Author: José Carlos
Date: 16:45:03 09/03/02
Go up one level in this thread
On September 03, 2002 at 06:25:18, Uri Blass wrote:
>On September 02, 2002 at 20:26:30, Vincent Diepeveen wrote:
>
>>On September 01, 2002 at 12:08:21, Uri Blass wrote:
>>
>>Uri, i posted my answer at 11:55, you answerred at 12:08.
>>That's 13 minutes.
>>
>>How the HELL can you have looked up MTD in the meantime,
>>what it is and what it is doing, and where are your experiments
>>with MTD?
>>
>>For MOVEI it sure works as long are your eval is stupid.
>>
>>It would remove a lot of discussion about fritz here too.
>>like the stupid comments you don't nullmove in your PV.
>>
>>WHICH PV DO YOU TALK ABOUT?
>>
>>What kind of nonsense do you do in your program.
>>
>>CODE PLEASE.
>>
>>In fritz it's very easy i bet:
>>
>>int mtd() { // of course with a goto instead of a very slow function call
>> // hashtable probe
>> if( depth <= verysmalldepth ) {
>> lookupfasthahstable(L1 cache) // L2 cache lookup too slow of course
>> return(qsearch(bound)); // of course no function call but a goto, faster...
>> }
>> else if( depth < mediumdepth ) {
>> lookupslowhashtable(L2 cache); // main memory factor 100 slower
>> else // depth movei nevergets
>> lookupveryslowmemory(main memory);
>> // no returns of course in case of a truebound, because returns are too
>> // slow. Only a jump
>>
>> if( !incheck )
>> TryNullmove(); // best move to make first
>
>I admit that I do not understand this code
>
>I do not know how to tell my program to look at L1 cache or L2 cache.
>Maybe I will know better after I read the books that I ordered from amazon.
There's no instruction to tell the program to read from L1 or L2. What Vincent
means (if I get him right) is using very small tables that fit in cache (when
you have a small piece of data that you use very often, it's very likely to stay
in cache).
Apart from this, I think the above code is simply ridiculous.
José C.
>The only optimization that I cared about in movei is optimization of the
>algorithm.
>
>I guess that the only place that I use the very slow memory is in book moves(I
>do not use tablebases).
>
>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.