forked from OSchip/llvm-project
[libc] Align to 32B instead of 16B for optimized memcmp
This commit is contained in:
parent
d52f506192
commit
791d88f35f
|
@ -39,7 +39,7 @@ static int memcmp_impl(const char *lhs, const char *rhs, size_t count) {
|
|||
return ThreeWayCompare<HeadTail<_32>>(lhs, rhs, count);
|
||||
if (count <= 128)
|
||||
return ThreeWayCompare<HeadTail<_64>>(lhs, rhs, count);
|
||||
return ThreeWayCompare<Align<_16>::Then<Loop<_32>>>(lhs, rhs, count);
|
||||
return ThreeWayCompare<Align<_32>::Then<Loop<_32>>>(lhs, rhs, count);
|
||||
}
|
||||
|
||||
LLVM_LIBC_FUNCTION(int, memcmp,
|
||||
|
|
Loading…
Reference in New Issue