This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Big find_cross_jump performance regression.
- To: Brad Lucier <lucier at math dot purdue dot edu>
- Subject: Re: Big find_cross_jump performance regression.
- From: Michael Matz <matzmich at cs dot tu-berlin dot de>
- Date: Mon, 22 May 2000 19:50:22 +0200 (MET DST)
- cc: gcc at gcc dot gnu dot org, feeley at iro dot umontreal dot ca
Hi,
On Mon, 22 May 2000, Brad Lucier wrote:
> I did a clean build this morning and got the same thing (this is not
> profiled, so it's faster):
>
> jump : 726.39 (99%) usr 0.04 (15%) sys 726.90 (99%) wall
>
> This is on alphaev6-unknown-linux-gnu.
Must be something depending on the machine description. Could you try with
-Os? I get normal results. Even for -O2 with a profiled compiler:
% ./cc1 -O2 ./new_t-c-2.i
garbage collection : 1.32 ( 1%) usr 0.01 ( 1%) sys 1.33 ( 1%)
parser : 8.48 ( 9%) usr 0.22 (28%) sys 8.71 ( 9%)
varconst : 0.00 ( 0%) usr 0.01 ( 1%) sys 0.00 ( 0%)
integration : 0.00 ( 0%) usr 0.00 ( 0%) sys 0.00 ( 0%)
jump : 8.49 ( 9%) usr 0.20 (25%) sys 8.69 ( 9%)
CSE : 2.55 ( 3%) usr 0.02 ( 2%) sys 2.57 ( 3%)
global CSE : 29.07 (31%) usr 0.16 (20%) sys 29.23 (31%)
loop analysis : 0.17 ( 0%) usr 0.00 ( 0%) sys 0.17 ( 0%)
CSE 2 : 1.86 ( 2%) usr 0.00 ( 0%) sys 1.86 ( 2%)
flow analysis : 1.78 ( 2%) usr 0.02 ( 2%) sys 1.80 ( 2%)
combiner : 1.50 ( 2%) usr 0.00 ( 0%) sys 1.50 ( 2%)
if-conversion : 0.25 ( 0%) usr 0.03 ( 4%) sys 0.28 ( 0%)
regmove : 0.52 ( 1%) usr 0.00 ( 0%) sys 0.52 ( 1%)
local alloc : 1.77 ( 2%) usr 0.00 ( 0%) sys 1.77 ( 2%)
global alloc : 5.34 ( 6%) usr 0.03 ( 4%) sys 5.37 ( 6%)
reload CSE regs : 18.51 (20%) usr 0.01 ( 1%) sys 18.52 (20%)
flow 2 : 2.08 ( 2%) usr 0.01 ( 1%) sys 2.09 ( 2%)
if-conversion 2 : 0.08 ( 0%) usr 0.02 ( 2%) sys 0.10 ( 0%)
peephole 2 : 0.16 ( 0%) usr 0.00 ( 0%) sys 0.16 ( 0%)
scheduling 2 : 6.57 ( 7%) usr 0.04 ( 5%) sys 6.63 ( 7%)
shorten branches : 0.47 ( 1%) usr 0.00 ( 0%) sys 0.47 ( 1%)
reg stack : 0.00 ( 0%) usr 0.00 ( 0%) sys 0.00 ( 0%)
final : 0.58 ( 1%) usr 0.02 ( 2%) sys 0.60 ( 1%)
symout : 0.00 ( 0%) usr 0.00 ( 0%) sys 0.00 ( 0%)
rest of compilation : 0.77 ( 1%) usr 0.00 ( 0%) sys 0.77 ( 1%)
TOTAL : 92.35 0.80 93.78
In the last days I see no change to jump.c which could cause that (by
looking only short time at the diffs ;) ) Only the change from 2 to
BRANCH_COST (which is 5 for alpha) in the minimum argument. Hmm, no that
can't be the reason...
Ciao,
Michael.