Sloppy 0.1.0 is here!

[Jim Ablett]

Hi Ilari,

Yes, should have defined below >

Code: Select all

#define PRIu64 "I64u"

I recompiled, but benchmark the same.

In perft.c (line 218) the "divide" node count is multiplied by 3. I don't understand why.

Yes sorry about sending that in the src. The compiled exe didn't use that.
I was testing to see how much out exactly the nodecount was compared to your build.

Jim.

[Ilari Pihlajisto]

But because Sloppy 0.1.0 is also released under the GPL (version 3), everything appears to be OK.

Yup. And I intend to keep Sloppy GPL'd in the future as well. Any forks or clones will be taken as compliments and signs of success, not as competition.

I'll send you a Mac OS X executable and the modified source code as soon as I succeed.

Thanks a bunch. I don't own a Mac myself, but I'll be happy to host an OSX version too.

I recompiled, but benchmark the same.

Which benchmark (bench or perft)? "perft 7" should definitely display a wrong node count if you print it with %d. The right number is 3195901860, your build gives me -1099065436.

Yes sorry about sending that in the src. The compiled exe didn't use that.

No problem. But the build definitely does use it, at least the "all processors" version does. Here's a perft comparison:

My builds:

Code: Select all

White: divide 7
0: a3 106743106
1: b3 133233975
0: c3 144074944
1: d3 227598692
1: f3 102021008
0: e3 306138410
1: g3 135987651
0: h3 106678423
1: a4 137077337
0: b4 134087476
1: c4 157756443
0: d4 269605599
0: f4 119614841
1: e4 309478263
0: g4 130293018
1: h4 138495290
0: Na3 120142144
1: Nc3 148527161
1: Nh3 120669525
0: Nf3 147678554
Perft(7): 3195901860 nodes.

Your build:

Code: Select all

White: divide 7
0: a3 320229318
1: b3 399701925
0: c3 432224832
1: d3 682796076
1: f3 306063024
0: e3 918415230
1: g3 407962953
0: h3 320035269
1: a4 411232011
0: b4 402262428
1: c4 473269329
0: d4 808816797
0: f4 358844523
1: e4 928434789
0: g4 390879054
1: h4 415485870
0: Na3 360426432
1: Nc3 445581483
1: Nh3 362008575
0: Nf3 443035662
Perft(7): -1099065436 nodes

[Jim Ablett]

Hi Ilari,

Here's a new build with the fixes. My Apologies.
Looks like you got one of my test builds by mistake.

http://myfreefilehosting.com/f/75a5432037_1.41MB

Rgds,
Jim.

[Ilari Pihlajisto]

Perfect, thanks!

I have added your changes to my own source code, so Sloppy 0.1.1 (when released) should compile with MSVC++ right away.

[Vladimir Medvedev]

Thank you for Sloppy. Source code is very readable and self-documented. I enjoy reading it. I plan to include Sloppy into next iteration of my open-source engines rating pool.

[Ilari Pihlajisto]

Okay, Sloppy 0.1.1 is now MSVC-friendly. I had a problem with mismatching hash keys, and it turned out that the RAND64 macro (in hash.c) wasn't working the same way with every compiler. I'll get the new release out hopefully by the end of the week.

Thank you for Sloppy. Source code is very readable and self-documented. I enjoy reading it. I plan to include Sloppy into next iteration of my open-source engines rating pool.

Thanks. I'm glad someone else can understand my code.

[Ryan Benitez]

Hello Vladimir!

Just two questions.

1. Isn't Fruit 2.1 released under GPL?
2. Doesn't GPL imply distribution of source code for any derivatives of GPL'd software?

The answer to both questions is "yes". But because Sloppy 0.1.0 is also released under the GPL (version 3), everything appears to be OK.


Tord

I read his eval function and I my understanding is that because it is written in his own code he would not have to GPL it if he did not want to. I am happy he did choose to GPL the program though. Not because of Fruit but because it helps computer chess overall to have more GPL chess engines.

Ryan

[Ilari Pihlajisto]

I read his eval function and I my understanding is that because it is written in his own code he would not have to GPL it if he did not want to.

That would be a slippery slope though. Had I not decided to release under GPL, I probably wouldn't have examined Fruit's evaluation at all.

I am happy he did choose to GPL the program though. Not because of Fruit but because it helps computer chess overall to have more GPL chess engines.

I happy about it too.

[YvesLejeail]

Hi,
looks like creating a strong chess engine is easy
How many time did you spent on your project (just curious and a bit jealous)?
Yves

[Ilari Pihlajisto]

Hi,
looks like creating a strong chess engine is easy
How many time did you spent on your project (just curious and a bit jealous)?
Yves

I haven't counted the man-hours, but it probably took about 4h/week for 2 years. When I started I wasn't really familiar with the C language, so learning it took a lot of that time.

Btw, looks like Sloppy 0.1.1 is going to be at least 50 ELO stronger than 0.1.0.

[YvesLejeail]

Hi,
very good work and efficient learning ! I could give wife, children, dog, against a 50 elo bonus for my chess engine !
yves

[Denis P. Mendoza]

Ilari,

I haven't heard news about your latest works on Sloppy. Thanks to valuable tips from the only active compiling guru, Master Jim, I finally built a faster 64-bit version of your Sloppy 0.1.1.

I just can't figure which one is best, due to limited time. I just tested using Opremp, and it seems a nice improvement. Here's the Intel compiles to test. You'll be surprised. Enjoy!

http://www.zshare.net/download/68786040e06ac4/