Testing LazySMP

[Graham Banks]

I sent you a private message.

Thank you for your message. You can download LazySMP 4.0 from the following link:

If it's a genuine and original effort on your part, testers would take an interest in your engine..

I released the latest version of the program and fixed the previous bugs. If you test it, I will continue to develop it, otherwise this is the last version of the LazySMP and I never develop it again.

According to Task Manager, your engine is only using 5.7MB hash. Is there a way to set it to use 256MB?

[LazySMP]

According to Task Manager, your engine is only using 5.7MB hash. Is there a way to set it to use 256MB?

LazySMP don't use a Transposition Table. Why do you want to set the hash to 256MB? this is not a bug.

Well, that is not too bad, for an engine without TT. It cannot be that buggy. The high depth observed by Ed suggests it does a lot of forward pruning.

[chessica]

I sent you a private message.

Thank you for your message. You can download LazySMP 4.0 from the following link:

If it's a genuine and original effort on your part, testers would take an interest in your engine..

I released the latest version of the program and fixed the previous bugs. If you test it, I will continue to develop it, otherwise this is the last version of the LazySMP and I never develop it again.

According to Task Manager, your engine is only using 5.7MB hash. Is there a way to set it to use 256MB?

sorry the engine is not working for me

[Brunetti]

I released the latest version of the program and fixed the previous bugs.

Code: Select all

Games Completed = 4000 of 4000 (Avg game length = 13.532 sec)
Settings = Gauntlet/128MB/4000ms+100ms/M 600cp for 4 moves, D 160 moves/EPD:c:\Chess\LB\UBC\BRUCE.epd(32000)
Time = 3717 sec elapsed, 0 sec remaining
 1.  LazySMP 4.0 popcnt       1534.5/4000	1299-2230-471 (38.36%)	(loss: m=199 t=297 i=0 a=1734)	(draw: r=188 i=115 f=102 s=4 a=62)	(avg: tpm=161.9 d=13.32 nps=3936367)
 2.  BikJump v2.01 (64-bit)   737.0/1000	678-204-118 (73.70%)	(loss: m=4 t=0 i=0 a=200)	(draw: r=45 i=42 f=11 s=2 a=18)	(avg: tpm=136.5 d=9.10 nps=0)
 3.  Adam 3.3                 681.0/1000	618-256-126 (68.10%)	(loss: m=2 t=0 i=0 a=254)	(draw: r=53 i=30 f=31 s=0 a=12)	(avg: tpm=105.3 d=7.26 nps=4801126)
 4.  Aristarch 4.51           455.5/1000	415-504-81 (45.55%)	(loss: m=17 t=0 i=0 a=487)	(draw: r=22 i=21 f=26 s=1 a=11)	(avg: tpm=44.1 d=6.24 nps=2434817)
 5.  Apollo Release 1.2.1     592.0/1000	519-335-146 (59.20%)	(loss: m=10 t=0 i=0 a=325)	(draw: r=68 i=22 f=34 s=1 a=21)	(avg: tpm=132.6 d=9.52 nps=659428)

Estimate 2050. In 7% of cases, it lost on time without even making a move; it's possible the time control is too fast.

Alex

[chrisw]

I released the latest version of the program and fixed the previous bugs.

Code: Select all

Games Completed = 4000 of 4000 (Avg game length = 13.532 sec)
Settings = Gauntlet/128MB/4000ms+100ms/M 600cp for 4 moves, D 160 moves/EPD:c:\Chess\LB\UBC\BRUCE.epd(32000)
Time = 3717 sec elapsed, 0 sec remaining
 1.  LazySMP 4.0 popcnt       1534.5/4000	1299-2230-471 (38.36%)	(loss: m=199 t=297 i=0 a=1734)	(draw: r=188 i=115 f=102 s=4 a=62)	(avg: tpm=161.9 d=13.32 nps=3936367)
 2.  BikJump v2.01 (64-bit)   737.0/1000	678-204-118 (73.70%)	(loss: m=4 t=0 i=0 a=200)	(draw: r=45 i=42 f=11 s=2 a=18)	(avg: tpm=136.5 d=9.10 nps=0)
 3.  Adam 3.3                 681.0/1000	618-256-126 (68.10%)	(loss: m=2 t=0 i=0 a=254)	(draw: r=53 i=30 f=31 s=0 a=12)	(avg: tpm=105.3 d=7.26 nps=4801126)
 4.  Aristarch 4.51           455.5/1000	415-504-81 (45.55%)	(loss: m=17 t=0 i=0 a=487)	(draw: r=22 i=21 f=26 s=1 a=11)	(avg: tpm=44.1 d=6.24 nps=2434817)
 5.  Apollo Release 1.2.1     592.0/1000	519-335-146 (59.20%)	(loss: m=10 t=0 i=0 a=325)	(draw: r=68 i=22 f=34 s=1 a=21)	(avg: tpm=132.6 d=9.52 nps=659428)

Estimate 2050. In 7% of cases, it lost on time without even making a move; it's possible the time control is too fast.

Alex

I got a lot of UCI problems to go away when the engine switched from polling the keyboard every N nodes or T milliseconds, to putting the keyboard poller in its own thread and the engine in another. That way the engine/poller combination can act without random polling delays (mostly). Possibly this chess engine is not using a polling thread?

[LazySMP]

I got a lot of UCI problems to go away when the engine switched from polling the keyboard every N nodes or T milliseconds, to putting the keyboard poller in its own thread and the engine in another. That way the engine/poller combination can act without random polling delays (mostly). Possibly this chess engine is not using a polling thread?

Thank you for your attention into the matter.

Code: Select all

void loop(std::function<void(void)> uci, unsigned int interval)
{
    std::thread([uci, interval]()
        {
            while (true)
            {
                auto x = std::chrono::steady_clock::now() + std::chrono::milliseconds(interval);
                uci();
                std::this_thread::sleep_until(x);
            }
        }).detach();
}

[LazySMP]

Code: Select all

Games Completed = 4000 of 4000 (Avg game length = 13.532 sec)
Settings = Gauntlet/128MB/4000ms+100ms/M 600cp for 4 moves, D 160 moves/EPD:c:\Chess\LB\UBC\BRUCE.epd(32000)
Time = 3717 sec elapsed, 0 sec remaining
 1.  LazySMP 4.0 popcnt       1534.5/4000	1299-2230-471 (38.36%)	(loss: m=199 t=297 i=0 a=1734)	(draw: r=188 i=115 f=102 s=4 a=62)	(avg: tpm=161.9 d=13.32 nps=3936367)
 2.  BikJump v2.01 (64-bit)   737.0/1000	678-204-118 (73.70%)	(loss: m=4 t=0 i=0 a=200)	(draw: r=45 i=42 f=11 s=2 a=18)	(avg: tpm=136.5 d=9.10 nps=0)
 3.  Adam 3.3                 681.0/1000	618-256-126 (68.10%)	(loss: m=2 t=0 i=0 a=254)	(draw: r=53 i=30 f=31 s=0 a=12)	(avg: tpm=105.3 d=7.26 nps=4801126)
 4.  Aristarch 4.51           455.5/1000	415-504-81 (45.55%)	(loss: m=17 t=0 i=0 a=487)	(draw: r=22 i=21 f=26 s=1 a=11)	(avg: tpm=44.1 d=6.24 nps=2434817)
 5.  Apollo Release 1.2.1     592.0/1000	519-335-146 (59.20%)	(loss: m=10 t=0 i=0 a=325)	(draw: r=68 i=22 f=34 s=1 a=21)	(avg: tpm=132.6 d=9.52 nps=659428)

Estimate 2050. In 7% of cases, it lost on time without even making a move; it's possible the time control is too fast.

Alex

Thank you for your testing. I also believe that the time control is too fast. Considering that in 93% of cases time is not lost, we can safely say that the main engine problems have been solved. Of course, as I mentioned before, this is the last version of my engine and I won't update it anymore.

I may try SPCC chess, CEGT, SSDF, ipman chess and many others. CCRL are not the only ones.

[JVMerlino]

Considering that in 93% of cases time is not lost, we can safely say that the main engine problems have been solved.

[Graham Banks]

I may try SPCC chess, CEGT, SSDF, ipman chess and many others. CCRL are not the only ones.

Good idea.

[a_node_uncut]

Hijacking this thread a bit. But what's the usual procedure to submit an engine on the rating lists? I'm planning to release my engine in a few weeks but I'm not at all familiar with how the rating lists work.