Release youtube and java applets


Dark Resistor

Still Fresh
Joined
Oct 21, 2008
Messages
1
How will pandora handle youtube without Adobe Flash Player?? maybe using the same codec for youtube (or wathever) as iphone who has not suport for flash.
And will pandora be able to run java applets++, i am going to use pandora at the university with some java applets.

An exaple of what kind of applet there: http://www.falstad.com/circuit/

Thanks, and exuse my bad english.
 
The browser will be able to recognize sites with flash video, and know how to isolate these streams (for known sites) so they can be opened in a media player.
 
I will wait for Maemo. I'm sure it will be ported. Tor said, that he will do it partly. Someone from Nokia said, that they wouldn't be restrictive towards someone who tries to make it work on non Nokia devices. If Maemo works, java and flash will work.
 
There's no Java in maemo, so getting maemo running won't handle that problem, unfortunately. There are some 3party efforts going on for Java and Java applets, and maybe something will happen with v5 of maemo, but the current state is: No java.
 
Tor said:
There's no Java in maemo, so getting maemo running won't handle that problem, unfortunately. There are some 3party efforts going on for Java and Java applets, and maybe something will happen with v5 of maemo, but the current state is: No java.
Sorry, I mixed that with Java in the browser :confused:
 
Well, it's the same thing.. no java in the browser unless there's java applet support. We're not there yet, but hopefully we'll get there eventually.
 
Javascript is a different story - that'll work just fine. Netscape (who, as far as I know, started the whole Javascript thingy) gave this scripting language a terribly misleading and confusing name.
 
Hey guys ... sorry this isn't clear for me. Javascript should surely run, agreed assuming an interpreter is most likely already integrated in the browser be it firefox, webkit, etc. But it's not clear to me why full-blown Java wouldn't run? Certainly it is ported to ARM (most ARM11's have java acceleration in hardware). Surely there must be a JVM for OMAP3?

I should think that, given a JVM, Dark Resistor's app should run just fine. What am I missing?

e.g.: http://www.lalena.com/games/

Yek
 
Well, some JVM is likely. It will not be anywhere near as fast as Java on the PC, though, due to lack of JIT. It seems unclear whether we will actually be able to use Jazelle at all, or whether it would help even if we could. But yes, that circuit application should run, or at least there are no theoretical objections.
 
As long as we're talking about Java, I've found this interesting "Web Start" thing that allows .. something cool to happen. JOGL happened, to say the least, without my explicit installation of it.

Is there going to be a Web Start setup for Pandora?
 
The built-in java support in the ARM CPU needs licensing or you won't be able to use it, afaik. In practice it means it's not documented. So it's not muched used either, at least not on open platforms.
 
Tor said:
The built-in java support in the ARM CPU needs licensing or you won't be able to use it, afaik. In practice it means it's not documented. So it's not muched used either, at least not on open platforms.
I think you mean "won't be able to officially use it". Open Pandora Ltd. can't provide support, that doesn't mean someone else won't reverse engineer it, like with the DSP driver.

Not saying it will happen (for either), but "won't" (or "will not") is a strong statement.
 
First off, Java has nothing at all to do with Javascript. Nothing. Call Javascript by it's correct name btw: ECMAScript (Well, Javascript is the predecessor to ECMAScript etc; standard specs are complicated).
So yeah, pretty much all browsers that run in a GUI have Javascript support nowadays, it isn't too complicated.

Then, Java is a whole different story. You'll need 4 things for full Java support: A JVM, the Java core libraries, a Java browser plugin and the javaws executable (Java WebStart). If we have a JVM, the rest is trivial since most JVMs bundle core libs, and if they don't we can just take them from somewhere else since Java is platform independent. The Java-plugin is also platform-independent. javaws is just a launcher executable that takes a jnlp-file and downloads the correct JARs for it, then launches a JAR, not really rocket science.

We might get some problem with Java3D though, since it isn't ported to OGLES afaik, and javaws won't work if the thing you download contains platform-dependent libs like OpenGL-stuff etc (javaws downloads the correct binaries for the platform, but if there are none compiled it won't work); you'd have to port those platform-dependent libs.

Finally, interpreted Java will certainly be fast on the Pandora; at least 8 times faster than interpreted Python if I recall my benchmarks correctly. If we get any kind of JIT, it will be even faster. And Jazelle, HotSpot etc can bring Java close to (but not quite at, of course) C performance, but we'll need some benefactors or serious hacks for that.
 
dflemstr said:
Finally, interpreted Java will certainly be fast on the Pandora; at least 8 times faster than interpreted Python if I recall my benchmarks correctly. If we get any kind of JIT, it will be even faster. And Jazelle, HotSpot etc can bring Java close to (but not quite at, of course) C performance, but we'll need some benefactors or serious hacks for that.
I already spoke to Sun regarding their JVM. Unfortunately, there's no discount for open source projects. :|

The Pandora has a lot of processing power, so although something like JamVM isn't ideal, it'll still be fast enough for simple apps. What'd I'd love to see is JamVM utilizing the Jazelle bounds checker for arrays - if that's even possible. I imagine that'd be yet another small speedup, if someone could figure it out. ^_^
 
Kramy said:
I already spoke to Sun regarding their JVM. Unfortunately, there's no discount for open source projects. :|

The Pandora has a lot of processing power, so although something like JamVM isn't ideal, it'll still be fast enough for simple apps. What'd I'd love to see is JamVM utilizing the Jazelle bounds checker for arrays - if that's even possible. I imagine that'd be yet another small speedup, if someone could figure it out. ^_^
You spoke to Sun, for real? :D

Anyways, the largest bottleneck in a JVM has always been that you can't store objects on "the stack" afaik. It gets really annoying when you do heavy math computations with vector classes and you can't have immutable vector types because of this bottleneck, forcing you to create mutable vector types, effectively just exposing two float variables with wrapper methods. Very annoying. I'd like to use code that creates, like, 30 000 objects per second, which isn't possible without the optimizations that have been done in the big JVMs.

By the way: will we get good float performance in the JVM packaged with the Pandora? Can it be compiled to use the DSP/SIMD extensiosn or what have you? Don't know what could be vectorized in a JVM, really, but would be awesome if we were able to use floats like we do on all other platforms, for heavy math.
 
dflemstr said:
Kramy said:
I already spoke to Sun regarding their JVM. Unfortunately, there's no discount for open source projects. :|
You spoke to Sun, for real? :D
Yes, I was emailing them constantly in January & February with questions, then forwarding the important responses to project members like Chip.

Java SE Embedded 1.6 for ARM isn't the most optimized JRE/JVM out there, but it'd certainly beats open source projects. It might even be close in performance to C, but that's not saying much.

I recall from gp32x, some board members were complaining about one gcc compile being slower than another identical compile on the GP2X. Other board members said assembly often boosted performance 50+%, which is a far bigger gain than you'd expect on x86. Java on x86 is very efficient, but it sounds like it's not that hard to beat C on some of these ARM processors. :wink:

dflemstr said:
Anyways, the largest bottleneck in a JVM has always been that you can't store objects on "the stack" afaik. It gets really annoying when you do heavy math computations with vector classes and you can't have immutable vector types because of this bottleneck, forcing you to create mutable vector types, effectively just exposing two float variables with wrapper methods. Very annoying. I'd like to use code that creates, like, 30 000 objects per second, which isn't possible without the optimizations that have been done in the big JVMs.
I don't have a very good understanding of that low level stuff, but I suspect since Java 1.6 for ARM was developed for SoCs with more memory(like the OMAP 3530), it must have heavier optimizations than earlier versions of java.

dflemstr said:
By the way: will we get good float performance in the JVM packaged with the Pandora? Can it be compiled to use the DSP/SIMD extensiosn or what have you? Don't know what could be vectorized in a JVM, really, but would be awesome if we were able to use floats like we do on all other platforms, for heavy math.
Hell if I know! :lol: Sun's JVM would probably use the weak FPU rather than NEON - but if the Pandora were actually using that JVM, you could always try begging that they optimize it more. ;)

Honestly, NEON optimizations are more likely in a JVM like JamVM, where any one of us can recompile it.

I doubt any JVM will use the DSP without some hefty redesigning.
 
I really dont see the fuss over youtube on the pandora...

just go to keepvid.com or any of the other BAJILLION youtube downloader websites out there and just watch the mp4/swf on VLC or mplayer or whatever else you like. a lot of other sites have the same deal.

the problem I was adobe for is mainly things like (the coincidentally named) pandora.com for some good tunes, certain websites that are flash based, and of course flash games. youtube isn't a specific enough problem for me to get flustered about it.
 
Would be cool to be able to play some java-based games like Runescape or Wurm Online on the Pandora, or to be able to develop games with it... Considering JavaFx would make the games really nice.
BTW if NEON optimizations are so good, then JamVM optimized for NEON would be a better bet, I'd say.
 
On Firefox, get the GreaseMonkey extension and then download any of the multitudes of scripts that directly embed the flv or mp4 file into the YouTube page. Problem pseudosolved.

Alternatively, wait for FF3.5 and YouTube to implement HTML5. Then you don't have to worry about flash at all. :D
 
Back
Top