------- Comment #5 from ubizjak at gmail dot com  2007-02-24 19:09 -------
(In reply to comment #2)
> I have verified that revision 119252:
> 
> http://gcc.gnu.org/ml/gcc-cvs/2006-11/msg00907.html
> breaks -mtune=nocona. Jan, can you take a look? Thanks.

Something is still wrong with unrolled_loop stringop handling.

Patch at http://gcc.gnu.org/ml/gcc-patches/2007-02/msg01937.html and the
following patch that disables unrolled loop stringops (for nocona) bootstraps
OK for -mtune=nocona. This applies to -march=athlon64 as well.

Index: i386.c
===================================================================
--- i386.c      (revision 122286)
+++ i386.c      (working copy)
@@ -789,10 +789,10 @@
   COSTS_N_INSNS (44),                  /* cost of FSQRT instruction.  */
   {{libcall, {{12, loop_1_byte}, {-1, rep_prefix_4_byte}}},
    {libcall, {{32, loop}, {20000, rep_prefix_8_byte},
-             {100000, unrolled_loop}, {-1, libcall}}}},
+             /*{100000, unrolled_loop},*/ {-1, libcall}}}},
   {{libcall, {{6, loop_1_byte}, {48, loop}, {20480, rep_prefix_4_byte},
    {-1, libcall}}},
-   {libcall, {{24, loop}, {64, unrolled_loop},
+   {libcall, {{24, loop}, /*{64, unrolled_loop},*/
              {8192, rep_prefix_8_byte}, {-1, libcall}}}}
 };



-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ubizjak at gmail dot com
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2007-02-24 19:09:51
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30770

Reply via email to