[llvm-bugs] [Bug 42050] New: gvn produces wrong code with "-loop-rotate -load-store-vectorizer -instcombine -gvn -sink -reg2mem -instcombine -gvn"

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=42050

            Bug ID: 42050
           Summary: gvn produces wrong code with "-loop-rotate
                    -load-store-vectorizer -instcombine -gvn -sink
                    -reg2mem -instcombine -gvn"
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: cszide at 163.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 22037
  --> https://bugs.llvm.org/attachment.cgi?id=22037&action=edit
.bc file of the source code

$clang -v
clang version 9.0.0 (trunk 355281)
Target: x86_64-unknown-linux-gnu
Thread model: posix
InstalledDir: /home/jack-zhou/Documents/llvm/llvm_truck/llvm/build4/bin
Found candidate GCC installation: /usr/lib/gcc/i686-linux-gnu/8
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/5
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/5.5.0
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/6
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/6.5.0
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/7
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/7.3.0
Found candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/8
Selected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/7.3.0
Candidate multilib: .;@m64
Candidate multilib: 32;@m32
Candidate multilib: x32;@mx32
Selected multilib: .;@m64

$gcc small.c -o small0.out && ./small0.out
0

$clang -O3 -c -emit-llvm  -mllvm -disable-llvm-optzns small.c

$clang small.bc -o small1.out && ./small1.out
0

$opt -loop-rotate -load-store-vectorizer -instcombine -gvn -sink -reg2mem
-instcombine -gvn small.bc -o small-opt.bc

$clang small-opt.bc -o small2.out && ./small2.out
1

without gvn, the output is correct.
$opt -loop-rotate -load-store-vectorizer -instcombine -gvn -sink -reg2mem
-instcombine  small.bc -o small-opt2.bc

$clang small-opt2.bc -o small3.out && ./small3.out
0

However, when I delete any pass before gvn, the output also is correct. For
example, delete loop-rotate pass,

$opt  -load-store-vectorizer -instcombine -gvn -sink -reg2mem -instcombine 
small.bc -o small-opt3.bc

$clang small-opt3.bc -o small4.out && ./small4.out
0

---------------------------------------
#include<stdio.h>
struct {
  unsigned char a;
} b;
int c;
int d() {
  short int ***e[9];
  short int ****f[][9] = {&e[2], &e[2], &e[2], &e[2], &e[2],
                        &e[2], &e[2], &e[2], &e[2]};
  unsigned char *g = &b.a;
  for (c = 7; c; --c)
    *g = f[0][1] != &e[2];
  return 0;
}
void main() {
  d();
  printf("%d\n",b.a);
}

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20190529/609fdd40/attachment-0001.html>