This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug target/72805] AVX512: invalid code generation involving masks


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72805

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2016-08-04
     Ever confirmed|0                           |1

--- Comment #1 from Uroš Bizjak <ubizjak at gmail dot com> ---
Indeed, there are a couple of intrinsics in avx512fintrin.h defined under

#ifndef __OPTIMIZE__

with a wrong cast to (__mmask8) instead of (__mmask16), e.g.:

 #define _mm512_cmp_epi32_mask(X, Y, P)                                 \
-  ((__mmask8) __builtin_ia32_cmpd512_mask ((__v16si)(__m512i)(X),      \
-                                          (__v16si)(__m512i)(Y), (int)(P),\
-                                          (__mmask16)-1))
+  ((__mmask16) __builtin_ia32_cmpd512_mask ((__v16si)(__m512i)(X),     \
+                                           (__v16si)(__m512i)(Y), (int)(P), \
+                                           (__mmask16)-1))

I have an obvious patch.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]