bug-gnubg
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Benchmarks on server class machines and resultingchange


From: Michael Petch
Subject: Re: [Bug-gnubg] Benchmarks on server class machines and resultingchange requests
Date: Fri, 11 Sep 2009 01:16:54 -0600
User-agent: Microsoft-Entourage/12.20.0.090605

The file “batch” explicitly sets to 2, but I noticed that not all your scripts concatenate the “batch” file. Bench.sh does, bench2.sh does, bench3 doesn’t, and bench4-.sh (and subfile bech4-sub). Maybe I am tired at this hour and am missing it as well. It seems like some are 0 ply and some are 2 ply?

On 11/09/09 1:01 AM, "Ingo Macherius" <address@hidden> wrote:

I shouldn't write mails that early in the morning. See the file "batch" for why it wasn't 0ply. 2ply was set explicitly ...

Ingo

 
-----Original Message-----
From: Michael Petch  [mailto:address@hidden]
Sent: Friday, September 11, 2009  8:41 AM
To: Ingo Macherius; address@hidden
Subject: Re:  [Bug-gnubg] Benchmarks on server class machines and resultingchange  requests


The 0 ply might explain why your performance is  marginal at around ~10% (This is a guess until I do a run on your  scripts/data). But at 0 ply, each move is basically independent. Inputs get  fed to the, the results come out, they get added to the cache but the  likelyhood of that cache getting reused is slim. If you do 1 ply all the  possible distinct rolls (21 of them) get analyzed, so generally the next move  in the game is already cached from the previous data.

The only thing  that had me real curious on the data was how a large cache (especially now  that I know it was 0 ply) has significant overhead as your charts suggest  (> 2^25) except that it went out to swap space and was paging data and was  more costly than going to the neural net to recalculate the numbers. I would  have expected it to reach a threshold and stay there. My testing on 2 ply  suggests that. Now I am using at present the code as of September 10th, so  I’ll also produce the data with an older CVS release similar to yours and see  what I get with your data on my system, and compare it to the latest – but do  it on 0 ply,1ply and 2 ply.

I’m runnign some 4 ply cache tests ovr the  next 24 hours. I’ll queue up your scripts when its complete and see what  happens.

On 10/09/09 11:42 PM, "Ingo Macherius" <address@hidden> wrote:

 
Ah, one missing detail: I didn't use any .gnubgrc, so all  settings are whatever the defaults for a newly built binary were on August  1st 2009. Probably that means 0 ply.

Ingo




reply via email to

[Prev in Thread] Current Thread [Next in Thread]