https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117323
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target|x86_64-*-* i?86-*-* |
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Full testcase:
```
#include <algorithm>
void
foo (short* p1, short* p2, char* __restrict p3, int n)
{
for (int i = 0; i != n; i++)
{
short sum0 = p1[i];
short sum1 = p2[i];
sum0 = std::max<int>(0, std::min<int>(127, sum0));
sum1 = std::max<int>(0, std::min<int>(127, sum1));
p3[i] = static_cast<char>(sum0 * sum1 / 128);
}
}
void
foo1 (short* p1, short* p2, char* __restrict p3, int n)
{
for (int i = 0; i != n; i++)
{
short sum0 = p1[i];
short sum1 = p2[i];
sum0 = std::max<int>(std::min<int>(127, sum0), 0);
sum1 = std::max<int>(std::min<int>(127, sum1), 0);
p3[i] = static_cast<char>(sum0 * sum1 / 128);
}
}
```
foo1 works for x86_64 but foo fails.
foo works for aarch64 but foo1 fails.
Due to the way gimplification works in the opposite order for function
arguments.
