Mplayer Altivec test version for G4 and PA6T-1682M

SAM440EP-Flex 800MHz, R9250 (Superman.flv):

mplayer alfa:

BENCHMARKs: VC: 53.234s VO: 9.658s A: 0.000s Sys: 0.569s = 63.461s
BENCHMARK%: VC: 83.8841% VO: 15.2186% A: 0.0000% Sys: 0.8974% = 100.0000%

mplayer beta2:

BENCHMARKs: VC: 51.088s VO: 7.415s A: 0.000s Sys: 0.590s = 59.093s
BENCHMARK%: VC: 86.4530% VO: 12.5487% A: 0.0000% Sys: 0.9983% = 100.0000%

mplayer beta2 with skiploopfilter:

BENCHMARKs: VC: 37.012s VO: 7.365s A: 0.000s Sys: 0.570s = 44.946s
BENCHMARK%: VC: 82.3463% VO: 16.3858% A: 0.0000% Sys: 1.2679% = 100.0000%

@K-L

Maybe, I think just down to when I draw the graphics, so I think so.
I also like to fix flickering effect when re-sizing, don't like that.

@ALL developers.

All my fixes to mplayer is uploaded to repo,
only thing I have not uploaded, is the broken vo_p96_pip, as that one needs a lot of work. And vo_comp as I'm working on it.

I hope some of you contribute whit your fixes as well.

https://code.google.com/p/mplayer-amigaos/source/list

@LiveForIT
Quote:

From my side can say that all my interest only in va_comp. That is only "fix" which i need, as all other fixes in SVN are pure for compiling things.

And you not need to make it perfect before uploading. It still worth to upload current version of va_comp too. SVN is not for perfect-ready-to-use-code , but also for work in progress too (and, mostly for work-in-progress in the first place, or then it not SVN, but just some "ready-sources"). I also need it as soon as possible (your current version), so i can plays with it till i have motivation. Its just a bit sadly when i worry much, help with all stuff, and now you just hold it "till its not fully ready". Upload it to SVN, that what SVN for.

Or, as i ask, just send me it in mails (kas1e@yandex.ru), but plz, don't just disappear now with closing code for yourself with saying "when it will be done, only then". I want to build muimplayer with it too, and all my interest before with share all the help to you was because i need vo_comp for muimplayer. Even current non-finished version are fine, as i need to play with it, do some tests, and implement inside things i need , bring betas to users to see how well it integrated to whole muimplayer, and so on.

So plz, share it with me.

Quote:

Don't worry about that one. Joerg already works on, and will share it.

@joerg

New beta knocks a few seconds off the superman.flv and more off some other clips I tested.

Is buffering still being used in this latest beta?

Cheers

@LiveForIt

Quote:

kas1e is right, you don't need to duplicate the effort because once done we might use the same p96_pip (from Joerg) in all versions, but in meantime would be cool if you can commit in Google your current comp driver (even if not 100% finished), so atleast Roman can start to add it into his MUI MPlayer port !

Another step for merge all efforts

@kas1e

Apart the new p96 driver, do you think it will be possible to update to the new r37148 revision code (updated by Joerg) to be use directly into your current MUI port ?

Maybe it's possible that using a recent version some problems related to the corrupted video will be "auto-fixed"...

I'm getting problems using comp with 720x576 files. it only shows about 4/5 of the width whether window or fullscreen.

Playing movies3:TV/Ultimate Force - Season 4/Ultimate Force - 4x05 - Slow Bomb.avi.
AVI file format detected.
[aviheader] Video stream found, -vid 0
[aviheader] Audio stream found, -aid 1
VIDEO: [DIVX] 720x576 24bpp 25.000 fps 1198.8 kbps (146.3 kbyte/s)
VO: [comp] Welcome man !.
==========================================================================
Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
Selected video codec: [ffodivx] vfm: ffmpeg (FFmpeg MPEG-4)
==========================================================================
==========================================================================
Opening audio decoder: [mp3lib] MPEG layer-2, layer-3
AUDIO: 44100 Hz, 2 ch, s16be, 64.0 kbit/4.54% (ratio: 8000->176400)
Selected audio codec: [mp3] afm: mp3lib (mp3lib MPEG layer-2, layer-3)
==========================================================================
[AO SDL] Samplerate: 44100Hz Channels: Stereo Format s16be
AO: [sdl] 44100Hz 2ch s16be (2 bytes per sample)
Starting playback...
Movie-Aspect is 1.78:1 - prescaling to correct movie aspect.
VO: [comp] 720x576 => 1024x576 Planar YV12
720 x 576
Open_Window:301 -- Task 0x52845AC0
VO: Disabling blanker
Cgx_ControlBlanker:876
[swscaler @ 0x4feadbb0]BICUBIC scaler, from yuv420p to argb using AltiVec
[swscaler @ 0x4feadbb0]Warning: dstStride is not aligned!
->cannot do aligned memory accesses anymore


Using the mplayer GUI and setting scale/zoom the 720x520 shows the full width in rougly the right aspect.

@joerg

Haven't seen any comments about your version on an x1000, on mine it brings up a GR if I try to run it.

@sundown

Quote:Probably because most X-1000 users use a RadeonHD gfx card.

Quote:Did you try with -nosound?
The audio parts in the SDL version I'm currently using (SDL-1.2.13-SVNr49-sdk.lha from code.google.com/p/os4sdl) seem to be broken.

New beta version: http://www.sendspace.com/file/5oiy3r

@joerg

Quote:

Can't you use the (SDL) audio part from the Google repository ?
Aniway what's new in this beta 3 ?

Maybe DMA is now used seing the speed up on Sam440:


On Sam440ep o/c 733Mhz and Radeon M9:

Beta3 from Joerg with p96_pip:

BENCHMARKs: VC: 49.497s VO: 4.990s A: 0.000s Sys: 0.980s = 55.467s
BENCHMARK%: VC: 89.2369% VO: 8.9957% A: 0.0000% Sys: 1.7674% = 100.0000%

Formerly with beta2 and p96_pip:

BENCHMARKs: VC: 52.296s VO: 7.978s A: 0.000s Sys: 1.082s = 61.355s
BENCHMARK%: VC: 85.2345% VO: 13.0024% A: 0.0000% Sys: 1.7631% = 100.0000%


On AmigaOne XE G4 1,26 and AGP 9000 Pro:

Beta3 from Joerg with p96_pip:

BENCHMARKs: VC: 21.683s VO: 4.502s A: 0.000s Sys: 0.963s = 27.147s
BENCHMARK%: VC: 79.8702% VO: 16.5834% A: 0.0000% Sys: 3.5464% = 100.0000%

Formerly with beta2 and p96_pip:

BENCHMARKs: VC: 21.420s VO: 4.714s A: 0.000s Sys: 0.914s = 27.048s
BENCHMARK%: VC: 79.1936% VO: 17.4286% A: 0.0000% Sys: 3.3778% = 100.0000%

Not much difference on my G4 but a huge difference on Sam440. So DMA or not ?

@K-L

Your bench with beta 3 seems pretty good !
But the test will be even more interesting when the audio will be implemented, an example:

With the current version of the p96 driver (on Sam440) is not so easy to found a 360p (MP4 videos) that is always 100% in sync with the audio and the video ..
A classic situation when you are trying to watch a video from YouTube ecc ..

So all that speedup is very welcome, just audio is needed to test this speed in a "real" situation, crossing finger for beta 4

SAM440EP-Flex 800MHz, R9250 (Superman.flv):

mplayer beta3:

BENCHMARKs: VC: 50.667s VO: 4.739s A: 0.000s Sys: 0.578s = 55.985s (beta2 59.093s)
BENCHMARK%: VC: 90.5022% VO: 8.4649% A: 0.0000% Sys: 1.0329% = 100.0000%


mplayer beta3 with skiploopfilter:

BENCHMARKs: VC: 36.948s VO: 4.693s A: 0.000s Sys: 0.594s = 42.235s (beta2 44.946s)
BENCHMARK%: VC: 87.4836% VO: 11.1111% A: 0.0000% Sys: 1.4053% = 100.0000%

@mr2

skiploofilter is only useful for HD videos that MPlayer can't cope with. And it's much uglier. No need to test with this option IMHO.

@K-L
Quote:

As you can see it gives a massive improvement (55s vs 42s) and Superman.flv is not a HD movie. Sometimes an ability to see a content of a movie with an acceptable speed is more importan than its quality

@all

I don't seem to be able to get the -benchmark option to work, could some one show a shell line to get it working please.

@sundown

mplayer -nosound -benchmark -quiet superman.flv

@mr2

Totally agree.

On a downloaded clip at 360mp4, I can get the beta9 by kas1e to work nicely with the following options -


vo=p96_pip
lavdopts=skiploopfilter=all
quiet=1
Framedrop="1"

Decent enough quality and plays in sync.
Can't get it to stream with those options though.

The framedrop option when streaming renders the video unwatchable, the skiploopfilter does not.

As the version by Joerg is faster than the version by kas1e, Sam440EP ITX owners might just be able to stream youtube videos at a decent enough quality.

Cheers

@samo79

Quote:Except for trying to include a file which doesn't exist, an useless IDOS->Printf(), screwed up comments and changing the size of a buffer there are not differences in the sources. Maybe the buffer size change is important: http://www.sendspace.com/file/e4i0qh

@joerg

I think I downloaded, LibSDL from OS4Depot.net.

What file is missing? Maybe its because your using different SDK then I'm using.

I know some commit comments are not correct, but there are no way to edit the comments after you have committed the changes.

If you are referring to comments in ao_sdl file then that's because every thing was written in on line, as my text editor does not have line wrapping, I simply divided the text string.

There are some left over Printf() form when I was debugging some crashes on exit, it was down to how mplayer was linked I believe, some how some of the SDL exit code was called instead of the correct routines as they where using the same function name in different .o files, in the vo_cgx_wpa I believe or was it vo_comp.

I Just did not understand way my uninit() code was not called, every thing looked prefect, but whit using Printf() I discovered that wrong one was called, I can remove the Printf() now that problem is solved.

I did not change the size of the buffer, so if that's different then maybe its what your looking for.

https://code.google.com/p/mplayer-amig ... trunk/src/libao2/ao_sdl.c