Faile 0.6 for Win32 on Pentium Pro, PII or PIII

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 17 June 2000 03:08:31:
Look here:
ftp://38.168.214.175/pub/faile/
In particular...
Binary only:
ftp://38.168.214.175/pub/faile/failef.exe
Opening book:
ftp://38.168.214.175/pub/faile/FAILE.OPN
Book, binary from VC++ and GCC, source build scripts, etc.:
ftp://38.168.214.175/pub/faile/FAILE.ZIP
It's only a tiny bit faster than the stock version (1 or 2%).
I did make a lot of changes for safety (initializations, etc.) which probably accounts for how tiny the gain is.
If you don't have one of the CPU's in the subject line, it probably won't be faster at all.


My FTP site

[CL]

Geschrieben von: / Posted by: CL at 17 June 2000 11:50:34:
Als Antwort auf: / As an answer to: Faile 0.6 for Win32 on Pentium Pro, PII or PIII geschrieben von: / posted by: Dann Corbit at 17 June 2000 03:08:31:

Look here:
ftp://38.168.214.175/pub/faile/
In particular...
Binary only:
ftp://38.168.214.175/pub/faile/failef.exe
Opening book:
ftp://38.168.214.175/pub/faile/FAILE.OPN
Book, binary from VC++ and GCC, source build scripts, etc.:
ftp://38.168.214.175/pub/faile/FAILE.ZIP
It's only a tiny bit faster than the stock version (1 or 2%).
I did make a lot of changes for safety (initializations, etc.) which probably accounts for how tiny the gain is.
If you don't have one of the CPU's in the subject line, it probably won't be faster at all.

May be there will be a special version for fritz very soon, too.

[Pete Galati]

Geschrieben von: / Posted by: Pete Galati at 17 June 2000 17:21:57:
Als Antwort auf: / As an answer to: Re: Faile 0.6 for Win32 on Pentium Pro, PII or PIII geschrieben von: / posted by: CL at 17 June 2000 11:50:34:

Look here:
ftp://38.168.214.175/pub/faile/
In particular...
Binary only:
ftp://38.168.214.175/pub/faile/failef.exe
Opening book:
ftp://38.168.214.175/pub/faile/FAILE.OPN
Book, binary from VC++ and GCC, source build scripts, etc.:
ftp://38.168.214.175/pub/faile/FAILE.ZIP
It's only a tiny bit faster than the stock version (1 or 2%).
I did make a lot of changes for safety (initializations, etc.) which probably accounts for how tiny the gain is.
If you don't have one of the CPU's in the subject line, it probably won't be faster at all.

May be there will be a special version for fritz very soon, too.

That would be a good program for Chessbase to offer as an engine. It would be a rather plesant playing strength, compared to playing Fritz itself which is a rather brutal heavy handed program to play against. faile.eng would be good.
Pete

[Gian-Carlo Pascutto]

Geschrieben von: / Posted by: Gian-Carlo Pascutto at 17 June 2000 21:45:28:
Als Antwort auf: / As an answer to: Re: Faile 0.6 for Win32 on Pentium Pro, PII or PIII geschrieben von: / posted by: Pete Galati at 17 June 2000 17:21:57:

May be there will be a special version for fritz very soon, too.

That would be a good program for Chessbase to offer as an engine. It would be
a rather plesant playing strength, compared to playing Fritz itself which is a
rather brutal heavy handed program to play against. faile.eng would be good.
Pete

Might as well make one for Sjeng too. It shares the WinBoard code with
Faile.
I sure wouldn't complain...I can't seem to get the adapter working
decently.
--
GCP

[Pete Galati]

Geschrieben von: / Posted by: Pete Galati at 17 June 2000 22:20:06:
Als Antwort auf: / As an answer to: Re: Faile 0.6 for Win32 on Pentium Pro, PII or PIII geschrieben von: / posted by: Gian-Carlo Pascutto at 17 June 2000 21:45:28:

May be there will be a special version for fritz very soon, too.

That would be a good program for Chessbase to offer as an engine. It would be
a rather plesant playing strength, compared to playing Fritz itself which is a
rather brutal heavy handed program to play against. faile.eng would be good.
Pete

Might as well make one for Sjeng too. It shares the WinBoard code with
Faile.
I sure wouldn't complain...I can't seem to get the adapter working
decently.
--
GCP

My limited experience with their adapter is that Winboard programs using the adapter don't seem like they work as well as the Crafty and Comet .eng that Chessbase offers for download. But I may be mistaken. It just seems like they're working better in Winboard than in a Chessbase interface with the adapter, but Crafty & Comet .eng seem to work real good.
Pete

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 18 June 2000 09:37:57:
Als Antwort auf: / As an answer to: Faile 0.6 for Win32 on Pentium Pro, PII or PIII geschrieben von: / posted by: Dann Corbit at 17 June 2000 03:08:31:
After running a large number of games, it appears that the programs are approximate statistical equals. In addition, by repeating speed NPS calculation tests under high priority as single tasks, it appears the speed advantage is really very close to a push. Here are the results, with failef being the new version:
Program Elo + - Games Score Av.Op. Draws
1 Failef : 2306 56 59 149 51.7 % 2294 0.0 %
2 Faile_06 : 2294 59 56 149 48.3 % 2306 0.0 %
Considering that the ELO difference is about ten with a standard deviation spread of over 100, any increase in strength at all is questionable at best.
It seems that any modest speed improvements I have attempted have been almost exactly offset by attempted safety improvements.
Because I have not managed to create any measurable statistical speed-up with this version, I think it prudent to run the original in my contest, since it is very thoroughly tested, having had a year of service by this time.


My FTP site