[Bug c/17680] New: Gcc generates mull for a power of 2

["terpstra at ito dot tu-darmstadt dot de" <gcc-bugzilla at gcc dot gnu dot org>]

unsigned long long foo(unsigned long x)
{ return (unsigned long long)x * 16777216; }

compiled with gcc -O -S -o beh.s -c beh.s generates:
        pushl   %ebp
        movl    %esp, %ebp
        movl    $16777216, %eax
        mull    8(%ebp)
        popl    %ebp
        ret

Interestingly, when the code is:
unsigned long long foo(unsigned long x) {
        unsigned long long y = x;
        return y * 16777216;
}

gcc generates:
        pushl   %ebp
        movl    %esp, %ebp
        movl    8(%ebp), %eax
        movl    $0, %edx
        shldl   $24, %eax, %edx
        sall    $24, %eax
        popl    %ebp
        ret

Normally, I would just use a temporary like above.
However, in the context where this missed optimization occured there was a method:
foo(unsigned long a, unsigned long b)
{
 unsigned long long z = (unsigned long long)a * (unsigned long long)b;
 ...
}

This construction performs a single mul, whereas with the temporary approach it
does a full-scale 64bit multiplication. Really, I just want the best of both
worlds. ie: that I leave the code as above (and thus use only one mul), but when
the code foo(x, 16777216) appears, use shifts as expected.

-- 
           Summary: Gcc generates mull for a power of 2
           Product: gcc
           Version: 3.4.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: terpstra at ito dot tu-darmstadt dot de
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17680