Closed GoogleCodeExporter closed 9 years ago
Better reproducer that survives -O3:
#include <emmintrin.h>
int main() {
__m128i d;
__m128i * volatile p = &d;
__m128i dst_pixel = _mm_load_si128(p);
__m128i z =_mm_srli_epi16(dst_pixel, 8);
return z[0];
}
Original comment by euge...@google.com
on 20 Feb 2014 at 9:12
MemorySanitizer instrumentation pass has this magical function
maybeHandleSimpleNomemIntrinsic that takes care of all the vector arithmetic
(in a very approximate but safe way), but does not handle certain shifts
(namely, vector-by-scalar).
Original comment by euge...@google.com
on 20 Feb 2014 at 9:18
r202712, r202713 add support for all bitshift X86 intrinsics.
Original comment by euge...@google.com
on 3 Mar 2014 at 2:45
Adding Project:MemorySanitizer as part of GitHub migration.
Original comment by gli...@google.com
on 30 Jul 2015 at 9:22
Original issue reported on code.google.com by
earth...@google.com
on 19 Feb 2014 at 4:12