zullil wrote:Seems like a rather complicated way to benchmark. I mean, why should I have to download Arena?
How about using Stockfish's standard built-in bench command?
Can't really tell what you're doing. Is Stockfish being asked to search just the position below until Qg4 appears?
[D]rnbq1rk1/pppnbppp/4p3/3pP1B1/3P3P/2N5/PPP2PP1/R2QKBNR w KQ - 0 7
I don't agree with you... I did not start creating benchmarks from yesterday.. since more than 10 years I am creator...and experience is already talking, understand ?
So...actually is not complicated, if you have fast internet, to download it will take just a few seconds...
Otherwise... it will take more time to download Arena, to download exactly same Stockfish versions, to create position...to set time control, to set hashtables and so on... really this mesagge is more complicated...
I spent many efforts to replay you,,, in this time you could send me a benchmark
But however, no problem if you wish you can pıost your benchmark from here too, on this forum i mean...
Best,
Sedat
Last edited by Sedat Canbaz on Sat Feb 08, 2014 1:28 am, edited 1 time in total.
Using the built in bench would make it possible for users of any operating system that Stockfish can run on to contribute. Then there is no need to download a special Stockfish and Arena, which I don't have. It does not make a big difference which version of Stockfish is used, as long as it does the benchmark with the same 30 positions of course. Okay, people can cheat but not many would I hope. With a screenshot you could still check if the benchnumber is correct so it is not super easy to fake the results. The normal bench only uses one thread though.
Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Eelco de Groot wrote:Using the built in bench would make it possible for users of any operating system that Stockfish can run on to contribute. Then there is no need to download a special Stockfish and Arena, which I don't have. It does not make a big difference which version of Stockfish is used, as long as it does the benchmark with the same 30 positions of course. Okay, people can cheat but not many would I hope. With a screenshot you could still check if the benchnumber is correct so it is not super easy to fake the results. The normal bench only uses one thread though.
Eelco
Yes...this is my list, my rules... do you understand ?
If you dont accept, sad, but no problem
Btw, no patience too see your Stockfish benchmark list
Sedat Canbaz wrote:
Yes...this is my list, my rules... do you understand ?
Yes, understood but it is still a fair point to mention when something will limit contributions such as using an operating system dependent component. I like your idea anyway even if operating system constraints exclude me from potential contribution. Good luck with the list.
Sedat Canbaz wrote:
I don't agree with you... I did not start creating benchmarks from yesterday.. since more than 10 years I am creator...and experience is already talking, understand ?
So...actually is not complicated, if you have fast internet, to download it will take just a few seconds...
Otherwise... it will take more time to download Arena, to download exactly same Stockfish versions, to create position...to set time control, to set hashtables and so on... really this mesagge is more complicated...
I spent many efforts to replay you,,, in this time you could send me a benchmark
But however, no problem if you wish you can pıost your benchmark from here too, on this forum i mean...
Best,
Sedat
Does Arena even run on linux or OS X? Without Wine? All I see in your download are .exe files. Is this Windows only? Do I have to install Wine just to use your method?
Perhaps you could simply explain how the benchmark is done? I'm supposed to run one particular compile of one version of SF, with specific settings, on one position until the search does what?
Sedat Canbaz wrote:
I don't agree with you... I did not start creating benchmarks from yesterday.. since more than 10 years I am creator...and experience is already talking, understand ?
So...actually is not complicated, if you have fast internet, to download it will take just a few seconds...
Otherwise... it will take more time to download Arena, to download exactly same Stockfish versions, to create position...to set time control, to set hashtables and so on... really this mesagge is more complicated...
I spent many efforts to replay you,,, in this time you could send me a benchmark
But however, no problem if you wish you can pıost your benchmark from here too, on this forum i mean...
Best,
Sedat
Does Arena even run on linux or OS X? Without Wine? All I see in your download are .exe files. Is this Windows only? Do I have to install Wine just to use your method?
Perhaps you could simply explain how the benchmark is done? I'm supposed to run one particular compile of one version of SF, with specific settings, on one position until the search does what?
Thanks...but I never used Linux...that's why I have no any idea about Arena GUI runs on Linux or not...
Btw, looking at your current results:14517 Nodes/second are very good indeed
What is your hardware and how many cores, ponder off, HT ON or OFF ?
But however, I plan to publish the benchmark results especially which are done via Arena 2.0.1GUI (1024 Hashtable size, ponder off, 120min+30sec...) otherwise the results can lead to misunderstandings
Sedat Canbaz wrote:
Yes...this is my list, my rules... do you understand ?
Yes, understood but it is still a fair point to mention when something will limit contributions such as using an operating system dependent component. I like your idea anyway even if operating system constraints exclude me from potential contribution. Good luck with the list.
