Hi all, and happy 4th of July.
On Saturday 04 July 2009 12:06:26 Shlomi Fish wrote:
> Hi all!
>
> Here's what new on the version control repository's trunk since the 2.32.1
> release:
>
> 1. I did some work on the Games-Solitaire-Verify module, mostly legal
> cleanups and released a new version - 0.08.
>
> 2. I converted the soft_thread struct to use a C union (see:
> http://en.wikipedia.org/wiki/Union_%28computer_science%29 ) for the scan
> type- specific data to save some space.
>
> This change proved to be relatively tricky and took me a long time to get
> all tests to pass with it.
>
> 3. I added support for the Sun Studio C/C++ compiler to the Makefile.gnu.
> The speed it runs the benchmarks (after compiled with its -fast flag) are
> worse than gcc.
>
> 4. I added more links to http://fc-solve.berlios.de/docs/ .
>
> 5. As I said, I updated the Architecture document.
>
> 6. I added a libfreecell-solver.pc pkg-config (see
> http://pkg-config.freedesktop.org/ ) file to the installation process.
>
> 7. The time-threads-num.bash script was enhanced in several ways.
>
> 8. Added a --minsize option to Tatzer to enable the CMake MinSizeRel. It's
> still a bit larger than the Makefile.gnu's generated binaries due to the
> use of -fPIC.
>
> 9. Added some tests for split-cmd-line.c and even fixed a bug or two there.
>
> 10. I finally decided to get to enhancing the atomic moves preset. I used
> the analyse.pl script in the meta-moves directory and found the boards
> which proved to be intractable for all the individual scans. Then I found
> scans that solved them efficiently.
>
> I uncovered a bug in the presets' generation code where I blacklisted some
> scans globally instead of for each preset-space (atomic moves, meta-moves,
> etc.). This was fixed.
>
> I also discovered that the --scans-synergy option caused some boards to
> become intractable, and so had to time everything with "--scans-synergy
> none".
>
> The result of all of it was the "toons-for-twenty-somethings" atomic moves
> preset that is now included there.
>
I forgot to note that toons-for-twenty-somethings solves the first 22,000
deals in 1060.46217894554 seconds (using the non-threaded solver). This is
much worse than "-l gi" which solves all 32,000 deals in 113 seconds.
Regards,
Shlomi Fish
--
-----------------------------------------------------------------
Shlomi Fish http://www.shlomifish.org/
"Humanity" - Parody of Modern Life - http://xrl.us/bkeut
God gave us two eyes and ten fingers so we will type five times as much as we
read.
Received on Sat Jul 04 2009 - 02:21:18 IDT