Senpai 1.0 (new engine)

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Xann
Posts: 127
Joined: Sat Jan 22, 2011 7:14 pm
Location: Lille, France

Re: Senpai 1.0 (new engine)

Post by Xann »

Hi Mark!

I don't smoke, but I do computer chess.
I tried to quit but ... it's not easy.
Say NO to computer chess ;)
mvk
Posts: 589
Joined: Tue Jun 04, 2013 10:15 pm

Re: Senpai 1.0 (new engine)

Post by mvk »

Xann wrote: I have been working on a new GPL engine: Senpai.
Very nice and great to see you back!

Can you enlighten us about the cause of the 50% TT bug if you have found it? I think I have seen it before, long long time ago, but not even sure if it was in my own program or elsewhere...
[Account deleted]
Joerg Oster
Posts: 937
Joined: Fri Mar 10, 2006 4:29 pm
Location: Germany

Re: Senpai 1.0 (new engine)

Post by Joerg Oster »

Nice to see you're back!

Senpai runs just fine under LinuxMint on my AMD FX-8120.
Another great participant for the next TCEC!
Thanks!
Jörg Oster
Gerd Isenberg
Posts: 2250
Joined: Wed Mar 08, 2006 8:47 pm
Location: Hattingen, Germany

Re: Senpai 1.0 (new engine)

Post by Gerd Isenberg »

Hi Fabien,

Wow, great news!
And congrats also for the 10th Fruit 1.0 anniversary.

I see, Senpai has little endian file:rank mapping. Allows cheaper pawn moves. Little bit disappointed about so many loops so far ;-)

For sliders, you traverse not only the first blocker but all pieces on those rays?

Code: Select all

bit_t piece_attacks_from(int pc, int f, const board::Board & bd) {

   assert(pc != piece::PAWN);

   bit_t ts = Piece_Attacks[pc][f];

   for (bit_t b = bd.all() & Blockers[pc][f]; b != 0; b = bit::rest(b)) {
      int sq = bit::first(b);
      ts &= ~Behind[f][sq];
   }

   return ts;
}
I guess you will gain some speedup later ;-)

Cheers & much fun,
Gerd
Xann
Posts: 127
Joined: Sat Jan 22, 2011 7:14 pm
Location: Lille, France

Re: Senpai 1.0 (new engine)

Post by Xann »

I wish I knew the cause. I tried everything from random numbers, different ways of measuring the percentage, switching to unsigned numbers. None of this made any difference. So I guess one of the remaining candidates would be the scoring formula to decide which entry to overwrite.
Xann
Posts: 127
Joined: Sat Jan 22, 2011 7:14 pm
Location: Lille, France

Re: Senpai 1.0 (new engine)

Post by Xann »

TCEC uses Linux?
But ... Senpai must first pass some stability test I suppose; we couldn't test SMP very much.
Xann
Posts: 127
Joined: Sat Jan 22, 2011 7:14 pm
Location: Lille, France

Re: Senpai 1.0 (new engine)

Post by Xann »

Hi Gerd!

BTW I used CPW a lot when beginning the engine. This is so much easier than reading hundreds (if not thousands) of old posts.

Yes, I loop over all blockers. I somehow find this simple and elegant with the endgame in mind. In the middle game, I agree it's a sure way to cripple NPS though. There's a reason why nobody uses this.
Gerd Isenberg
Posts: 2250
Joined: Wed Mar 08, 2006 8:47 pm
Location: Hattingen, Germany

Re: Senpai 1.0 (new engine)

Post by Gerd Isenberg »

Xann wrote: Yes, I loop over all blockers. I somehow find this simple and elegant with the endgame in mind. In the middle game, I agree it's a sure way to cripple NPS though. There's a reason why nobody uses this.
Ok, this deserves an own sliding piece attack site devoted to you ;-)
Last edited by Gerd Isenberg on Mon Mar 17, 2014 8:16 pm, edited 1 time in total.
Xann
Posts: 127
Joined: Sat Jan 22, 2011 7:14 pm
Location: Lille, France

Re: Senpai 1.0 (new engine)

Post by Xann »

Are you sure? This looks like the most basic way to do bitboards.
tralala
Posts: 16
Joined: Fri Jan 10, 2014 11:49 pm

Re: Senpai 1.0 (new engine)

Post by tralala »

JuLieN wrote:As Fabien is a bit busy right now, I'll take a shot and emulate what his answers will probably be. :)
Steve Maughan wrote:+ After an absence of a few years, what was the motivation for writing Senpai?
The clones wars were fading, so I needed to release some new strong open sources. Plus we need a new R[edited out by moderator]a 5.
Steve Maughan wrote:+ How long did it take you write Senpai?
A pair of hours in a six years span.
Steve Maughan wrote:+ Why did you start a new codebase? Why not continue with the development of Fruit?
Fruit is lame, it's only about 2800.
Steve Maughan wrote:+ Have you accomplished everything you set out to do with Senpai, or is this a work-in-progress?
The answer is contained in the name of the engine.
Steve Maughan wrote:+ LMR became popular after it was introduced in Fruit - is there any new "secret sauce" in Sanpai which you think may be of interest to other programmers?
There's a big hint hidden in the release readme !
Steve Maughan wrote:+ Why did you choose C++11 as the development language?
Trendy is funny.

*hiding* :oops:
:D :D :D Too bad there is no liking functionality.