| Message ID | 20250425205618.360232-5-adhemerval.zanella@linaro.org |
|---|---|
| State | Accepted |
| Commit | c8775c0423f594543994eea268065e71b76effb0 |
| Headers | show |
| Series | Fix UB in math implementations | expand |
On 4/25/25 4:54 PM, Adhemerval Zanella wrote: > The left shift overflows for 'int', use uint64_t instead. It syncs > with CORE-MATH commit bbfabd99. > > Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. LGTM. Reviewed-by: Carlos O'Donell <carlos@redhat.com> > --- > sysdeps/ieee754/flt-32/e_sinhf.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/sysdeps/ieee754/flt-32/e_sinhf.c b/sysdeps/ieee754/flt-32/e_sinhf.c > index 754b84abe0..6c8c1db404 100644 > --- a/sysdeps/ieee754/flt-32/e_sinhf.c > +++ b/sysdeps/ieee754/flt-32/e_sinhf.c > @@ -3,7 +3,7 @@ > Copyright (c) 2022-2024 Alexei Sibidanov. > > The original version of this file was copied from the CORE-MATH > -project (file src/binary32/sinh/sinhf.c, revision 572ecec). > +project (file src/binary32/sinh/sinhf.c, revision bbfabd99). OK. > > Permission is hereby granted, free of charge, to any person obtaining a copy > of this software and associated documentation files (the "Software"), to deal > @@ -83,8 +83,8 @@ __ieee754_sinhf (float x) > double h2 = h * h; > int64_t jp = asuint64 (ia + 0x1.8p52); > int64_t jm = -jp; > - double sp = asdouble (TB[jp & 31] + ((jp >> 5) << 52)); > - double sm = asdouble (TB[jm & 31] + ((jm >> 5) << 52)); > + double sp = asdouble (TB[jp & 31] + ((uint64_t)(jp >> 5) << 52)); > + double sm = asdouble (TB[jm & 31] + ((uint64_t)(jm >> 5) << 52)); OK. Both cast to uint64_t. > double te = C[0] + h2 * C[2]; > double to = (C[1] + h2 * C[3]); > double rp = sp * (te + h * to); OK. commit bbfabd993a71b049c210b0febfd06d18369fadc1 Author: Paul Zimmermann <Paul.Zimmermann@inria.fr> Date: Mon Mar 17 12:29:20 2025 +0100 various fixes to avoid undefined behavior for binary32 functions (contributed by Ganesh Ajjanagadde <gajjanag@alum.mit.edu> see https://sympa.inria.fr/sympa/arc/core-math/2025-03/msg00007.html)
diff --git a/sysdeps/ieee754/flt-32/e_sinhf.c b/sysdeps/ieee754/flt-32/e_sinhf.c index 754b84abe0..6c8c1db404 100644 --- a/sysdeps/ieee754/flt-32/e_sinhf.c +++ b/sysdeps/ieee754/flt-32/e_sinhf.c @@ -3,7 +3,7 @@ Copyright (c) 2022-2024 Alexei Sibidanov. The original version of this file was copied from the CORE-MATH -project (file src/binary32/sinh/sinhf.c, revision 572ecec). +project (file src/binary32/sinh/sinhf.c, revision bbfabd99). Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal @@ -83,8 +83,8 @@ __ieee754_sinhf (float x) double h2 = h * h; int64_t jp = asuint64 (ia + 0x1.8p52); int64_t jm = -jp; - double sp = asdouble (TB[jp & 31] + ((jp >> 5) << 52)); - double sm = asdouble (TB[jm & 31] + ((jm >> 5) << 52)); + double sp = asdouble (TB[jp & 31] + ((uint64_t)(jp >> 5) << 52)); + double sm = asdouble (TB[jm & 31] + ((uint64_t)(jm >> 5) << 52)); double te = C[0] + h2 * C[2]; double to = (C[1] + h2 * C[3]); double rp = sp * (te + h * to);