The Week in Chess

Tuesday, April 30, 2013

Stockfish 130429 Tests

The latest build of Stockfish dated April 29, 2013 was released.

Change log:
Author: Marco Costalba
Date: Mon Apr 29 00:55:32 2013 +0200
Timestamp: 1367189732

Temporary revert "Expose EvalInfo struct to search"

It is not needed for the release and introduces
a slowdown, although very small.

Probably it will be readded after the release.

No functional change.


I noticed that the past few releases since April 25 contained some annoying updates.  The release on April 25 just before the TCEC final crashed on my computer. I was about to complain in one of the forums when I saw similar experience already posted by some users. It got fixed swiftly but the resulting tests were discouraging when it was thrashed by the previous build. Then another update came with this log:

Author: Marco Costalba
Date: Thu Apr 25 21:52:26 2013 +0200
Timestamp: 1366919546

Store Eval::Info in Search::Stack

Instead of a pointer. This should fix the issue of
remaining with a stale pointer when for instance calling
IID, but also null search verification, singular search
and razoring where we call search with the same ss
pointer. In this case ss->ei is overwritten in the
search() call and upon returning remains stale.

This patch could have a performance hit because Eval::Info
is big (176 bytes) and during splitting we copy 4 ss entries.

On the good side, this patch is a clean solution.

Proposed by Gary.

No functional change.  


I think that the last changes were something not to be proud about. The crash should not happen if it was thoroughly tested by the Stockfish team.  It was so embarrassing that I suspect there was little quality testing procedures that were followed. And then the regression that was known beforehand was published and then quickly reverted. Oh! my God...

Despite of this, I continued to test the latest version but no longer sure if it will be consistent with my previous tests. It was some kind of 100 round robin games between my 64-bit compilation with the April 29 build against the 64-bit auto-build at and Houdini 3 Pro x64.

Stockfish 130429 x64 m10i pgo - Compiled with VS 2010 with full optimization, PGO, IPP, MKL and TBB, beat Houdini 3 Pro by just 1 point and Stockfish 13042900 x64 with 10 points.

Stockfish 13042900 x64 - Auto-built with MinGW by lost to Houdini 3 Pro and Stockfish 130429 x64.

Houdini 3 Pro - Compiled with VS 2010 by +Robert Houdart won over Stockfish 13042900 x64 with 8 points but lost to Stockfish 130429 x64 with 1 point.

Here is the crosstable:

Engines 1 2 3 Points Win Loss Draw
Stockfish 130429 x64 m10i pgo xxx 35-34-31 33-23-44 105.5 68 57 75
Houdini 3 Pro x64 34-35-31 xxx 38-30-32 103.5 72 65 63
Stockfish 13042900 x64 23-33-44 30-38-32 xxx 91.0 53 71 76

These are run in AMD Quad Core with 4GB RAM, Windows 7 64bit virtualized under PcLinuxOS 2013.  The results are just for testing and not intended for serious statistics.

Download the winning engine Stockfish 130429 x64 here.

Download computer chess engines tournament games here.

No comments:

Post a Comment

Chessdom News