https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71847
Bug ID: 71847
Summary: powerpc64le: Potential rlwinm optimisation
Product: gcc
Version: 7.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: anton at samba dot org
Target Milestone: ---
The following code:
unsigned long foo(unsigned long x)
{
unsigned long y = (x & 0xffffffffUL);
return y | (y << 32);
}
Results in 3 instructions on gcc:
foo:
rldicl 3,3,0,32
sldi 9,3,32
or 3,9,3
blr
Clang does it in 1, knowing that rotate32 starts with two 32 bit copies of the
data:
foo:
rlwinm 3, 3, 0, 1, 0
blr
