Re: [Bug-gnubg] Re : Re: Optimal settings for MacBookPro

[Philippe Michel]

On Mon, 29 Nov 2010, pierre zakia wrote:

> I did the test again with only X11 running (no Airport, no Ethernet connection). 
> The results are:
> 1 thread: 52 000 000
> 2 thread: 90
> 3 thread: 104
> 4 thread: 40 
> 5 thread: 40 
> 6 thread: 39 
> 7 thread: 39
> 8 thread: 39
>
> 3 thread is still the winner. Why ?

I don't know, but there is more in the above numbers than "3 is the 
winner". It looks like with 4 or more threads, there is only one active, 
giving the same speed than with one, minus some overhead.

Note that when you set the thread number in gnubg to 4, the process really 
has 5 threads : one controller, mostly inactive, and 4 workers that do 
evaluations, rollouts or speed tests. It looks like, in your case, it just 
doesn't work with more total threads than cpu cores.

> I am running the build 0.9.0 (downloaded from gnubg.org) that was compiled for 
> OSX Leopard (in June 2008)
>
> As I wrote, it runs under OSX Snow Leopard, but I never had any issues with this 
> build like crash, freeze, blurred 3D or something else.

The fact that it was build for an older OS may be an explanation. 
Hopefully Louis Zulli's proposed test will clarify this. As far as I can 
read in the changelog, there were bug fixes to it since then, but 
multithreading was working mostly ok in 2008.