Peg2 / 1 ghz / os4.1 with latest stuff. In generall almost everythink working, but slow. Some not so "heavy" games works fine. Some videos works fine too. Some time a have crashes, but i am not sure what cause the problem - original source code of gnash, or some bugs of aos4.1 itself (if you in interest i can save the Grim logs).
In general the main problem - it's pretty slow even in original windowses modes. I am not expert, but imho, HW rendering (over OGL (minigl) ), can help a lot with that ?
I tested both - swfdec on morphos and gnash on os4 (on the same hardware), and well.. they both slow in generall. But we all expected that it will be like that. Fab say that when (if) someone will add HW support for swfdec (cairo?) then it will speedup a lot swdec on mos, so, that make me think that if you will try to add HW acceleration for Gnash, it will speedup it a lot.
How hard add HW to gnash btw ? As i know, some OGL already in gnash, and maybe that will be not so hard to add minigl support ? Also as i know that OGL support are experemintal, but if it works already for some "main" parts (like render some crap over it) then it's good start.
Can't try youtube vids thorugh (not have python installed at moment).. Just trying it as: aos4-gnash http://youtubeballablab with no luck, but i assume it will be slow too ..
Cairo backend is already included in the OS4 version of Gnash. But for the release version i've used AGG that is more precise than Cairo and since Cairo is not HW accelerated is better. To use cairo instead of AGG, gnash needs only a simply recompilation. The youtube problem is not related to any backend because in the gnash testsuite there is a file called "easyvideo.swf" that load from disk a file called "media.flv" (any valid flv video) and that is played with 40% of CPU on my A1. The youtube problem is into the streaming part of gnash but we (me and gnash developers) haven't find it..
I am not about Cairo in general, i am about add MiniGL support (over experemental OGL support) in Gnash. Becouse if Cairo not HW rendered , so then it's not need to use at all with gnash of course.
Did i understand you right, that Gnash can play (not streamy) youtube vids on 40% load of CPU ? (full-speed, music, etc ?)
I am not about Cairo in general, i am about add MiniGL support (over experemental OGL support) in Gnash. Becouse if Cairo not HW rendered , so then it's not need to use at all with gnash of course.
I don't think our OGL implementation will help so much and sice HW Cairo is planned it would be the best solutiono even if it not so advanced like agg backend in Gnash
Quote:
Did i understand you right, that Gnash can play (not streamy) youtube vids on 40% load of CPU ? (full-speed, music, etc ?)
I've only had time to try a couple of things so far.
1 .?Tried playing a local .flv using easyvideo.swf
First was a download of the pianeta report from youtube (got with getvideo) this was very slow, could not play it.
Tried another video from a clients site which I know to be a "proper" flv and this worke very well. Played with about 60% cpu usage.
That with SAM Flex 766 mhz with radeon 9250 and 32bit workbench screen.
I think the flvs downloaded by getvideo are in a different codec.
2. Then I tried playing a video direct from youtube.com using the python script. Worked okay, all the gadgets seems to work in the player. Actually speed was too slow.
3. Tried streaming some music from my "station" at iacmusic.com
This worked okay. 100% cpu usage whilst streaming but audio was unbroken. The window opened with the wrong shape looked like x and y were swapped.
yes, the video are the worst part of gnash if streamed from network. 60% of cpu is not so bad on a SAM even because it isn't hw accelerated. for example a video that i play with Gnash on my A1 G4 933mhz use 40% of cpu speed while with mplayer that has the overlay on it use only 20% of cpu (and mplayer eat many cpu that gnash..)
So the first thing to speedup is the video reprodution.. maybe Cairo accelerated can help
In fact, i doubt it's streaming being your problem. Here that easyvideo.swf + some lowres flv also takes 35% cpu with swfdec, but the actual problem (at least in swfdec, but i suspect in gnash too) is that the usual video players (youtube in particular) add much overhead with their slider, time status and so on. Youtube in particular hurts swfdec very much, because it forces it to take a slow render path because of some shapes.
And of course, youtube has slowly migrated to h264 content, which makes it worse too (because decoding lowres flv takes about 2% cpu with ffmpeg :)).
yeah, i know.. but we must admit that Flash is also slow on the PPC with macosx. My Dual G5 with a geforce 6800 256MB play swf not so fast like my dual core intel laptop.. so we can't expect miracles at the end.. we must only hope to play videos at a "decent" rate..
Gnash is working on my sam 667,Tooooo slow. Only tested youtube,i get problems during or after i have seen the movie in fullscreen mode(the 2 times i tried.).
@kas1e H/w accelerated Cairo is "planned" by Hyperion. As we should know by now "it's done when it's done", and no sooner! Maybe Hyperion's "most ambitious project ever" will have some bearing on it (either delaying it, or perhaps being part of it).
yes... but only i have something to work on my hands.. at moment i'm trying to recompile OWB. Yesterday i've sent some patches to OWB guys to configure the program again.. The OWB development has move on in the meanwhile..
i cannot confirm anything but this is the good thing of the open source programs.. Anyone can contribute to it and even if a guy don't want to continue the development anyone can do. I'm the first one that has compiled OWB for any platforms available in our world. MOS/AROS/OS4/OS3 and ALL of these platforms has used as base mine changes and, even if anyone could do an OWB version from scratch, they have found at least a 50% of work already done..