This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Stepanov test (still) failing ?!?
- To: gcc at gcc dot gnu dot org, mark at codesourcery dot com
- Subject: Re: Stepanov test (still) failing ?!?
- From: Paolo Carlini <pcarlini at unitus dot it>
- Date: Mon, 11 Jun 2001 22:58:55 +0200
- Organization: Universita' della Tuscia
- Reply-To: pcarlini at unitus dot it
Hi,
FWIW eventually I ran the Stepanov test on my PII/400 system
(Gcc-20010611). This is a typical (reassuring!) summary (-O2):
test absolute additions ratio with
number time per second test0
0 0.39sec 128.21M 1.00
1 0.43sec 116.28M 1.10
2 0.41sec 121.95M 1.05
3 0.38sec 131.58M 0.97
4 0.40sec 125.00M 1.03
5 0.39sec 128.21M 1.00
6 0.38sec 131.58M 0.97
7 0.40sec 125.00M 1.03
8 0.38sec 131.58M 0.97
9 0.39sec 128.21M 1.00
10 0.39sec 128.21M 1.00
11 0.53sec 94.34M 1.36
12 0.53sec 94.34M 1.36
mean: 0.41sec 121.17M 1.06
Total absolute time: 5.40 sec
Abstraction Penalty: 1.06
As already mentioned by Mark Mitchell the overall AP is now similar to
that of Gcc2.95.2 (and much better than before! Thansk!) The individual
tests however show an important difference: 1-10 are almost perfect
whereas tests 11-12 ("use wrapped pointers wrapped in a reverse-iterator
adaptor wrapped in a reverse-iterator adaptor") ran much slower and are
alone responsible of the still perfectible score.
For comparison the following is a typical Gcc2.95.2 summary:
test absolute additions ratio with
number time per second test0
0 0.40sec 125.00M 1.00
1 0.41sec 121.95M 1.02
2 0.43sec 116.28M 1.07
3 0.41sec 121.95M 1.02
4 0.42sec 119.05M 1.05
5 0.42sec 119.05M 1.05
6 0.42sec 119.05M 1.05
7 0.42sec 119.05M 1.05
8 0.42sec 119.05M 1.05
9 0.42sec 119.05M 1.05
10 0.42sec 119.05M 1.05
11 0.42sec 119.05M 1.05
12 0.41sec 121.95M 1.02
mean: 0.42sec 119.94M 1.04
Total absolute time: 5.42 sec
Abstraction Penalty: 1.04
Cheers,
Paolo Carlini.