Author: Uri Blass
Date: 02:21:35 12/06/02
Go up one level in this thread
On December 06, 2002 at 05:03:13, Richard Pijl wrote:
>On December 06, 2002 at 03:53:26, Uri Blass wrote:
>
>>I decided to build a bench command for movei.
>>
>>Here is an example for first bench command that calculate number of nodes at
>>depth 7 and print information about the nodes for every position.
>
>I tried to level the amount of time between positions some time ago, so that
>every included position would take about the same time.
>In addition, I also accept a parameter in the bench command which will be used
>to increase the selected depth. So bench 0 will perform a fixed depth search for
>every position to a depth specific for that position. Bench 1 will search one
>deeper for every position and so on.
>>
>>choosemove(0) means that it does not print the analysis but if the number of
>>nodes is the same I do not need the analysis.
>>
>>void bench(void)
>>{
>> int totalnodes=0;
>> max_depth=7;
>> setup("1r3rk1/3qn1bp/p1np2p1/1ppNpp2/2P2P2/1PNPP1P1/PB4QP/1R3RK1 w - - 0 1");
>> choosemove(0);
>> printf(" %d ",nodes);
>> totalnodes+=nodes;
>> setup("1rb2rk1/1pq3bp/1Np1n3/p1Pp1p1n/1P1Np3/P2PP1P1/1BQ3BP/R4RK1 w - - 0 1");
>> choosemove(0);
>> printf(" %d ",nodes);
>> totalnodes+=nodes;
>> setup("r4rk1/2q1bppp/2bp2n1/1p1Bp3/1BnPP3/2P1NN2/5PPP/R2Q1RK1 w - - 0 1");
>> choosemove(0);
>> printf(" %d ",nodes);
>> totalnodes+=nodes;
>> setup("3qrrk1/pb5p/1p1p2p1/2p1pp2/1PPbPPn1/P2P2P1/1BQ3BP/1R1N1R1K w - - 0 1");
>> choosemove(0);
>> printf(" %d ",nodes);
>> totalnodes+=nodes;
>> printf(" %d ",totalnodes);
>>}
>>
>>I think that if the total number of nodes are the same then probably the nodes
>>for every position is the same.
>>
>>My questions are the following:
>>1)What statistics should I include in the bench commands
>>to do it more productive.
>
>I haven't got much besides nodes either. I print only totals for the positions
>together as the normal statistics output can be used per position (just type
>'post' after starting bench).
>The statistics printed for the total set are: nodes, qnodes, time and knps.
I understand but in this way you cannot find a bug that happen only in games and
not in analysis.
possible ideas to catch more bugs are also to tell the program to play against
itself in fixed depth and count nodes to finish the game but I wait to hear from
more expereinced programmers.
>
>>
>>I am sure that more statistics is needed and not only the number of nodes at
>>fixed depth.
>>
>>I guess that it may be better if I write a function
>>nextstring(epdfile) to get the next string.
>>so I can use one loop instead of a lot of setup commands.
>
>I'm using a fixed set of positions so I don't really care about this.
>Richard
I understand but I want to use a set of hundreds of positions(the GCP test suite
includes 183 positions and wac include 300 positions) and I do not want to print
hundreds of lines in the source of bench.
Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.