GBuella added a comment.
In https://reviews.llvm.org/D46683#1094662, @craig.topper wrote:
> What effect does this have?
Nothing important really, I just guessed it doesn't cost.
One contrived example I could come up with in 2 minutes:
#include <x86intrin.h>
void x(char *restrict a __attribute__((align_value(64))), char *restrict b,
const char *c)
{
_movdir64b(b, c);
for (int i = 0; i < 32; ++i)
a[i] = b[i];
}
The silly memcpy loop above becomes 2 `movaps` instructions, instead of two
`movups` instructions.
Repository:
rC Clang
https://reviews.llvm.org/D46683
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
