[Bug middle-end/22067] Inconsistent multiply by immediate

dje at gcc dot gnu dot org gcc-bugzilla@gcc.gnu.org
Sat Sep 17 04:28:00 GMT 2005


------- Additional Comments From dje at gcc dot gnu dot org  2005-09-17 04:28 -------
I think the patch in attachment 9747 that subtracts the cost of operand0 is the
correct solution.  One generally does want the cost of the operands included, so
rtx_costs is correct.  choose_mult_variant would consider too long a sequence if
it *added* the cost of operand0, so the correct solution appears to be
subtracting the cost.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dje at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22067



More information about the Gcc-bugs mailing list