Messages

1

Software / Uclibc X-compile Toolchain For Z (3.4.3)

« on: February 14, 2005, 02:12:49 pm »

Well, I decided to tarball up the version of gcc that I used for building the mplayer that I made. If anyone's interested in it, that is...

What it will not let you do:
* Compile dynamically linked applications (... that work on any Sharp-based ROM, at least. I could be wrong but I believe that none of the ROMs out there are based on uClibc (why? In my limited tests, Qt/Embedded compiled fine and almost even worked on my 5600...))
* Use anything that Sharp has released in terms of binaries (source may work, no idea...)

What it will let you do:
* Compile statically linked console applications using the latest release version of gcc

http://web.mit.edu/~anrp/www/zaurus/toolchain/

If you're interesting in compiling AGAWA Koji's sources, you'll probably also want the Intel IPP, which you can get a non-commercial license for here:
http://www.intel.com/software/products/ipp/noncom.htm

Couple technical notes about the toolchain:
* Compiled 100% for softvfp-style floating point (necessary for compatibility with Intel IPP)
* Has C, C++, and Fortran compilers included
* Should be installed as /opt/arm-embedded-devel-xscalefvp
* The script env_setup.sh should set up the environment for basic Xcompiling

-anrp

2

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 04, 2005, 04:57:44 pm »

I'll look into making a C3000 version... shouldn't be very hard, I just can't test it...

Also, will look into making a good set of commandline shell scripts around mencoder.

3

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 02, 2005, 09:22:11 am »

yeah no rtprio for me.
any ideas why the equalizer is so slow. I dont have acess to a linux machine but if you do could you tell me if the behaviour is the same in x86 version of mplayer.

tovarish
[div align=\"right\"][a href=\"index.php?act=findpost&pid=65105\"][{POST_SNAPBACK}][/a][/div]

The problem is that af_equalizer is very float-heavy, so even using softfloat doesn't help enough to bring it up to realtime. I'm sure there are integer implementations of an equalizer out there, I may give a shot at adding one...

4

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 10:57:32 pm »

So, for the first run, you do:
mplayer-opt-fm -vo w100 -dr -double -vop rotate=1 -quiet -rtprio 2 file.avi

..and it's fine. Then, you run
mplayer-opt-fm -fs file.avi

and it hangs?

You can try doing the exact same sequence you're using to get it to crash minus the "-rtprio 2" to see where it's failing....
[div align=\"right\"][a href=\"index.php?act=findpost&pid=65027\"][{POST_SNAPBACK}][/a][/div]

i actually use the other options plus -fs  for fullscreen
yes the hanging doesnt occur without the -rtprio option
tovarish
[div align=\"right\"][a href=\"index.php?act=findpost&pid=65044\"][{POST_SNAPBACK}][/a][/div]

Well, that makes sense. If, for whatever reason, mplayer gets stuck in an infinite loop while running with real time priority, the machine will freeze. Understand that it has not actually crashed, it's just that the linux scheduler will always prefer a real time scheduled process over a mere mortal process (like QT,) so the machine becomes unusable. (This is a concern for any real-time process on linux.)

I may be able to hack in a so-called rtsafety patch to mplayer to help in that situation... otherwise, there's not much you can do about it at this point (besides not use rtprio )

5

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 09:37:03 pm »

I'd like to take this post to say something important:

Doing rotation in software via -vop rotate=1 on the fly will cause up to a 40% performance hit in MPlayer overall.

Any media files you want to play on a Cx60 series zaurus really should be pre-rotated....

6

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 08:10:59 pm »

i am using the matrix reloaded trailer to do my testing. the hang (not a crash, i have to do a battery reset  ) occurs usually when i try a fullscreen playback immediately after normal playback. i run it with the foll switches

-vo w100 -dr -double -vop rotate=1 -quiet -rtprio 2
and
-fs
for fullscreen

tovarish
[div align=\"right\"][a href=\"index.php?act=findpost&pid=65024\"][{POST_SNAPBACK}][/a][/div]

So, for the first run, you do:
mplayer-opt-fm -vo w100 -dr -double -vop rotate=1 -quiet -rtprio 2 file.avi

..and it's fine. Then, you run
mplayer-opt-fm -fs file.avi

and it hangs?

You can try doing the exact same sequence you're using to get it to crash minus the "-rtprio 2" to see where it's failing....

7

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 07:39:49 pm »

your mplayer-opt-fm hangs sometimes, but indeed it is faster than atty's build.

doest the -rtprio 2 switch really make a difference. i also get an error about /dev/rtc not being readable

tovarish
[div align=\"right\"][a href=\"index.php?act=findpost&pid=65019\"][{POST_SNAPBACK}][/a][/div]

Hang? How so? Post an example file, maybe? (If mplayer crashes while running with -rtprio 2, it will most likely hang the zaurus due to the nature of that switch...)

8

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 05:55:49 pm »

OK, new binary available.

Link to listing directory (file you're interested in is mplayer-opt-fm)

Benchmarks for it on tmm-11.avi are posted below (look at the link above, files C860-benchmark*txt for other benchmarks on other builds...)

Code: [Select]

mplayer-anrp-02-01-2005 (nofastmath)

./mplayer-n -nosound -benchmark /mnt/card/tmm-11.avi -vo w100 -dr -quiet -rtprio 2

BENCHMARKs: VC: 425.598s VO:   2.240s A:   0.000s Sys:  69.684s =  497.522s
BENCHMARK%: VC: 85.5437% VO:  0.4502% A:  0.0000% Sys: 14.0062% = 100.0000%

---
mplayer-atty-01-31-2005
./mplayer-atty  -nosound -benchmark /mnt/card/tmm-11.avi -vo w100 -dr -quiet -rtprio 2

BENCHMARKs: VC: 421.666s VO:   2.153s A:   0.000s Sys:  70.022s =  493.842s
BENCHMARK%: VC: 85.3850% VO:  0.4361% A:  0.0000% Sys: 14.1790% = 100.0000%

---
mplayer-anrp-02-01-2005 (fastmath)

./mplayer -nosound -benchmark /mnt/card/tmm-11.avi -vo w100 -dr -quiet -rtprio 2

BENCHMARKs: VC: 423.364s VO:   2.218s A:   0.000s Sys:  67.239s =  492.820s
BENCHMARK%: VC: 85.9064% VO:  0.4500% A:  0.0000% Sys: 13.6436% = 100.0000%

[...]

---
mplayer-anrp-02-01-2005 (highopt-fastmath)

./mplayer-opt-fm -nosound -benchmark /mnt/card/tmm-11.avi -vo w100 -dr -quiet -rtprio 2

BENCHMARKs: VC: 415.543s VO:   2.301s A:   0.000s Sys:  68.331s =  486.174s
BENCHMARK%: VC: 85.4720% VO:  0.4732% A:  0.0000% Sys: 14.0548% = 100.0000%

[...]

9

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 05:40:58 pm »

2 anrp: looking forward to your new builds. I'm sure that MP3 VBR problem will be reported very soon and Agava will update the sources. I'm realy hopign to have mplayer build, which makes the best out of Zaurus!
[div align=\"right\"][{POST_SNAPBACK}][/a][/div]

Any chance you could send me (any) MP3 VBR file that you know has problems?
[div align=\"right\"][a href=\"index.php?act=findpost&pid=64978\"][{POST_SNAPBACK}][/a][/div]

Try this one:

[a href=\"http://cacko.oesf.org/downloads/kino2/vbr.mp3]http://cacko.oesf.org/downloads/kino2/vbr.mp3[/url]
[div align=\"right\"][a href=\"index.php?act=findpost&pid=64995\"][{POST_SNAPBACK}][/a][/div]

Hm, amusingly enough, the latest SVN version doesn't work with that file (which means my current working build doesn't play it... although this looks like a simple fix, more in a sec.)

Edit: Yep, got it working. Will post updated binary in a few minutes.

10

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 04:37:18 pm »

2 anrp: looking forward to your new builds. I'm sure that MP3 VBR problem will be reported very soon and Agava will update the sources. I'm realy hopign to have mplayer build, which makes the best out of Zaurus!
[div align=\"right\"][a href=\"index.php?act=findpost&pid=64970\"][{POST_SNAPBACK}][/a][/div]

Any chance you could send me (any) MP3 VBR file that you know has problems?

11

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: February 01, 2005, 03:01:22 pm »

Working on making a newer one from AGAWA Koji's (atty.jp) sources, having a bit of trouble matching his speed so I'm going to fiddle with CFLAGS a bit  

I'm not sure what kind of video chip the 6000 has, but I'm relatively (un)sure that it's not just pure fbdev. Also, this build will run slower than maximally possible on the C3000, because the C3000 has a newer CPU that allows for more optimization as well as a special video chip (PXA27x built-in LCD controller - bvdd) that I don't include the driver for in my binary (since I'm not optimizing the binary for the PXA27x anyway...)

Edit: whoa, that file (RL_MQB_320x240_512_128.avi) is totally different from the file I made. Noticably faster on my latest build from atty.jp's sources than his build... while my file is slower on the same build..!

12

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: January 31, 2005, 07:20:44 am »

One preliminary result is that this version of mplayer can now play WMA files properly, although I had to overclock my Zaurus. Processor load is still very high - 60-70%, but prevsous versions couldn't do even that.
[div align=\"right\"][{POST_SNAPBACK}][/a][/div]

Atty has released today new version of Mplayer for Zaurus 1.0. This version is about 10% faster than version from this topic.
Link:
Zaurus C7xx,8xx series [a href=\"http://atty.jp/?plugin=attach&refer=Zaurus%2Fmplayer&openfile=mplayer-w100_1.0.0-1_arm.ipk]http://atty.jp/?plugin=attach&refer=Zaurus...1.0.0-1_arm.ipk[/url]
Zaurus C3000 series
http://atty.jp/?plugin=attach&refer=Zaurus...1.0.0-1_arm.ipk
+ driver
http://atty.jp/?plugin=attach&refer=Zaurus...0.1.0-1_arm.ipk

Best regards,

Konstantin
[div align=\"right\"][a href=\"index.php?act=findpost&pid=64634\"][{POST_SNAPBACK}][/a][/div]

OK, I just tested it. It is about 3% faster than my version (not 10%) but a new source release is out, thanks for pointing me to it. Maybe I can get it faster :-)

(benchmarks below)

Code: [Select]

mplayer-static-softvfp-ipp:

BENCHMARKs: VC: 446.442s VO:   2.349s A:   0.000s Sys: 111.881s =  560.673s
BENCHMARK%: VC: 79.6261% VO:  0.4190% A:  0.0000% Sys: 19.9548% = 100.0000%

---
mplayer-atty:

BENCHMARKs: VC: 474.755s VO:   2.605s A:   0.000s Sys: 124.336s =  601.697s
BENCHMARK%: VC: 78.9027% VO:  0.4330% A:  0.0000% Sys: 20.6643% = 100.0000%

---
mplayer-static-softvfp-ipp-nofastmath (same as top minus -ffast-math)

BENCHMARKs: VC: 451.735s VO:   2.251s A:   0.000s Sys: 112.733s =  566.719s
BENCHMARK%: VC: 79.7106% VO:  0.3972% A:  0.0000% Sys: 19.8922% = 100.0000%

---
mplayer-atty-01-31-2005

BENCHMARKs: VC: 433.065s VO:   2.282s A:   0.000s Sys: 115.162s =  550.508s
BENCHMARK%: VC: 78.6663% VO:  0.4145% A:  0.0000% Sys: 20.9192% = 100.0000%

13

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: January 30, 2005, 11:43:49 am »

I just ran another benchmarking run, and it may be that the the fast-math-enabled binary is slightly faster on the C860... which is strange, because it wasn't before. o.O The difference is minimal (1%) but is is there.. I may run some more benchmarks to confirm that.

14

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: January 30, 2005, 11:02:04 am »

Well, the reason I built this is that it is built with the latest release compiler (gcc-3.4.3), and built statically (works on X11 and Qt ROMs - only has fbdev and w100 video output drivers, though.) Video decoding is about 8% faster than the mplayer included with Kino2.

Oh, and because I could.

15

Software / New Optimized Mplayer For Pxa25x-based Zaurus Mode

« on: January 30, 2005, 10:48:01 am »

maslovsky: Guess that was enough interest for me   Anyway, updated the binaries on that site, now they have mp3lib disabled and tremor enabled (and changed the top post to reflect that.)