Could you describe what your benchmark involves?
i.e. which set of games, and some description of the
solutions obtained?
From: mailto:fc-solve-discuss_at_yahoogroups.com
Sent: Monday, June 01, 2015 3:19 AM
To: Freecell Solving Discussions
Subject: 1.47% Speed Improvement in the Benchmarks in the git HEAD
Hi all,
lately I've been doing a lot of cleanups and optimisations in the Freecell
Solver code, such as: adding const annotations, making functions inline,
removing dead code, eliminating malloc()ed pointers, and converting to stack
initialisers. Yesterday night I've done a new benchmark using the
GCC profile-guided optimisation scheme , and here are the new results:
*
https://github.com/shlomif/fc-solve/blob/master/fc-solve/benchmarks/shlomif-core-i3-desktop-machine.txt#L425
Solved in 9.11893510818481 seconds instead of 9.25326490402222 seconds:
$ \
perl ~/bin/improvement-percent from 9.25326490402222 to 9.11893510818481
1.47308643217388%
So seems like it was worth it.
In the process I discovered that GCC has a flag called “-Ofast” which aims to
improve on -O3 , while possibly breaking some standards-compliant code. It
seems to build Freecell Solver fine though , but I wasn't able to use it to
get a lower run time with it than I did with -O3 .
One thing I'd like to try is to add a compile time flag to exclude the
patsolve code which isn't utilised by the benchmark and benchmark using it.
Bye all!
— Shlomi Fish
--
----------------------------------------------------------
Shlomi Fish http://www.shlomifish.org/
Parody of "The Fountainhead" - http://shlom.in/towtf
Chuck Norris made the baby Jesus stop crying.
— http://www.shlomifish.org/humour/bits/facts/Chuck-Norris/
Please reply to list if it's a mailing list post - http://shlom.in/reply .
Received on Mon Jun 01 2015 - 06:20:08 IDT