HOUDINI 5.01

[Ray Bongalon]

Houdini 5.01 is now available, all users will be informed by e-mail.
Houdini 5.01 contains some corrections and interface improvements, for a detailed list see http://www.cruxis.com/chess/manual/version_history.htm
Playing strength and analysis are unchanged compared to Houdini 5.

Thank you for reading,
Robert

Thank you.

[Damir]

Hi Robert

I have noticed that Houdini 5.01 has no numa support for Fritz GUI. Runs on my machine at about 65%.
Maybe you could see in Brainfish source how to code,so you can fix the Fritz GUI problem. The Fritz GUI Problem is tricky, as Fritz is not Numa aware...

[Houdini]

Houdini 5.01 is now available, all users will be informed by e-mail.
Houdini 5.01 contains some corrections and interface improvements, for a detailed list see http://www.cruxis.com/chess/manual/version_history.htm
Playing strength and analysis are unchanged compared to Houdini 5.

Thank you for reading,
Robert

When I run the installation another registration check is done. Does this reduce my allotment of 4 activations?

There is no real limit on the number of activations you use.
If you reach the limit, just send us an e-mail and we'll increase the limit.

We have used a very tolerant activation policy since 2012, basically there is no limit on installing on computers that you personally own.

[Houdini]

Hi Robert

I have noticed that Houdini 5.01 has no numa support for Fritz GUI. Runs on my machine at about 65%.
Maybe you could see in Brainfish source how to code,so you can fix the Fritz GUI problem. The Fritz GUI Problem is tricky, as Fritz is not Numa aware...

NUMA support does not depend on the GUI. It should work in every GUI you are using.

[Houdini]

allotment of 4 activations?

What?! Good to know, keeps me from ever buying Houdini.

It's incorrect, see our post above.

[tpoppins]

Thank you for the update, Robert; the new debug log feature is quite handy. Did you see this request, by the way? ;)

There's also what appears to be a cosmetic (but irritatating) bug that clutters analysis with almost identical (except for the node counts and times) lines:



Same depth, seldepth, eval and PV - why does Houdini print the same info twice for the same iteration? I thought that might be a bug with the Deep Fritz GUI but no, it isn't:

Code: Select all

<< info currmove f3f4 currmovenumber 1
<< info multipv 1 depth 21 seldepth 52 score cp 0 time 15330 nodes 436449542 nps 28470000 tbhits 0 hashfull 556 pv f3f4 e5c4 b3c4 c5c4 g5f6 g7f6 e4e5 f6g7 b2b3 c4c5 c3e4 c5d5 e4c3 d5c5
<< info currmove a2a3 currmovenumber 2
...
<< info currmove b3f7 currmovenumber 37
<< info multipv 1 depth 21 seldepth 52 score cp 0 time 15348 nodes 436966014 nps 28470000 tbhits 0 hashfull 557 pv f3f4 e5c4 b3c4 c5c4 g5f6 g7f6 e4e5 f6g7 b2b3 c4c5 c3e4 c5d5 e4c3 d5c5
<< info depth 22
<< info currmove f3f4 currmovenumber 1
<< info multipv 1 depth 22 seldepth 52 score cp 0 time 15634 nodes 445307350 nps 28483000 tbhits 0 hashfull 563 pv f3f4 e5c4 b3c4 c5c4 g5f6 g7f6 e4e5 f6g7 b2b3 c4c5 c3e4 c5d5 e4c3 d5c5
<< info currmove a2a3 currmovenumber 2
...
<< info currmove b3f7 currmovenumber 37
<< info multipv 1 depth 22 seldepth 52 score cp 0 time 16151 nodes 459949196 nps 28478000 tbhits 0 hashfull 575 pv f3f4 e5c4 b3c4 c5c4 g5f6 g7f6 e4e5 f6g7 b2b3 c4c5 c3e4 c5d5 e4c3 d5c5
<< info depth 23
<< info currmove f3f4 currmovenumber 1
<< info multipv 1 depth 23 seldepth 52 score cp 0 time 16246 nodes 462631659 nps 28476000 tbhits 0 hashfull 576 pv f3f4 e5c4 b3c4 c5c4 g5f6 g7f6 e4e5 f6g7 b2b3 c4c5 c3e4 c5d5 e4c3 d5c5
<< info currmove a2a3 currmovenumber 2
...
<< info currmove b3f7 currmovenumber 37
<< info multipv 1 depth 23 seldepth 52 score cp 0 time 20953 nodes 594813962 nps 28388000 tbhits 0 hashfull 703 pv f3f4 e5c4 b3c4 c5c4 g5f6 g7f6 e4e5 f6g7 b2b3 c4c5 c3e4 c5d5 e4c3 d5c5

Haven't seen this behavior with any other engine.

[tpoppins]

Odder still: Houdini resolves an iteration then displays two fail-highs and another apparently resolved iteration - all with the same depth value 30.

r1bq1rk1/bpppnp1p/p4p2/4P3/B3P3/2P2N2/PP3PPP/RN1Q1RK1 b - - 0 1

Code: Select all

Analysis by Houdini 5.01 Pro x64-popc:

[...]
10...Ng6 11.exf6 Qxf6 12.Qd2 d5 13.exd5 Nf4 14.g3 Nh3+ 15.Kg2 Kh8 16.Bd1 Rg8 17.c4 Bf5 18.Bc2 Rae8 19.Nc3 Bxc2 20.Qxc2 Nf4+ 21.Kh1 Nxd5 22.Qd3 Nxc3 23.bxc3 c6 24.Rad1 Rd8 25.Nd4 b5 26.Kg2 b4 27.Qf3 Qxf3+ 28.Kxf3 bxc3 29.Nxc6 Rxd1 
  +/= (0.52)  Depth: 30/61   00:10:52  18498MN, tb=23
10...Ng6 
  +/= (0.43 ++)  Depth: 30/61   00:11:10  19012MN, tb=23
10...Ng6 
  +/= (0.35 ++)  Depth: 30/61   00:11:51  20173MN, tb=23
10...Ng6 11.exf6 Qxf6 12.e5 Qg7 13.Nbd2 d5 14.exd6 Bh3 15.Ne1 Nh4 16.g3 cxd6 17.Qh5 Bxf1 18.Kxf1 Ng6 19.Ne4 f5 20.Ng5 Ne5 21.Rd1 Rf6 22.Nd3 Rh6 23.Nxe5 Rxh5 24.Bb3+ Kh8 25.Nef7+ Qxf7 26.Nxf7+ Kg7 27.Kg2 f4 28.gxf4 
  = (0.23)  Depth: 30/61   00:12:30  21290MN, tb=23
10...Ng6 11.exf6 Qxf6 12.e5 Qg7 13.Nbd2 d5 14.exd6 Bh3 15.Ne1 Nh4 16.g3 cxd6 17.Qh5 Bxf1 18.Kxf1 Ng6 19.Ne4 f5 20.Ng5 Ne5 21.Rd1 Rf6 22.Nd3 Rh6 23.Nxe5 Rxh5 24.Bb3+ Kh8 25.Nef7+ Qxf7 26.Nxf7+ Kg7 27.Kg2 f4 28.gxf4 
  = (0.23)  Depth: 30/61   00:12:30  21291MN, tb=23

[peter]

Hi!
One problem remained with the update at my Intel 12core and Win7x64Pro with Fritz14 as well as with Shredder13- GUI, the hash taken by the engine isn't as big as set by the GUI- command.

Independently from the amount "given", H5 takes 256Mb only, as for what Windows Task Mananger shows. That wasn't so as for a short time after installation of H5, then the problem arose, after installation of 5.01 it disappeared again for a while to be here again now.

That doesn't depend on how much RAM- allocation has happened already, it's the same with newly started Windows, just the same with 32G large pages usable according to lp- test of H5.exe,

[Houdini]

Hi!
One problem remained with the update at my Intel 12core and Win7x64Pro with Fritz14 as well as with Shredder13- GUI, the hash taken by the engine isn't as big as set by the GUI- command.

Independently from the amount "given", H5 takes 256Mb only, as for what Windows Task Mananger shows. That wasn't so as for a short time after installation of H5, then the problem arose, after installation of 5.01 it disappeared again for a while to be here again now.

That doesn't depend on how much RAM- allocation has happened already, it's the same with newly started Windows, just the same with 32G large pages usable according to lp- test of H5.exe,

Are you using Large Pages?
Windows reports Large Memory Page usage differently, IIRC it's not shown under the "Working Set" but only as committed memory.
Just look at the engine output to see how much hash is used, Houdini shows the amount and tells you whether Large Pages have been used successfully.

[peter]

Windows reports Large Memory Page usage differently, IIRC it's not shown under the "Working Set" but only as committed memory.

Thanks for the prompt answer! Had not chosen committed memory to be shown in Taskmanager-menu,