Your message dated Tue, 11 Aug 2015 11:37:16 +0200
with message-id <55c9c24c.4090...@debian.org>
and subject line fixed
has caused the Debian Bug report #783543,
regarding miscompile of _mm256_permutevar8x32_epi32
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact ow...@bugs.debian.org
immediately.)


-- 
783543: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=783543
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems
--- Begin Message ---
Package: llvm-toolchain-3.6
Version: 1:3.6-2
Severity: serious

clang 3.6 in debian miscompiles _mm256_permutevar8x32_epi32
it swaps the two first arguments:


$ apt-cache policy clang-3.6
clang-3.6:
  Installed: 1:3.6-2
  Candidate: 1:3.6-2
  Version table:
 *** 1:3.6-2 0
        500 http://ftp.de.debian.org/debian/ unstable/main amd64

$ cat test.c
#include <immintrin.h>

__m256i fun (__m256i a, __m256i o) {
    return _mm256_permutevar8x32_epi32(a, o);
}


$ clang-3.6 test.c -mavx2 -c -O2
$ objdump -d test.o

test.o:     file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <fun>:
   0:   c4 e2 7d 36 c1          vpermd %ymm1,%ymm0,%ymm0
   5:   c3                      retq


the correct result is:
vpermd %ymm0,%ymm1,%ymm0

clang 3.5 seems to produce the right result so this is a 3.6 regression

--- End Message ---
--- Begin Message ---
fixed in 1:3.6.1-1

--- End Message ---

Reply via email to