RE: FLASH: Flash 5/6 aka DirectFlash! Some suggestions

[Damien Morton]

>

From: ownerchinwag [dot] com On Behalf Of Branden Hall

>

> I see your points, however, if you look at what Flash 5 gives you, an XML

> parser, socketed connection ability, color manipulation, sound control,

> etc... all on top of a rendering engine that handles not only vectors, but

> also bitmaps, including transparency, 200K is pretty crazy.


All good things, and getting it into a 200K plugin, is, as you say, a

remarkable acheivement. I am very much looking forward to working with Flash

5 and exploring it capabilities to the fullest.


> Also, adding a video system to Flash would add a HELL of a lot more than

> 200K to the size..Just look at the size of the quicktime or even vivo

> plugins.  Besides, MM is trying to get Flash everywhere, and in PDAs,

> phones, and web-top boxes the law of the land is the K, period.

> And having

> a combined plugin thing like Director does would not only add

> quite a bit to

> the plugin size, but it would also make the plugin less cross-platform,

> which is not acceptable.


Flash on PDAs, phones, etc, is a good thing. I would'nt necessarily suggest

a combined plug-in based architecture. In fact, a unified platform is one of

the wonderfull things about Flash. I would in fact suggest incorporating as

well-balanced set of capabilities into a single unified player platform.


On the other hand, these PDAs, phones and set-top boxen will undoubtedly

have or require video capabilities. Just because they are small, doesnt also

mean they arent capable. I have implemented simple video capabilities on

machines far less powerfull than the least powerfull PDA you can get a hold

of today, let alone the PDA or cellphone of one or two years from now. That

video player was tiny, used almost no RAM, and operated on a 1MHz 8-bit

machine. Compression was bad, and resolution was minscule, but it was

possible. Software technology has improved substantially since then, as has

the hardware.


I was reading today of a linux powered wrist-watch. It used a version of the

ARM processor equivalent to a 100Mhz Pentium, had 8MB RAM, 8 MB Flash RAM,

and would run for 4 days between recharges. This was a development

prototype. I believe that Palm is going to be using ARM chips in their next

range of handhelds. Lets assume that these kinds of capabilities are the

baseline for cellphones and PDAs that might incorporate Flash players.


> As for the math... the Flash renderer, is, unless I am mistaken, scanline

> based, hence trig and all that kind of stuff isn't needed by the renderer.

> Thus, there is nothing to "expose" to actionscript, sorry to say.  Then,

> when it comes to newtons method, etc, that kind of stuff will have to be

> done *somewhere*, might as well be in the SWF its self, that way

> the plugin

> can worry about more troublesome things.


Scan-line renderer or not, the trig is in there. When you

SetProperty("myObject", _rotation, 2.5 * x), there is trig going on behind

the scenes turning that into a matrix. I can pretty much guarantee you that

that trig is implemented in assembler, in the plug-in.


> If you compare what Unreal has code wise to do its software rendering,

> including GL, and directX, its pretty nuts how *much* code is

> required to do

> that kind of stuff fast.  Not only that, but a huge amount of the

> computations inside of those kind of games are pre-processed when the maps

> are made.  Also, bitmap calcuations are, but nature a LOT easier

> than vector

> calculations... just yank open a big ass PDF and you will see what I mean.


There is a great deal of commonality between a 3d texture renderer and a

vector graphics renderer.



> -= Branden J. Hall

> -= Senior Interactive Developer/Instructor

> -= Fig Leaf Software - "We've got you covered!"

>

>

> -----Original Message-----

>

From: ownerchinwag [dot] com [ownerchinwag [dot] com]On">mailto:ownerchinwag [dot] com]On Behalf Of Damien

> Morton

> Sent: Tuesday, August 08, 2000 4:10 PM

>

To: flasherchinwag [dot] com

> Subject: RE: FLASH: Flash 5/6 aka DirectFlash! Some suggestions

>

>

> Not being able to see your code, Branden, Ive had to make a few

> assumptions,

> but Id love to hear how your code performs if optimised.

>

> Im not sure that the use of floating points would be slower than integers,

> given the fundamental overhead of running an interpreter. At any

> rate, there

> should be virtually no difference.

>

> Exporting Newtons method where a square root is called doesnt

> make sense to

> me. When outputting Flash 4, this is no doubt necessary, but surely not in

> the case in Flash 5. It would better in the plug-in itself, and a

> whole slew

> of math/trig functions are probably already in there for the actual vector

> drawing routines anyway, so it should 'simply' be a matter of

> exposing them

> to actionscript.

>

> My basic point is that fast code enables a realm of experiences that go

> beyond scripted animations and into creating actual behaviours.

> Your version

> of SodaPlay, Branden, is a great example of this, and I think that further

> enabling this kind of thing will enhance Flash as a platform.

>

> The smallness of the plug-in comes up again and again, and I wonder if the

> balance between plug-in size and plug-in capabilities has been

> well-struck.

> A 200K plug-in is a good thing (tm). Is it better than a 400K plug-in that

> has some combination of much faster code execution and rendering, video

> sprites and images compressed 6 times better than JPEG? I wonder.

>

> Ive been playing Unreal Tournament a bit lately, in software

> rendering mode.

> It runs at over 30 frames per second, and is doing a _whole_ lot more work

> than a 1-spring 2-body dynamic system. At a guess, its doing somewhere

> between a thousand times and a million times more work. Why does

> it only use

> 2 times the CPU of the sprung.swf demo?

>

> Im not disparaging the Flash dev team. Flash is a cool tech, and

> obviously,

> in getting a product out the door, there are limits to be reached and

> compromises to be made. Keep up the good work and keep pushing

> the envelope.

>

> ------------------

> Damien Morton

> Senior Programmer

> Dennis Interactive

>

> > -----Original Message-----

>

> From: ownerchinwag [dot] com [ownerchinwag [dot] com]On">mailto:ownerchinwag [dot] com]On Behalf Of Branden

> > Hall

> >

> > Actually, that system does some just-plain-complex(tm) math...

> > lots of fun

> > calculus stuff (integration via Euler's method), and I am kinda

> > sloppy with

> > my floating points...  I'm sure if I went in and did a few

> assembler-style

> > tricks to keep the code from ever using floats, it would speed up quite

> > nicely. The simple fact of the matter is that you have a 200K

> > plugin that is

> > able to run some *serious* code pretty damn fast.  More than

> > likely, F5 took

> > a sort-of RISC approach to be able to do all it does with that

> > little...  in

> > other words, the F5 plugin is probably optimized up to wazoo for

> > doing basic

> > operations, and have all complex operations exsist as

> combinations of the

> > basic ones. For example, when you export say, a square root call Flash

> > probably includes the math needed to do a newtons method to get

> > the sqrt in

> > the SWF...  Otherwise I don't see how in the hell the plugin is still so

> > damn small.

> >

> > Anyhoo... you have to look at all of the factors involved with

> > why flash is

> > what it is...

> > Besides that, I honestly do not see video coming to Flash for a

> > while... it

> > would just bloat the plugin too much.. and once browsers get their act

> > together they should all be able to do what I can do in IE... put a

> > quicktime above my flash movie.  Finally, as for JPG2000... I agree

> > totally... wavelets are the future!

> >

> > -= Branden J. Hall

> > -= Senior Interactive Developer/Instructor

> > -= Fig Leaf Software - "We've got you covered!"

> >

> >

> > -----Original Message-----

>

> From: ownerchinwag [dot] com [ownerchinwag [dot] com]On">mailto:ownerchinwag [dot] com]On Behalf Of Damien

> > Morton

> >

> > Heres a couple of suggestions for Flash 6:

> >

> > Ive been looking at Branden Hall's sprung.swf Flash 5 demo. A

> 1-spring, 2

> > body system consumes 50-60% of my 600MHz CPU. Now Branden is a

> > smart guy, so

> > I cant really imagine that hes implemented that demo too

> shabbily. If this

> > is anything like what we can expect, then the performance of

> ActionScript5

> > leaves a lot to be desired. Suggestion: Make ActionScript byte code

> > interpretation much much faster. For games, for anything

> > requiring a modicum

> > of physical simulation or computation, a faster interpreter is needed.

> <snip>

>


>



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 Full flasher archive now available online at:

 http://www.chinwag.com/flasher/archive.shtml

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

To unsubscribe or change your list settings go to

http://www.chinwag.com/flasher or email

helpchinwag [dot] com