[Pachi] Regression ?

Stanislaw Halik sthalik at misaki.pl
Fri Apr 15 12:43:39 CEST 2016


On 2016-04-10 03:37, Petr Baudis wrote:
> I will try with four, but that's as high I can go for now...
>
> sthalik's struct uct_node layout changes could have had some unforeseen
> threading consequences wrt. lockless stats updates, but that should have
> happenned *before* the regression you reported.

In Message-ID: <57011BF6.6030208 at free.fr> lemonsqueeze said that 32e02d4 
is fine.

It could theoretically have problems with same (aligned) word 
rw-accessed by multiple threads without synchronization, that needs stop 
and resync of all CPUs. But that only happens with races, which are the 
proper error.

I haven't dug deep into pachi's logic but the lockless architecture 
looks quite sound so far.

sh



More information about the Pachi mailing list