Physics performance challenge to J.Kesselman...

Game Mechanics
#1

Open letter and challenge to Mr Kesselman, in response to a debate in his latest blog http://weblogs.java.net/pub/wlg/1198

Jeff,

As you appear to be so confident that Java is suitable for multiplatform games development, and that it will perform as well as native code, I want you to put your money where your mouth is, so to speak. You have told me that a physics engine will perform just as fast in a Java VM as it will in native code, so I propose you financially sponsor a competition to build a simple Java physics simulation that will match or outpeform a native code version, using exactly the same simulation algorithm, collision detection algorithm and maths. If you or any other entrant can match (within 0.5%) or beat the speed of my native code version, the fastest version wins, otherwise I win. The simulation will be set up so that it behaves like a real game loop and not like a micro-benchmark.

Now given that it will be just my code and time vs the whole Java games development community, Sun’s resources, experts on Java optimisation like yourself, experts you quote like Professor Kendall, you shouldn’t have any problems winning your argument, should you?

Now I don’t want to leave this unanswered, so if you fail to accept I (and anyone else reading this) will have to assume you don’t believe the views you have published. The rules will be negotiated to ensure a real and fair test scenario for multiplatform games development in Java, so that will not count as an excuse to decline my challenge.

I await your response,

Andrew J Larder.

#2

Jeff’s blog is http://weblogs.java.net/pub/wlg/1198.

#3

Thats not the assumption I’d make, I’d assume he’d made an assumption about you. :wink:

Kev

#4

Hey, I’m not the one who 18 months ago was promising that a PS2 VM announcement was ‘around the corner’ ::slight_smile:

#5

Hmm PS2 or physics, which are we talking about?

Actually thats sort of rhetorical because it doesnt matter either way. As I already explained to Mr. Larder in private, and as you folks all already know, we in the GTG (actually it wasn’t the GTG yet) were at least as disappointed as you folks about the fact that the PS2 VM talks failed to produce results.

We haven’t forgotten the importance of this. It contineus to be a hot button on our list, and we continue to persue avenues to get you guys VMs on consoles. As I told him. you guys will be the first to be notified if anything really solid happens but I’m not going to get your hopes up again too early. Some of us do learn from our experiences.

I also think Mr. Lrder had a good idea. But sincee’s the one who cares he should put up the prize, at least 4 dgits worth. If he does that I’m sure you guys would be happyto compete, wouldn’t you? (Sicne we are talking an individual that prize better be in verifiable escrow account. and the contest rules clearly and legally stated.)

Otherwise we’d rather use our meager resources to reward you guys for writing good games in things like the current game-programming contest.

And with that, I’m plonking this troll. Does “Land of the Trolls” still exist? If so this thread should clearly bne moved.

#6

Sorry to hear that, but also glad - I’m really busy at the moment so I could do without a diversion. The real reason you wouldn’t touch this with a barge-pole is that you know full well the Mac VM is still about 7x slower than C++. Of course you wouldn’t want to make that too public, would you???

You’re right though Jeff, I do care, because I like many other developers would love to dump C++ and use Java (believe it or not John Carmack takes an occasional rain check in this direction, but I don’t recall Doom 3 being a pure Java game).

I guess next year Jeff will be telling everyone about the great Quake23 clone FullSail have, and of course nobody will ever see it.

I wish I could say it was a pleasure Jeff.

Ta ta.

#7

You might want to take a break from your ‘Real Game’ and hit up a counselor! You’ve got some deep rooted anger management issues you could work on.

#8

Dude your right. I’m taking all of this far too seriously!

Jeff’s (I won’t resort to Mr Kesselman…) role is to push the positive points of Java and sweep the negatives under the carpet … he wouldn’t be doing his job if he didn’t.

If only Java were that good then I’d have time to see that counsellor :wink:

#9

Wow! Now that was a good troll! Warts and all. Have you considered politics?

Kev

#10

Hasn’t Tony got a space since Alastair Campbell quit? Perhaps you could put a good word in for me :wink:

Seriously though - why if someone asks sound technical questions and then gets frustrated when no hard evidence is delivered, do they get labelled a ‘troll’? Hearsay holds no water with me, only runnable code. If Jeff wants me to believe that physics or whatever in Java is as good or better performing than C++ then he can send me a demo of exact equivalence in both languages, including source code. That way I can test it on Win32 and MacOS X. And if he wants me to think that he can cover the other 80% of the games market he can send me an evaluation VM for the PS2, XBox and GameCube. Renderware, NDL and Havok are quite willing to do that today.

#11

Probably cause when anyone gets frustrated they type things like:

on a “Java Gaming” forum. Wasn’t intended to offense but surely you can see how that could only be considered a troll? Infact, a mighty impressive troll given the context.

Kev

#12

[quote]why if someone asks sound technical questions and then gets frustrated when no hard evidence is delivered, do they get labelled a ‘troll’? Hearsay holds no water with me, only runnable code. If Jeff wants me to believe that physics or whatever in Java is as good or better performing than C++ then he can send me a demo of exact equivalence in both languages, including source code.
[/quote]
Since you’re new to this forum, use the search function. You’ll find dozens of threads which discuss exactly what you’re looking for (provided you really are looking for this). Use keywords like “benchmark”, “how not to interpret micro benchmarks”, “statistics and other lies”, and such, hehe. Jeff wrote many excellent articles in these threads, and usually he proves what he says. No reason for him to repeat himself X times I think.
You could also search the SUN site for some of Jeff’s interesting speeches on the Java performance topic, and don’t miss his book “Java Platform Performance: Strategies and Tactics” (published online on the SUN site).

#13

I’ll stump up some money. Let’s agree on some rules, specify an escrow company, etc.

It might also be a good idea for all these hot shots to contribute to the prize fund in some way as well. Put your money where your mouth is.

#14

The burden of proof lies with the one making the positive assertion. If Jeff says Java is good enough, Jeff must provide evidence. That’s how real debates work.

If Jeff had these statistics all along, why didn’t he post links on his weblog? Not everyone reading the weblog would know that this relatively insignificant forum even existed. And like I just said, even if people knew the forum existed, it is still Jeff’s responsibility to provide evidence or references. This should be placed on the weblog discussion, where people will actually SEE it.

And by the way, where is this mythical FPS that supposedly runs as fast as Quake 2?

#15

Excellent, so how much? Who are you? etc…

Kev

#16

[quote]Probably cause when anyone gets frustrated they type things like:

on a “Java Gaming” forum. Wasn’t intended to offense but surely you can see how that could only be considered a troll? Infact, a mighty impressive troll given the context.

Kev
[/quote]
Jeff made a bunch of assertions regarding Java’s fitness for a particular purpose, and people questioned it. This debate was started by Jeff, not Andy. And Jeff DID ask for others’ opinions.

If Andy had come into the forum without provocation, and started this discussion, that may be trolling. But asking for evidence in response to a weblog in which comments and debate are encouraged? That isn’t trolling.

You can’t just call someone a troll just because you disagree with them.

With all the effort you have all made to post replies, I’m suprised no-one thought to just post some links to threads containing this elusive evidence that Andy is wrong.

#17

http://java.sun.com/j2se/1.5.0/download.jsp, download the sdk and make your own micro benchmark, come back and we will happily explain to you why you are wrong and java is da bomb ;D

#18

You mean like this one that was posted on Jeff’s first reply in his blog discussion:

Which is frankly where this discussion should have stayed.

And as to trolling, if I went on to a “Cookering with Bananas” forum and use a flippant comment like “If were bananas were any use for cooking” that could be nothing other than a troll. Infact, I was quite impressed with it as a troll, nothing wrong with a good one :slight_smile:

Finally, I’ve yet to be convinced either way on this argument, so to say that I’d call the poster a troll because I disagree with his argument would be absolutely unfounded.

Kev

EDIT: It does seem that the only performance statistics anyone is going to be satisified with are ones that they’ve created themselfs (otherwise they’re going to be open to interpretation).

#19

Exactly. It’s easy to say ‘we are on parity with C compilers’. As mentioned previously politicians use language like this. But who’s compiler? Easier to say you are on parity with GCC (no offense to the guys that work on GCC as you have to cover all bases with one compiler), but maybe not Intel CPP, IBM’s PPC970 compiler etc.

Oddly this comparison above was with GCC, and anyway, how dismissive is Jeff of micro-benchmarks?

The debate started by Jeff and was about why games developers are not adopting Java. I responded, and got called a ‘troll’ (and he was the very first to call me this I believe). Will Jeff apologise for this and the other patronising language he has used with me? Hmmm … perhaps Jeff should not start debates in future?

#20

[quote]Excellent, so how much? Who are you? etc…

Kev
[/quote]
I assume most of you are in the US, so you’ll be wanting dollars.

How about $500? The rest can be made up from donations by the various participants in this little competition.

I think we should go for a prize fund of about US$1000.

The money comes out of my next paycheque. Can we have some rules worked out before the end of the month?

You would also need to specify an escrow company that we can all agree on, and which makes it easy for participants to submit their money.