv_v into
riscv_v_ok, until they're all gone and we can mechanically replace
riscv_v_ok with riscv_v_hw and riscv_v_okc with riscv_v.

Other uses of *riscv*_ok procs might be thrown into the list of
to-be-phased-out-and-renamed patterns.

I expect this would be a long and error-prone process.

Considering how often I've hit backported riscv testcases that fail in
the older branches, this should also be taken into account for the
phasing-out plan.

It's far too much for me to tackle :-(

> I realize there are tons of tests that don't follow a proper scheme but I 
> would 
> hope rewriting to be mostly mechanical?

I wish!

There's lack of clarity of intent in tests that require riscv_v.
Presumably a mechanical rewrite into requiring riscv_v_ok and
add-options riscv_v would do, but IIRC I found a number of them that
overrode compile flags, which made me feel that they should be tackled
by someone who knew what the tests meant to accomplish.

-- 
Alexandre Oliva, happy hacker            https://blog.lx.oliva.nom.br/
Free Software Activist     FSFLA co-founder     GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity.
Excluding neuro-others for not behaving ""normal"" is *not* inclusive!

Reply via email to