| Message ID | 20250425205618.360232-3-adhemerval.zanella@linaro.org |
|---|---|
| State | Superseded |
| 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 4d6192d2. > > 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_coshf.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/sysdeps/ieee754/flt-32/e_coshf.c b/sysdeps/ieee754/flt-32/e_coshf.c > index 5f6ff8c29f..382cd55cdf 100644 > --- a/sysdeps/ieee754/flt-32/e_coshf.c > +++ b/sysdeps/ieee754/flt-32/e_coshf.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/cosh/coshf.c, revision 5c58ea1). > +project (file src/binary32/cosh/coshf.c, revision de59ecfb). 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 > @@ -69,8 +69,8 @@ __ieee754_coshf (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); OK, fixes with (uint64_t) cast. commit de59ecfbe35b4e2dddbad2849f3cdde7835d5a48 (HEAD) Author: Paul Zimmermann <Paul.Zimmermann@inria.fr> Date: Thu Mar 13 17:06:24 2025 +0100 [coshf] fixed undefined behavior
diff --git a/sysdeps/ieee754/flt-32/e_coshf.c b/sysdeps/ieee754/flt-32/e_coshf.c index 5f6ff8c29f..382cd55cdf 100644 --- a/sysdeps/ieee754/flt-32/e_coshf.c +++ b/sysdeps/ieee754/flt-32/e_coshf.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/cosh/coshf.c, revision 5c58ea1). +project (file src/binary32/cosh/coshf.c, revision de59ecfb). Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal @@ -69,8 +69,8 @@ __ieee754_coshf (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);