[RFC PATCH] build: reduce use of AVX compiler flags
Varghese, Vipin
Vipin.Varghese at amd.com
Thu Apr 10 05:49:21 CEST 2025
[Public]
Snipped
> >
> > > > > > When doing a build for a target that already has the
> > > > > > instruction sets for AVX2/AVX512 enabled, skip emitting the
> > > > > > AVX compiler flags, or the
> > > > > > skylake-avx512 '-march' flags, as they are unnecessary.
> > > > > > Instead, when the default flags produce the desired output,
> > > > > > just use them
> > > > > unmodified.
> > > > > >
> > > > > > Depends-on: series-34915 ("remove component-specific logic for
> > > > > > AVX
> > > > > builds")
> > > > > >
> > > > > > Signed-off-by: Bruce Richardson <bruce.richardson at intel.com>
> >
> > Hi Bruce,
> >
> > From AMD end we have tested with gcc 12, and 13.
> > We tried cross checking with and without patch for libraries such as `librte_net_ice
> and librte_net_i40e`.
> >
> > We did not find much changes in the assembly code generated.
> > I hope this is expected? Can you please share any pointers?\
> >
> Yes, it's expected. This is a simplification/consolidation, not an attempt to optimize in
> any way.
Thank you Bruce for confirming, we are going ahead assuming this is correct
.we will run for any performance regression at our end and share tested or +1 soon.
>
> /Bruce
More information about the dev
mailing list