MikeB wrote: ↑Fri Jun 29, 2018 7:30 am
Nice position - thx for sharing.
Here current McBrain (not released) on a lowly Raspberry Pi Model 3B+(DGT-PI Housing) sees the mate in 8 at depth 32 and 765M nodes.
( Actually used this position to test Pi throttling - it ran for about ~20 minutes and maintained 1200 Mhz while using all 4 cores - the hardware falls in between iPad 4 and iPad 5 - Geekbench result:http://browser.geekbench.com/geekbench2/2659839),
The below analysis was captured on Mac Pro which was ssh-ed into the Pi via a WIFI connection.
[snip]
Update - incorporated Dann Corbit''s mobility idea into McBrain v9.5 ( thanks Dann!)
- but note, I was using Ivan Ivec's Depth R function - so this is new code marrying Dann's mobility function and Ivan's Depth R function - the marriage is a good one.
now less than 5 minutes on a lowly Pi, this 9.5 version has new minimal output option , and it will be released in current form. Current code is out on Github for those that can't wait. Will throw the macOS and Pi versions out there too.
MikeB wrote: ↑Sat Jun 30, 2018 5:24 am
note - I have found two cores optimal for the Pi - do not use all 4 cores , solution time longer with all 4 codes.
Did you try 3?
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
MikeB wrote: ↑Sat Jun 30, 2018 5:24 am
note - I have found two cores optimal for the Pi - do not use all 4 cores , solution time longer with all 4 codes.
Did you try 3?
Just kicked it off, the Pi has an auto throttle feature - that drops Mhz tp 1200 once a specified temp is reached. With two cores, I don't hit that threshold at all - so I can have two cores running at 1500 Mhz, in less than a minute or two it start throttling down 3 cores to 1200 , almost to the point that is at least 50%. The Pi smp capability is not like that of high end processors of today, designed for smp programs, so it not as effective. The processor was used in Apple iPad 4 and 5.
It's been running 5 minutes now it still hasn't;t found it with 3 cores - but my guess is that will find it soon- it's near the depth range where it usually finds it. (depth 28 to 32)
edit - 3 core was not good - not found in over 20 minutes
Henrik Dinesen wrote: ↑Sat Jun 30, 2018 1:53 am
It's obviously a problem for several strong engines, but then again it points to the many holes in the Dutch defence.
The move 2...h6 is a dubious response to 2.Bg5; and an early ...Nf6, allowing White to double the black pawns, is plain bad in this line: White gets a long-lasting strategic initiative, while Black's position is highly passive and unpleasant. That doesn't mean that 2.Bg5 makes the Dutch unplayable, Black has reasonable alternatives in 2...g6, 2...d6 or 2...c6. Even 2...h6 is playable, although the resulting mess may not be everyone's cup of tea.
current McBrain 9.6 ( not released) sees it in 48 seconds on my machine using a modified Monte Carlo search function written originally by Stefano Cardanobile
MikeB wrote: ↑Wed Jul 18, 2018 7:02 am
current McBrain 9.6 ( not released) sees it in 48 seconds on my machine using a modified Monte Carlo search function written originally by Stefano Cardanobile
