[Bug c/22072] New: bizarre code for int*int/2

["felix-gcc at fefe dot de" <gcc-bugzilla at gcc dot gnu dot org>]

int triangle(int a,int b) { int c; c=a*b/2; return c; }
emits this very bizarre code (at -O, -O2):
  mov a,%edx
  mov b,%eax
  imull %edx,%eax
  movl %eax,%edx
  shrl $31,%edx
  addl %edx,%eax
  sarl %eax
  ret
Why are the two instructions after the imull emitted?  Shouldn't this become
simply imull and sarl?  This code extracts the most significant bit of a*b and
adds it to a*b, then shifting the result right.  It almost looks as if trying to
round or something?  There is probably something obvious I'm overlooking here.
The analog code is generated for ppc, x86_64, sparc and mips.  Please explain.
I also tried with -Os, and the code becomes cltd (sign-extend 32 to 64 bits)
plus idivl with 2.  Could it be that the peephole optimizer converts the idivl
to a shift but forgets to remove the sign extend code?

-- 
           Summary: bizarre code for int*int/2
           Product: gcc
           Version: 3.4.4
            Status: UNCONFIRMED
          Severity: minor
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: felix-gcc at fefe dot de
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22072