This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Some numbers on gcc compiling itself
- From: Steven Bosscher <stevenb at suse dot de>
- To: gcc at gcc dot gnu dot org
- Date: Sat, 30 Oct 2004 15:17:47 +0200
- Subject: Some numbers on gcc compiling itself
- Organization: SUSE Labs
Hi,
These numbers are user times in seconds of GCC compiling its own
preprocessed sources of gcc mainline 20040914) on i686.
The numbers are averages of three runs, and I have only picked those
files for which the compile time is relatively large. Here is the top
10 slowdowns and the 2 largest speedups:
file 3.3.3 (hammer) cvs-20041030 slowdown %
insn-recog.i 28.62 41.57 45.25
insn-attrtab.i 27.58 37.69 36.66
parse.i 10.94 12.55 14.72
combine.i 6.81 8.02 17.77
gcc.i 4.2 6.46 53.81
reload.i 5.38 6.88 27.88
reload1.i 4.74 6.38 34.6
loop.i 4.99 5.76 15.43
genautomata.i 4.06 5.23 28.82
cse.i 3.92 4.77 21.68
insn-emit.i 6.9 5.92 -14.2
c-common.i 8.56 5.73 -33.06
The total compile time for all 643 .i files:
3.3.3 (hammer) cvs-20041030 slowdown %
TOTAL 417.38 484.75 16.14
The total compile time for all 643 .i files, *excluding* the top 10
slowdowns shown above:
3.3.3 (hammer) cvs-20041030 slowdown %
TOTAL 316.14 349.44 10.53
The "3.3.3 (hammer)" compiler is the SLES9 system compiler, which is
build with a profiledbootstrap, so part of the slowdown can probably
be explained with that. But it's still pretty depressing to see all
these slowdowns :-/
Gr.
Steven