Is Crafty 17.12 very slow?

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 14 July 2000 17:26:15:
I downloaded Crafty 17.12 and ran a benchmark, compared it to the previous version and discovered that 17.11 is 100 Knps faster on my machine. Does anyone get the same results?
Best wishes...
Mogens

[Frank Quisinsky]

Geschrieben von: / Posted by: Frank Quisinsky at 14 July 2000 19:34:58:
Als Antwort auf: / As an answer to: Is Crafty 17.12 very slow? geschrieben von: / posted by: Mogens Larsen at 14 July 2000 17:26:15:

I downloaded Crafty 17.12 and ran a benchmark, compared it to the previous version and discovered that 17.11 is 100 Knps faster on my machine. Does anyone get the same results?
Best wishes...
Mogens

On my machine (Dual PIII 733EB) 448.000 NPS, version 17.11 = 494.000 NPS) !
Regards
Frank

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 14 July 2000 19:50:03:
Als Antwort auf: / As an answer to: Is Crafty 17.12 very slow? geschrieben von: / posted by: Mogens Larsen at 14 July 2000 17:26:15:

I downloaded Crafty 17.12 and ran a benchmark, compared it to the previous version and discovered that 17.11 is 100 Knps faster on my machine. Does anyone get the same results?

I don't think the build on the UAB ftp server is one of the world famous "Nalimov" builds. Since there is no added functionality for me anyway, I will wait for 17.13 before I do another build.


my ftp site

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 14 July 2000 21:21:03:
Als Antwort auf: / As an answer to: Re: Is Crafty 17.12 very slow? geschrieben von: / posted by: Dann Corbit at 14 July 2000 19:50:03:

I downloaded Crafty 17.12 and ran a benchmark, compared it to the previous version and discovered that 17.11 is 100 Knps faster on my machine. Does anyone get the same results?

I don't think the build on the UAB ftp server is one of the world famous "Nalimov" builds. Since there is no added functionality for me anyway, I will wait for 17.13 before I do another build.

I just wanted to check if it was faster than 17.11, but it was a lot slower without significant knowledge added. Crafty 17.13 will be my next version as well, unless Nalimov builds a version of 17.12 that's very fast ). With deault hashsize the nps count were:
17.12: 396000 nps,
17.11: 493000 nps,
so the difference is significant. More so for Athlon than Pentium.
Best wishes...
Mogens

[Dieter Buerssner]

Geschrieben von: / Posted by: Dieter Buerssner at 15 July 2000 16:04:19:
Als Antwort auf: / As an answer to: Re: Is Crafty 17.12 very slow? geschrieben von: / posted by: Mogens Larsen at 14 July 2000 21:21:03:

I just wanted to check if it was faster than 17.11, but it was a lot
slower without significant knowledge added. Crafty 17.13 will be my next
version as well, unless Nalimov builds a version of 17.12 that's very
fast ). With deault hashsize the nps count were:
17.12: 396000 nps,
17.11: 493000 nps,

AMD K6-2 300, hash 6M hashp 4M:
17.12 : 98600
17.11 : 88700
No, it's not a typo. For me 17.12 is faster. BTW. The difference is
roughly comparable, to the difference between Yace compiled by Dann
and Yace compiled by myself.
I'd be interested, what the following produces on higher end machines
for Yace (in comparision to Crafty's result). It is roughly comparable to
Crafty's bench. It will clutter the screen, but at the end will report an
average nps. Just save as bench.bat in the Yace directory and run.
I get 118000 nps. (But usually in middle game I get somewhat higher
numbers).
Regards,
Dieter
REM bench.bat
REM Simulate Crafty's bench for Yace
REM If you changed the name of the executable, just change the last line
REM After the batch file was executed once, you can also just start Yace
REM and then type input bench.inp
REM Just search positions 2, 4, 8, 12, 22, 23 of
REM of Bratko-Kopec 24 position test suite
REM No sd in Yace yet, so use nodes or st
echo st 30 > bench.inp
echo use_book off >> bench.inp
echo test bench.ci >> bench.inp
REM echo quit >> bench.inp
echo noop BK.ci: Bratko-Kopec 24 position test suite > bench.ci
echo echo position 02 >> bench.ci
echo svfe 3r1k2/4npp1/1ppr3p/p6P/P2PPPP1/1NR5/5K2/2R5 W - - 0 1 >> bench.ci
echo srch d5 >> bench.ci
echo echo position 04 >> bench.ci
echo svfe rnbqkb1r/p3pppp/1p6/2ppP3/3N4/2P5/PPP1QPPP/R1B1KB1R W KQkq - 0 1 >> bench.ci
echo srch e6 >> bench.ci
echo echo position 08 >> bench.ci
echo svfe 4b3/p3kp2/6p1/3pP2p/2pP1P2/4K1P1/P3N2P/8 W - - 0 1 >> bench.ci
echo srch f5 >> bench.ci
echo echo position 12 >> bench.ci
echo svfe r3r1k1/ppqb1ppp/8/4p1NQ/8/2P5/PP3PPP/R3R1K1 B - - 0 1 >> bench.ci
echo srch Bf5 >> bench.ci
echo echo position 22 >> bench.ci
echo svfe 2r2rk1/1bqnbpp1/1p1ppn1p/pP6/N1P1P3/P2B1N1P/1B2QPP1/R2R2K1 B - - 0 1 >> bench.ci
echo srch Bxe4 >> bench.ci
echo echo position 23 >> bench.ci
echo svfe r1bqk2r/pp2bppp/2p5/3pP3/P2Q1P2/2N1B3/1PP3PP/R4RK1 B kq - 0 1 >> bench.ci
echo srch f6 >> bench.ci
echo rtrn >> bench.ci
yace < bench.inp

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 15 July 2000 18:40:37:
Als Antwort auf: / As an answer to: Re: Is Crafty 17.12 very slow? geschrieben von: / posted by: Dieter Buerssner at 15 July 2000 16:04:19:

AMD K6-2 300, hash 6M hashp 4M:
17.12 : 98600
17.11 : 88700
I'd be interested, what the following produces on higher end machines
for Yace (in comparision to Crafty's result). It is roughly comparable to
Crafty's bench. It will clutter the screen, but at the end will report an
average nps. Just save as bench.bat in the Yace directory and run.
I get 118000 nps. (But usually in middle game I get somewhat higher
numbers).

That's a very low node count per MHz. My old PII-266MHz did around 130000, but Crafty do have problems performing with the K6, so it's not that surprising. Fortunately, the K7 don't have the same problems.
I get this:
solution not found
6 tested, 3 found, 0 mates
66221925 nodes, 25019411 leavenodes in 174.85 sec 378742 n/s
test nodes 28711642 win nodes 93315 mate nodes 0
win time 0.35 mate time 0.00 av depth 10.833 (nm 10.833) maxdepth 32, tu 1022
log(win_nodes) = 10.53703 log(win_time) = -4.52637
I assume 378742 n/s is the average nps.
Best wishes...
Mogens