Quote:

I think I red something like this here: "It depends on the motivation of the guy who did it".

Quote:

For a more powerful MorphOS machine, the forthcoming Efika 8610 is the aswer. Developer programs are already open. There hasn't been any report about 8610 development board performance, as opposite to the ColdFire program, but hey, that's Gunnar von Boehn cracking it, a guy who can actually read and write assembly language, and bitch (with reason) about the questionable code that GCC compiler generates...

Quote:

Again, it's up to MorphOS Team. There's a comment on powerdeveloper.org from Matt "Neko" Sealey, of Genesi, saying that we won't "see much" of it. Alright, granted but... I can (I want to!) read this implying that it IS been worked on because, if we can't see "much", it's because there's SOMETHING...

Genesi wants to demo something pretty incredible: High definition video (1080p) on a single gigahertz CPU. It has been shown on YouTube, by a misteriuos japanese guy, I think. There has been a lot of discussion on powerdeveloper.org about wether this is impossible or not (it is a HUGE task), but what we saw was real: The MPC8610 evaluation board showing a short video clip in high definition. This processor is much more clever that some may think.
It seems to me that this demo was, indeed, a demo - pure demoscene focus: Raw hardware hitting, no operating system, no multitasking, no nothing. In other words: SEXY. It's supposed that Linux overhead would have made it impossible. What if our ace MorphOS could make it possible? Matt wants to see that.

It is possible on a Pegasos as long there is no heavy codec involved (read: no h264).
No need to wait for vapor hardware...

Quote:

Waiting for a "vapor" OS (and possibly a "vapor" player as well as I'm not sure the public MPlayer version can cope either) however is necessary, Linux won't cope (maybe, just maybe, if you can get HW accel to work). ;)


- CISC

Quote:

So, you are porting the 1080p video playback demo on your MPC8610 evaluation board, with MorphOS 2.1 beta...?

(me ducks)

Can anyone explain why MOS 2 Pup will not support Blizzard PPC from the scratch but the Bppc owners have to wait for an "unidentified time"?
I think that there are many owners of Amigas w/ Blizzard PPC cards and would like to check and buy the new Pup version -you can include me in the list-.
I know that donaitons and support are the main key points for an evolution of the supported hardware but why ignore or erase hardware that was supported on the previous Pup version -yes I am refering to BPpc-?

Quote:

Because 2.0 needs some extra work to work on PowerUP and nobody's been using/testing it in ages (afair my last real tests on PUP were done over 3 years ago). The Amiga is so slow compared to the Peg2 that I have 0 motivation to waste my free time on that.

Quote:

I wish. Though "porting" would be putting it a bit strong. I'd expect nothing less of MPlayer - The CISC Edition(tm) to just work OOTB... If a crippled piece of hardware like the Pegasos can do it, so can this evaluation board. ;)


- CISC

Hmm.. I'm not convinced that the mac mini will be faster than a peg2. If you guys could give data on this it would be interesting...


magnetic

@magnetic:

not convinced? it has faster cpu (1,3-1,5ghz), faster memory, real agp (dunno if it matters at all?). it has to be faster at least with a few %.

Sure faster ON PAPER - or on spec..

however, how good are MOS drivers to utlize these fast hw features?

lets get some real world results - would make interesting conversation, so how bout it.

|||

[ Edited by Velcro_SP 21.08.2011 - 06:37 ]

Quote:

Uhm, a faster CPU, faster bus and faster memory is just faster, period, no drivers needed. The only place a driver would have impact on speed would be gfx (well, that and network), but as the mini has a Radeon 9200 with a considerably better bus you can be sure it'll beat the peg hands down (except perhaps with certain ops with a better card) in that area as well...


- CISC

Quote:

We have to wait for the Efika 8610, which is a hard thing to do, because we all know that MorphOS already runs on the MacMini.
The MPC8610 (1 Ghz Altivec e600 core, with good memory bandwith and an awful lot of peripherals, including LCD controller) is already doing amazing things on the evaluation board, like playing back high definition (1080p) video, something that a triple-PPC-core, video-accelerated XBox 360 struggless to do...
Right, that video playback was a demo, squeezing out every CPU cycle, with no operating system behind. This is why I disturbed CISC here referring to it. He mentioned an hypothetical "MPLayer CISC edition", so perhaps he thinks he can do it. I've learnt a bit about video codecs lately, and the task of playing back MPEG4 (easy codec, of course) at 1080 lines is on the edge of impossible on this system. Yet it works...
The Efika8610 also does boring things, like running Linux, and quite zippy it was: Fedora Core 6, with 480p size DivX movie support by mplayer, Flash animation support by gnash, ekiga with USB cam demo on XGA display.
By the way, does anybody here know who is this "lonelywiddone" guy?

In the meantime, we have the announcement of AROS being released for the Efika before 18/05/2008... That hurts!

@sadddam

My iBook with a 1.33 G4 didn't score much better than the Pegasos2/G4 in MorphOS' memspeed test. Would have to dig it out of the closet and run that again for some numbers though...

My old '99 Celly with 100MHz SDRAMs still beats both machines anyway ;)

Quote:

I suppose you're talking about Simple Profile aka DivX/XviD, in which case it's certainly doable (that and MPEG2), however things get a bit more heavy once you move into AVC/H.264 domain. Granted you can speed up decoding quite a bit already by use of various trickery (but this will also degrade quality (although in many cases, not that visible)), but don't expect 1080p H.264 on any of the aforementioned hardware in this thread without serious improvements to the software decoder(s) or via hardware assist...

As for The CISC Edition, although a select few have already witnessed it in action for quite some time, I suppose not everyone knows about it. ;) Briefly explained it's simply my own little hobby-build of MPlayer (0.9x with handpicked merges from SVN version), a slightly more elaborate explanation is that it contains quite a few MorphOS-specific improvements/features, some additional AltiVec routines written by me and a bunch of tweaks to make it work just the way I like it (although I have been known to occationally take feature-requests). It also works as a test-platform for the avcodec/format/etc libraries.

If I made you want it now, I'm sorry to say it's not very likely I'll ever release it (due to certain logistics), chalk it up to me being another arsehole core developer tease if you want. :P

The good news however is that I share the sources with those select few (MPlayer is after all GPL), and that one of them is Fab, so chances are you will see atleast some (and probably already have) of these changes in his build... ;)


- CISC

Quote:

Sure, but just pushing 1920x1080 pixels thirty times a second is quite a task, even if it wasn't MPEG encoded... Not to mention that color information has to be converted from YUV to RGB. That is a lot of work, and very remarkable for a 1 Ghz CPU, with no assits for graphic acceleration!

Quote:

Indeed. That's what I had in mind already. I just wanted to point out that the demo shows something that many consider not possible. Well, just like all demos!

Quote:

Great! I didn't know you were also into AltiVec coding. What a pity that this technology is so rare. Let's hope it lives enough to see it well used.

Quote:

Not the least. You are a core developer that gives answers, and I think that many people here appreciate very much all the informations you give us. Of course you made me want "MPlayer CISC", but I understand that's your work, and whatever you do is up to you.
What we should do is learn a lot and make our own programs. Pity that I'm unable to code for MorphOS, coding is the most important use for a computer.

Quote:

That's all I can ask for, thankyou very much! It's important to make clear that the selected few MorphOS developers do share a lot of work and knowledge.

'99 celly?? you mean a celeron?? no waay!:) you're just kidding right?

but your ibook did score better than the peg2, if only with a few %, but did it as i said before:) not if the speed of mos wouldnt be enough even using my poor g3:)))

well, if you have a lil' time, maybe you can dig the mini out and do some test for the masses (and tease us:)

[ Edited by sadddam on 2008/4/23 12:41 ]

Quote:

Pushing those pixels isn't actually that much of a task, it only seems like it because of the poor busspeed of the pegasos, and you're wrong about the colorconversion, it's not necessary at all since most gfxcards support YUV natively (usually through overlay, but can also be done through 3D), and that means there's even less data to transfer (most clips are 4:2:0, ie 12bits as opposed to the 24bits you'd have to use to represent the same for RGB)...

Quote:

It's not as rare as you might think as there are millions upon millions of AltiVec(ish) capable units out there in the form of the current-gen consoles. Anywho, it's fun to figure out how to vectorize stuff (I always do like a challenge), and MorphOS has benefitted from that. ;)

Quote:

I'm always happy to share knowledge, not only to core people...


- CISC

Pushing 1920x1080 pixels is indeed something the pegasos can still do without too much trouble. When writing 64bits data to the "agp" bus, you can reach about 250MB/s. Now if you compute: 1920*1080*(3/2)*30 bytes per second (4:2:0 yuv format using 12 bits, as said above), which is "just" 88MB/s.

the main problem here is that not all graphic cards support that large overlay. When I once tested such large movie on my cheap gfx card I got overlay trouble and parts of the movie were cut off. ;)

Quote:

Right, I forgot these crappy cards are still being made (being the happy owner of a Radeon 9100), even the latest 9250 cards don't support 1920 pixels wide overlays IIRC, which is really sucky in this day and age... :P

To properly display 1080 clips on these cards you either have to scale down the video (mplayer has some nice options for this) or convert and display as RGB, both will significantly slow down playback (to the useless).


- CISC

And wich Card usable under MOS can do it?

Quote:

Off the top of my head the only ones I can think of right now is Radeon 8500/9100 and 9000Pro (I'm not sure if the non-pro version can or not), possibly 9800 too, but I don't know for sure...


- CISC

Quote:

Ah, and I forgot to mention that MPlayer/avcodec also has a rather nifty option to halve/quarter/octer(? ;) ) the decoded size (does not apply to H.264 yet though), saving lots of CPU-cycles in the process .. this means it's possible to play back 1920x1080 clips as 960x540 on much lower spec hardware (this ofcos defeats the whole HD purpose, but atleast it plays).


- CISC

Quote:

Well, I think that 1920 * 1080 *3 * 25 is something like 180 megabytes per second... Not much time left for decoding itself.

Quote:

The DIU unit of the MPC8610 only supports 32bit RGB (and a funny 8 bit palette mode, for nostagics). This is why this demo touches the impossible, there is no video acceleration, the CPU also has to deal with colour space conversion. Phew!

Quote:

It's always a pleasure to read your comments. I'm even beginning to understand some of your jokes!