Re: Issues and questions regarding PowerPC 32-bits e500v2 tuning

[Elinor Montmasson <elinor.montmasson@savoirfairelinux.com>]

Hello Peter,

----- Original Message -----
> From: "peter marko" <Peter.Marko@siemens.com>
> To: "Elinor Montmasson" <elinor.montmasson@savoirfairelinux.com>, "adrian freihofer" <adrian.freihofer@siemens.com>
> Cc: "yocto" <yocto@lists.yoctoproject.org>
> Sent: Wednesday, 21 October, 2026 15:50:22
> Subject: RE: Issues and questions regarding PowerPC 32-bits e500v2 tuning

> Hello Elinor,
> 
> I have already checked SPE on e500 before and can confirm that it needs to be
> disabled in favor of soft-fpu with newer GCC versions.
> It of course dramatically decreases performance, but what can we do...
> 
> As usual with Yocto, only the default settings are being tested and untested
> features slowly bit rot by time.
> For now, we don't plan to extend our pipeline to all possible tunes/variations.
> So ideally, you should send patches to fix your tune configuration and spin a
> regular local pipeline with Yocto master to catch if it would break again.
> 
> Best Regards,
>  Peter

Ok, thank you for your answer, at least it confirms my analysis.

Because GCC is not an option for SPE support, I am considering using
the LLVM/Clang toolchain through meta-clang, as it seems to support
PowerPC SPE.
I know that some specific recipes will require building with GCC anyway,
in that case fpu-soft will be enabled, but most of the distribution
could build with clang.

Is it something you have tested?

Best regards,
Elinor Montmasson