[Bug rtl-optimization/17236] inefficient code for long long multiply on x86

["bonzini at gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #21 from bonzini at gnu dot org  2007-12-19 12:43 -------
Created an attachment (id=14793)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14793&action=view)
two hunks only from the previous patch

Indeed, if I only use the regclass.c and local-alloc.c hunks, I get only one
spill!

        pushl   %ebx
        movl    8(%esp), %edx
        movl    16(%esp), %eax
        movl    20(%esp), %ecx
        imull   %edx, %ecx
        movl    12(%esp), %ebx
        imull   %eax, %ebx
        addl    %ebx, %ecx
        mull    %edx
        leal    (%ecx,%edx), %edx
        popl    %ebx

and no reg-reg move.

With Uros' patch it does use memory imull's, but also adds reg-reg moves.  The
next step should be to figure out why (with my patch and without his) reload
feels the need to put the memory operands from insns 7 and 8 into regs.


-- 

bonzini at gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #14792|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17236