On 02/19/2011 03:00 PM, f...@libero.it wrote: Hi,
>> ----Messaggio originale---- >> Da: joost.t.h...@planet.nl >> Data: 19/02/2011 13.37 >> A:<scid-users@lists.sourceforge.net> >> Ogg: Re: [Scid-users] I hate to wait [PATCH] >> >> On 02/19/2011 01:33 PM, f...@libero.it wrote: >> >> Hi, >> >> Yes, I posted about these stockfish anomalies before. >> >> And have some ideas in handling this; although I also hope that next sf >> releases will somehow start to behave... >> > Sorry, i didn't read your posts, i will google them. NP :-) They are mostly connected to the rework of the analysis module I did in 2K10. > Can you test if my patch breaks anything? I certainly will. First things first, unfortunately (and if you know what I mean; time has not really been on my side, lately). > Do you know which is the bottleneck? > formatPV maybe? That might one, indeed. I alread did some experiments skipping it, but it leads to "g1f3" formatted moves in the engine window (maybe not too bad) which in turn breaks the best variation board display functionality (which /is/ bad). Furthermore: It did not really help in improving the performance. I expect more results from a healthier translation mechanism of engine-formatted moves into scid-formatted moves. Currently, each time this is needed 1) a temporary game is created 2) to which the new engine-formatted moves are appended, 3) followed by a roll back to the original position 4) and a re-read of the scid-formatted moves from the game 5) and a destruction of the temporary game. I admit that the translation needed is not entirely trivial to accomplish (what piece is moving?), but this cannot be the most elegant approach (which is implemented in tcl, by the way). But: The fact that the receiving scid-end of the communication channel with the engine is implemented in tcl is certainly a big issue. Talking in producer/consumer terms: Since the engine (the producer) and pipe (OS) itself are implemented in native code, there is fundamentally no way that tcl as a receiver will be able to cope with all the data entering (if the engine starts spending all its energy in spitting data - and this is what SF does, especially if it sees a mating situation coming up). As a consumer you really deserve at least the same CPU attention as is given to the producer. If not, <fill in yourself>. Cheers, Joost. > Bye, > Fulvio > ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb _______________________________________________ Scid-users mailing list Scid-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scid-users
