This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Bug in ra-colorize.c:merge_moves?
On Thu, Feb 27, 2003 at 09:32:27PM +0100, Michael Matz wrote:
> Hi Christian,
>
> On Thu, 27 Feb 2003, Michael Matz wrote:
>
> > > Finally let me say a few words about my rational:
> >
> > All points correct. I guess, I will do something similar for the
> > new-regalloc-branch, and remove the non-detection of copy-insns involving
> > SPILL_SLOT_P's.
>
> Could you try the patch in
> http://gcc.gnu.org/ml/gcc-patches/2003-02/msg02347.html to see if this
> also fixes your performance issues?
I ran a somewhat longer test, the test is to compile gcc with the
tested compiler as hostgcc. Additional CFLAGS are -O2 and -ftime-report.
The time-report outputs are then added up by a per Skript.
The current new-regalloc-branch (with the patch included) gave these
top numbers (times in milliseconds):
user sys wall
parser: 48150 5100 52010
CSE: 80310 220 82220
global CSE: 81410 41150 123600
local alloc: 175740 1470 175030
TOTAL: 865730 63910 929210
This means that the performance issues of newra aren't solved yet but
here are the numbers that I got without the patch:
user sys wall
parser: 49670 5040 54920
CSE: 79560 250 76490
global CSE: 81760 41950 123330
local alloc: 598860 3230 601360
TOTAL: 1288580 66380 1351710
So this is definitely a significant improvement wrt. compile time speed!
Thanks.
regards Christian
--
THAT'S ALL FOLKS!