[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAH9TF6OC6Xf2VGSjpkDfVU12gRJfue1y-UDK=9rO6RnHtWn-fg@mail.gmail.com>
Date: Wed, 28 Aug 2024 18:31:42 +0200
From: Alex Rønne Petersen <alex@...xrp.com>
To: Alexander Monakov <amonakov@...ras.ru>
Cc: musl@...ts.openwall.com
Subject: Re: [PATCH] configure: prevent compilers from turning a * b +
c into fma(a, b, c)
On Wed, Aug 28, 2024 at 5:53 PM Alexander Monakov <amonakov@...ras.ru> wrote:
>
>
> On Wed, 28 Aug 2024, Alex Rønne Petersen wrote:
>
> > I've seen Clang do this for expressions in the fma() implementation itself,
> > which of course led to infinite recursion. This happened when targeting
> > arm-linux-musleabi with full soft float mode and -march=armv8-a. I imagine
> > it's possible for GCC to do similar silliness.
>
> musl passes -std=c99 to the compiler, and in all GCC releases so far* that
> disables FMA contraction (as opposed to -std=gnu99 or whichever -std=gnuXX
> is enabled by default, where unrestricted contraction is implicitly enabled,
> i.e. the non-standard and dangerous -ffp-contract=fast mode).
OK, that part can definitely be left out. I only added it for good
measure; I haven't actually seen the issue occur with GCC.
>
> Clang respects #pragma STDC FP_CONTRACT OFF, so that is available as
> a smaller hammer than disabling fma across the board. Breaking up
> contractable expression in fma*.c will work too.
That pragma indeed seems to work. I suppose we could just add it to
the fma*() implementations and call it a day? I'm unsure if there are
other functions that might be affected by this.
>
> [*] maybe modulo bugs in old releases where the backend doesn't respect
> -ffp-contract=off and which Glibc worked around with -mno-fused-madd.
>
> Alexander
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
