https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116615
Xi Ruoyao <xry111 at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |xry111 at gcc dot gnu.org
--- Comment #3 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
FYI on LoongArch it's claimed LOGICAL_OP_NON_SHORT_CIRCUIT=0 mostly helps FP
benchmarks, something like
/* { dg-options "-O2 -ffast-math -fdump-tree-gimple" } */
int
short_circuit (float *a)
{
float t1x = a[0];
float t2x = a[1];
float t1y = a[2];
float t2y = a[3];
float t1z = a[4];
float t2z = a[5];
if (t1x > t2y || t2x < t1y || t1x > t2z || t2x < t1z || t1y > t2z || t2y <
t1z)
return 0;
return 1;
}
See r14-8446. Maybe because FP comparison is more expensive.
I'm really not a fan of the change because without -ffast-math (which should
really not be used for 99.9% real-life code) the FP comparision isn't allowed
to be short-circuited anyway. Generally I'm much frustrated when all the
people seem using -ffast-math to bench things and justify changes.
