https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121617
Bug ID: 121617
Summary: arithmetic result or 'fma' result is zero when
denormals are zero but result is not zero and normal
Product: gcc
Version: 13.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: madams846 at hotmail dot com
Target Milestone: ---
For types 'float' and 'double', basic arithmetic or intrinsic functions fmaf
and fma have a normal result, but return zero when compiled with a switch for
denormals are zero -mdaz-ftz or -ffast-math or -Ofast. The result would be
sub-normal but rounds up to normal minimum per rounding mode. For types 'long
double' and 'float128', basic arithmetic and functions 'fmal' and 'fmaq' do not
return zero. This bug arose when testing free-standing code for a fused
multiply addition. This bug depends on whether gcc switch for denormals are
zero should allow for rounding mode.
Test program 'bugdemo.c':
#include <stdio.h>
#include <math.h>
#include <float.h>
int main() {
float af= FLT_MIN, bf= 1.0F - FLT_EPSILON/2.0F, cf= 0.0F ;
printf("af bf cf %7.18e %7.18e %7.18e \n",af,bf,cf);
printf("fmaf(af,bf,cf) %7.18e \n", fmaf(af,bf,cf) );
printf("af*bf %7.18e \n\n", af*bf );
double a= DBL_MIN, b= 1.0 - DBL_EPSILON/2.0, c= 0.0 ;
printf("a b c %7.18e %7.18e %7.18e \n",a,b,c);
printf("fma(a,b,c) %7.18e \n", fma(a,b,c) );
printf("a*b %7.18e \n\n", a*b );
long double al= LDBL_MIN, bl= 1.0L - LDBL_EPSILON/2.0L, cl= 0.0L ;
printf("al bl cl %7.24Le %7.24Le %7.24Le \n",al,bl,cl);
printf("fmal(al,bl,cl) %7.24Le \n", fmal(al,bl,cl) );
printf("al*bl %7.24Le \n\n", al*bl );
return 0 ;
}
when compiled with 'gcc bugdemo.c -lm' returns:
af bf cf 1.175494350822287508e-38 9.999999403953552246e-01
0.000000000000000000e+00
fmaf(af,bf,cf) 1.175494350822287508e-38
af*bf 1.175494350822287508e-38
a b c 2.225073858507201383e-308 9.999999999999998890e-01
0.000000000000000000e+00
fma(a,b,c) 2.225073858507201383e-308
a*b 2.225073858507201383e-308
al bl cl 3.362103143112093506262678e-4932 9.999999999999999999457899e-01
0.000000000000000000000000e+00
fmal(al,bl,cl) 3.362103143112093506262678e-4932
al*bl 3.362103143112093506262678e-4932
but when compiled with 'gcc bugdemo.c -lm -mdaz-ftz' returns:
af bf cf 1.175494350822287508e-38 9.999999403953552246e-01
0.000000000000000000e+00
fmaf(af,bf,cf) 0.000000000000000000e+00
af*bf 0.000000000000000000e+00
a b c 2.225073858507201383e-308 9.999999999999998890e-01
0.000000000000000000e+00
fma(a,b,c) 0.000000000000000000e+00
a*b 0.000000000000000000e+00
al bl cl 3.362103143112093506262678e-4932 9.999999999999999999457899e-01
0.000000000000000000000000e+00
fmal(al,bl,cl) 3.362103143112093506262678e-4932
al*bl 3.362103143112093506262678e-4932
My system:
WSL 5.10.102.1
Dell PC 64-bit 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
Windows 11 Home 24H2 OS build 26100.4351
gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-linux-gnu/13/lto-wrapper
OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa
OFFLOAD_TARGET_DEFAULT=1
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu
13.3.0-6ubuntu2~24.04' --with-bugurl=file:///usr/share/doc/gcc-13/README.Bugs
--enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --prefix=/usr
--with-gcc-major-version-only --program-suffix=-13
--program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id
--libexecdir=/usr/libexec --without-included-gettext --enable-threads=posix
--libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu
--enable-libstdcxx-debug --enable-libstdcxx-time=yes
--with-default-libstdcxx-abi=new --enable-libstdcxx-backtrace
--enable-gnu-unique-object --disable-vtable-verify --enable-plugin
--enable-default-pie --with-system-zlib --enable-libphobos-checking=release
--with-target-system-zlib=auto --enable-objc-gc=auto --enable-multiarch
--disable-werror --enable-cet --with-arch-32=i686 --with-abi=m64
--with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic
--enable-offload-targets=nvptx-none=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-fG75Ri/gcc-13-13.3.0/debian/tmp-gcn/usr
--enable-offload-defaulted --without-cuda-driver --enable-checking=release
--build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
--with-build-config=bootstrap-lto-lean --enable-link-serialization=2
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 13.3.0 (Ubuntu 13.3.0-6ubuntu2~24.04)
