Mplayer Altivec test version for G4 and PA6T-1682M

@K-L

Quote:Thanks. Sam440ep (666 MHz, onboard M9), MPlayer current svn.mplayerhq.hu revision, FFmpeg 2.2.1, new pip implementation (first tests only, not finished, doesn't use DMA yet):
BENCHMARKs: VC: 46.056s VO: 11.656s A: 0.000s Sys: 1.058s = 58.770s
BENCHMARK%: VC: 78.3673% VO: 19.8328% A: 0.0000% Sys: 1.7999% = 100.0000%

@Joerg

Wow ! That is really fast !

@joerg

That's awesome !
Mate if you still have some time can you look also at that 2 remained problems ?

P96_pip - OSD menu flickering issue
https://code.google.com/p/mplayer-amigaos/issues/detail?id=2

Picasso96 PIP not working when WB palette is locked
https://code.google.com/p/mplayer-amigaos/issues/detail?id=7

@K-L

can you write please the exactly strings for make the bench with mplayer i will make it on my pegasos and the Radeon9800 pro

Thanks

@K-L

Quote:No, but after converting it to mpeg1 it is
ffmpeg -i Superman.flv -f mpeg1video Superman.mpeg1
mplayer -nosound -quiet -benchmark Superman.mpeg1
BENCHMARKs: VC: 24.269s VO: 0.297s A: 0.000s Sys: 0.376s = 24.942s
BENCHMARK%: VC: 97.2644% VO: 1.1955% A: 0.0000% Sys: 1.5401% = 100.0000%


@samo79
OSD flickering is fixed in my new pip implementation.

@joerg

Quote:

Great !!
By the way only now i noted that you are not into the list of the committers, if you need i can contact Varthall to give you the access of the MPlayer Google repository

How hard would adding real overlay for radeonhd be?
Nice work by joerg. And liveforit.

@Antique

Quote:
You mean adding required hardware or doing it all in software ?

@TSK

Thinking of hardware accelerated yes. Maybe it's called something else, i think i recal Hans saying overlay is obsolete. In that case i'm thinking of the replacement.
Maybe it's undoable until the new 3d system is ready?

Anyhow, liveforit's composition made a massive boost compared to the old mplayer.

Antique : there won't be Overlay for RadeonHD (did you not read what Hans did write multiple times ?)

As far as I can tell, Composite2D rendering mode is really fast and allows you to watch your vidéos in Fullscreen.

This is the best that our developers could achieve.

If we can improve video playback on Sam440/AmigaOneXE/PegII thanks to an update p96_pip, this is great but, as you say, Composite2d has already made miracles on Sam460/X1000.

@Antique

Quote:
Textured video is the modern day equivalent.

Quote:
No, it doesn't have to wait for the 3D system to be ready.

Hans

@K-L

I think we can have more better performances on Sam 460 and on X1000 just need all the acceleration for the video (decode, player and so) running and good decode library with H264, YUV and all rendering did by GPU.

I had this result on an old p4 2.4ghz 1GB RAM with PCI->PCIE and a RadeonHD 7750 2GB DDR3 thank to a good tweaking of the decode library and a good video player.
with VLC with not tweaked decoding library the videos wasnt unplayable and much more slow compared with mplayer on pegasos2.

Video

@samo79
Quote:
No need to, I'm closely following this thread, and I have sent a PM to joerg :)

@joerg
Thanks for your work, I'm looking forward for your contribution, especially to update the ancient SVN mplayer revision to the latest one. I'd like to do a "vendor branching" as described here:

http://svnbook.red-bean.com/en/1.1/ch07s05.html

In a nutshell:
- Import the SVN revision you have worked on in a new directory under "branches/", merge the changes between the old and new official Mplayer in "branches/official" (which contains the old official r32634 version imported by abalaban), then finally merge the differences between the two versions in our trunk.

At this point, you should be able to import your updated PIP driver to the trunk. What I'd need to know is what is the SVN revision of svn.mplayerhq.hu you have worked on.

What I worry is if all the current versions of the VO drivers will have to be updated to work with the latest MPlayer sources, we'll see.

Varthall

@tslom

I don't think you know exaclty how works our video drivers nor the features that we can have on our system (it has been discussed many many times) or not.

Do not compare your vidéo because it has hardware decoding and *we do not have it* and we won't have it since drivers are closed source.

@Varthall

Quote:I didn't use the code.google.com version because I read notes_amigaos4, seems to be a mess to build it ...
The only thing I had to fix in the svn.mplayerhq.hu one was config.h and config.mak after running configure (maybe it would have worked better with a cross-compiler, but I did it on AmigaOS), adding my new pip implementation in libvo/video_out.c, and changing the paths for AmigaOS (PROGDIR:conf). But I didn't add any other AmigaOS sources (MPlayer-arexx.c, other AmigaOS vos and aos, etc.) yet, maybe they need more changes in the rest of the sources.

Quote:37148 and FFmpeg version 2.2.1 (release archive, not from git), but it doesn't matter, I only made very small changes in the mplayer sources and updating them to newer SVN revisions is no problem.
I can send you a svn diff and my config.h and config.mak if that helps.

First alpha version if someone wants to test it: http://www.sendspace.com/file/82kdca
New pip vo only and unusable sdl (always starts in full screen mode, very slow), no audio (except for sdl but that may not work), doesn't use DMA yet, and probably a lot of bugs

Quote:Only very small changes, for example control() was changed from int control(uint32_t request, void *data, ...) to int control(uint32_t request, void *data).

@joerg

Sam440EP ITX onboard graphics.

mplayer -nosound -quiet -benchmark Superman.flv

BENCHMARKs: VC: 65.823s VO: 11.540s A: 0.000s Sys: 0.950s = 78.313s
BENCHMARK%: VC: 84.0514% VO: 14.7354% A: 0.0000% Sys: 1.2132% = 100.0000%

Cheers

@Joerg

Also tested another video clip, I downloaded from SMTube.
Quality was set at 360.flv within SMTube when downloading.

Downloaded clip
http://www.youtube.com/watch?v=kLyFSGm4vTw

Plays much faster than it should.

Cheers

@Joerg

Thanks !

On My AmigaOne XE G4 1,26 Ghz :

MPlayer-Joearg's Alpha version:
BENCHMARKs: VC: 21.658s VO: 6.195s A: 0.000s Sys: 0.916s = 28.768s
BENCHMARK%: VC: 75.2825% VO: 21.5341% A: 0.0000% Sys: 3.1834% = 100.0000%


Formerly (NOn AltiVec versions only) :

MUI-Mplayer NON AltiVec OS4Depot 2011 release p96_pip:

BENCHMARKs: VC: 20.420s VO: 6.791s A: 0.000s Sys: 0.643s = 27.854s
BENCHMARK%: VC: 73.3109% VO: 24.3818% A: 0.0000% Sys: 2.3073% = 100.0000%

MUI-Mplayer **Beta6** NON AltiVec p96_pip:

BENCHMARKs: VC: 21.714s VO: 6.960s A: 0.000s Sys: 0.634s = 29.308s
BENCHMARK%: VC: 74.0901% VO: 23.7471% A: 0.0000% Sys: 2.1628% = 100.0000%

LiveForIt-Mplayer NON AltiVec p96_pip:

BENCHMARKs: VC: 22.520s VO: 6.665s A: 0.000s Sys: 0.558s = 29.743s
BENCHMARK%: VC: 75.7139% VO: 22.4095% A: 0.0000% Sys: 1.8766% = 100.0000%

Not much speed up but there is absolutely no tearing with your new p96_pip and thats is really great news.

@K-L

Quote:About 5-10% faster on your AmigaOne XE.
I'm only working on the VO, the VC speeds are irrelevant.

Quote:I haven't implemented EOSD yet but still use the old OSD method which directly renders into the video frames, in this version I'm using triple buffering to get rid of the OSD flickering.

@joerg

BIG thanks Joerg, it's very nice to see you updating to the very latest SVN, only problem we will have now 4 different releases to test and at some point we will got crazy without a proper merge !