Re: [PATCH 2/4] target/ppc: added solutions for building with disable-tcg

[David Gibson]

On Mon, Apr 12, 2021 at 08:40:47AM -0700, Richard Henderson wrote:
> On 4/11/21 10:08 PM, David Gibson wrote:
> > Not directly related to what you're trying to accomplish here, but the
> > whole vscr_sat thing looks really weird.  I have no idea why we're
> > splitting out the storage of VSCR[SAT] for the TCG case at all.  If
> > you wanted to clean that up as a preliminary, I'd be grateful.
> 
> That's about efficiently implementing the vector saturation instructions in
> TCG.  See GEN_VXFORM_SAT in translate/vmx-impl.c.inc.
> 
> The SAT bit is represented as a vector that e.g. compares the result of
> addition with the result of saturating addition.  We don't need to resolve
> that to a single bit until the VSCR register is read.

Aha, thanks for the input.

Long term, that might benefit from KVM specific code paths that don't
bother with this then.  However, for the first cut, it's simpler to
just keep the current representation, even though it's pretty odd for
KVM.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature