Author: Christophe Theron
Date: 23:11:28 11/10/99
Go up one level in this thread
On November 10, 1999 at 22:28:49, leonid wrote:
>On November 10, 1999 at 21:04:11, Christophe Theron wrote:
>
>>On November 10, 1999 at 17:51:07, leonid wrote:
>>
>>>On November 10, 1999 at 13:31:45, Christophe Theron wrote:
>>>
>>>>On November 10, 1999 at 07:15:37, leonid wrote:
>>>>
>>>>>
>>>>>>You can do something faster in assembly, but it takes such a long time to
>>>>>>develop it that in the end you lose your advantage.
>>>>>>
>>>>>>Because chess programming is about being creative, and assembly lengthens the
>>>>>>time between the idea and the implementation. That's the key.
>>>>>>
>>>>>>
>>>>>> Christophe
>>>>>
>>>>>In reality, it is not writing the code that is the most time consuming in
>>>>>programming (at least in mine) but verification of each version of logic.
>>>>>Verification for speed. Writing the code take hardly 5 or 10% from the total
>>>>>time for creating the game. This is why language must have so little impact on
>>>>>the time of writing the chess game.
>>>>>
>>>>>If the last change in my logic took some 5 hours for writing it, after 4 days of
>>>>>verification of positions I still don't know how much advantage I can obtain
>>>>>from the last change. I imagine that the same is true for everybody. This is why
>>>>>I would like to hear from you, or somebody else, how much really the time goes
>>>>>in writing the game compared with everything else.
>>>>>
>>>>>Leonid.
>>>>
>>>>between one and two hours a day.
>>>>
>>>>Anyway that's not the problem.
>>>>
>>>>Here is how I look at it: 100% of the time I spend in my sources is spend
>>>>reading C, not assembly, and for me that makes a big difference.
>>>>
>>>>When I'm not in my sources, I'm not working on Tiger. When my program is running
>>>>automatic tests I work on something else.
>>>>
>>>>
>>>> Christophe
>>>
>>>Can hardly imagine how you do your test. For me the test for speed is the
>>>verification of time that two logics ask for solving the same position. I must
>>>verify big number of positions in order to be certain that response is not
>>>aberration. And deposition of big number of different positions, taken very
>>>often from different sources, take time. To give you one idea about aberration.
>>>The last time I verified the new logic on the first 20 position, just asking the
>>>game to play on its own. The speed improvement was 160%. After this I took the
>>>positions from the Chess Life and tryed the same there on around next 18.
>>>Advantage was hardly 10%. Where I am? I still don't know. Tomorrow will continue
>>>my verification.
>>>
>>>Leonid.
>>
>>Being able to check if a change is an improvement or not is indeed the key to
>>really improve a program.
>>
>>It is very important to invest time to find a good testing methodology.
>>
>>
>> Christophe
>
>Almost impossible. Each time you have some new thing to try. The only way around
>the problem is to give all the try to somebody else. But this is impossible when
>you work on your own. Maybe your situation is different and this is how you are
>spending so much time with your code.
>
>Ho, maybe you could help me in solving my old mystery. Can you describe, in your
>way, at what speed now games search the position? For mate containing position I
>was able to find exact speed, but never found the way to know the speed for
>positional search. It make me wonder how far mine is from the good one.
>
>At what stage is your project?
>Leonid.
I don't understand your question.
Christophe
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.