bunch of small speed patches
Geoffrey Keating
geoffk@apple.com
Fri Sep 10 22:50:00 GMT 2004
On 10/09/2004, at 2:34 PM, Kaveh R. Ghazi wrote:
>
>> * gcc/config/rs6000/rs6000.c (print_operand): Use fputs
>> instead of fprintf for register names.
>>
>> Index: gcc/config/rs6000/rs6000.c
>> ===================================================================
>> RCS file: /cvs/gcc/gcc/gcc/config/rs6000/rs6000.c,v
>> retrieving revision 1.706
>> diff -u -p -u -p -r1.706 rs6000.c
>> --- gcc/config/rs6000/rs6000.c 8 Sep 2004 23:19:58 -0000 1.706
>> +++ gcc/config/rs6000/rs6000.c 10 Sep 2004 04:58:15 -0000
>> @@ -10062,7 +10062,7 @@ print_operand (FILE *file, rtx x, int co
>> /* Write second word of DImode or DFmode reference. Works on
>> register
>> or non-indexed memory only. */
>> if (GET_CODE (x) == REG)
>> - fprintf (file, "%s", reg_names[REGNO (x) + 1]);
>> + fputs (reg_names[REGNO (x) + 1], file);
>> else if (GET_CODE (x) == MEM)
>> {
>> /* Handle possible auto-increment. Since it is pre-increment
>> and
>
> IIRC, gcc has automatically done this transformation since at least
> version 3.0.
That's what I thought, but the profile showed a bunch of fprintf calls
coming from that location, so...
It may be that the transformation is broken. Is there a test for it?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2408 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20040910/083f38d2/attachment.p7s>
More information about the Gcc-patches
mailing list