https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90262
Bug ID: 90262
Summary: Inline small constant memmoves
Product: gcc
Version: 10.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: wilco at gcc dot gnu.org
Target Milestone: ---
GCC does not inline fixed-size memmoves. However memmove can be as easily
inlined as memcpy. The existing memcpy infrastructure could be reused/expanded
for this - all loads would be emitted first, followed by the stores. Large
memmoves could be inlined on targets with vector registers. Targets without
vector registers could emit an overlap check and use inlined memcpy for the no
overlap case.
void copy(int *p, int *q)
{
__builtin_memcpy(p, q, 24);
}
void move(int *p, int *q)
{
__builtin_memmove(p, q, 24);
}
copy:
ldp x2, x3, [x1]
stp x2, x3, [x0]
ldr x1, [x1, 16]
str x1, [x0, 16]
ret
move:
mov x2, 24
b memmove
The memmove could be expanded using the same number of registers:
ldp x2, x3, [x1]
ldr x1, [x1, 16]
stp x2, x3, [x0]
str x1, [x0, 16]
