[Bug tree-optimization/17050] New: Performance pessimisation as compared to GCC 3.4

[Bug tree-optimization/17050] New: Performance pessimisation as compared to GCC 3.4

[coyote at coyotegulch dot com]

The benchmark program attached hereto is representative of a regression I've
seen in a couple of large programs. Compiled with GCC 3.4 and prior, this
program runs in about 12 seconds on my test systems; using GCC 3.5, it takes
more than 30 seconds to run.

-- 
           Summary: Performance pessimisation as compared to GCC 3.4
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: coyote at coyotegulch dot com
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: x86_64-unknown-linux
  GCC host triplet: x86_64-unknown-linux
GCC target triplet: x86_64-unknown-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] Performance pessimisation as compared to GCC 3.4

[coyote at coyotegulch dot com]

------- Additional Comments From coyote at coyotegulch dot com  2004-08-16 15:35 -------
Created an attachment (id=6942)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=6942&action=view)
testcase


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [3.5 Regression] Performance pessimisation as compared to GCC 3.4

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
            Summary|Performance pessimisation as|[3.5 Regression] Performance
                   |compared to GCC 3.4         |pessimisation as compared to
                   |                            |GCC 3.4
   Target Milestone|---                         |3.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [3.5 Regression] Performance pessimisation as compared to GCC 3.4

[law at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |law at redhat dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [3.5 Regression] Performance pessimisation as compared to GCC 3.4

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-09-03 22:08 -------
Note that one needs to use -lm -lrt to compile the test case. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [3.5 Regression] Performance pessimisation as compared to GCC 3.4

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-09-03 22:19 -------
Hmm and also -std=c99 
Hmm, I would certainly be interested in learning what kind of system you are testing on.  On 
my Opteron I get the following, 
 
GCC 3.3-hammer (SUSE9 system compiler): 
run time: 96.195581 
 
GCC 3.5-20040903 (CVS HEAD) 
run time: 110.353146 
 
So, there is a regression, but not from 12s to >30s.  What flags did you use?? 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [3.5 Regression] Performance pessimisation as compared to GCC 3.4

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-08 17:50 -------
Can you try again as IV-OPT should have improved the preformance? 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [4.0 Regression] Performance pessimisation as compared to GCC 3.4

[dberlin at gcc dot gnu dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2004-11-26 22:33 -------
On my pentium-m, 3.4 gets:

dberlin@linux:~> gcc -O3 -std=c99 coyotebench_mole.i -ffast-math -lrt -lm
dberlin@linux:~> time ./a.out
run time: 23.438668

real    0m24.187s
user    0m23.147s
sys     0m0.368s

and the tree-cleanup-branch gets
dberlin@linux:~/gcc-tcb/build/gcc> ./xgcc -B./ -std=c99 -O3 -ffast-math
~/coyotebench_mole.i  -lrt -lm
timdberlin@linux:~/gcc-tcb/build/gcc> time ./a.out
run time: 8.626844

real    0m8.919s
user    0m8.803s
sys     0m0.039s


I haven't redone mainline yet.
But it's good to know that TCB has this problem fixed, which means it will
likely be gone in 4.1 if not sooner.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [4.0 Regression] Performance pessimisation as compared to GCC 3.4

[pcarlini at suse dot de]

------- Additional Comments From pcarlini at suse dot de  2004-11-26 22:48 -------
Seems fixed in mainline too. On my P4-2400:

4.0.0 20041126
--------------
run time: 8.684465
8.810u 0.010s 0:08.86 99.5%     0+0k 0+0io 133pf+0w

3.4.3
-----
run time: 35.656412
36.200u 0.010s 0:36.37 99.5%    0+0k 0+0io 133pf+0w

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050

[Bug tree-optimization/17050] [4.0 Regression] Performance pessimisation as compared to GCC 3.4

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2004-11-26 22:51 -------
Fixed then.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17050