[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [computer-go] Pattern Matcher



Mark,

If it helps, it is just another data point I used to convince me to keep plugging along in Java and exploiting distributed parallelism as a way to elevate large grain activities.

My current work is in GAs and ANNs. And the numbers I am getting sure seem to be similar to hand optimized C. I am doing huge amounts of floating point calcs and blowing the CPU caches all the time. I am still getting very desirable numbers, though...across 5 boxes...with very little manual effort.

Perhaps Vincent isn't a Java expert and does not know how to leverage particular types of optimization that would be Java specific. If he tried to do C-like optimizations to Java, he was likely defeating optimizations the Sun Hotspot JVM will make a run time. Worse, if he is comparing J2EE EJBs to C, I could see him thinking in terms of multiples. It could be he just lacks low level Java specific experience. I have heard it said, "It takes at least two years to be a master at something." I think Java optimization as distinct from C/C++ optimization is a "something".

Of course, given the way he communicates here, I would be very surprised if Vincent were to admit his "lacking" around the master level of Java optimization. I think he might fear it would make him look less "intelligent" than he would like for all of us to think he is. Besides, how could anything but assembler *possibly* be superior in any way to C? C (and C++ by association) is the language of supreme intelligence! All other languages languish in utter and abysmal design, implementation and performance! Only morons would consider otherwise.

BTW, I am hereby classifying myself as a moron of the Java breed.


Jim


Mark Boon wrote:


-----Original Message-----
From: computer-go-bounces@xxxxxxxxxxxxxxxxx
[mailto:computer-go-bounces@xxxxxxxxxxxxxxxxx]On Behalf Of Don Dailey
Sent: Monday, November 08, 2004 20:48
To: computer-go@xxxxxxxxxxxxxxxxx
Cc: computer-go@xxxxxxxxxxxxxxxxx; computer-go@xxxxxxxxxxxxxxxxx
Subject: Re: [computer-go] Pattern Matcher



Vincent,

John Tromp wrote a game benchmark that made a believer out of me.
Java appears to be only 30% slower or even better. There are probably
applications where it's much slower, but it doesn't seem to be board
games.

Rather than have him translate a Java program to C to prove it can be 3
times faster, it would be more practical to give Vincent a pointer to this
benchmark. I'm not familiar with it, so could you post a link? I'm rather
curious myself.

I would rather have thought my word about translating Goliath would have
been sufficient ;-) But I suppose we have a rather sceptical crowd here,
which I suppose is a good thing.

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/


_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/