[Mariano]

P.D.: I will try compile the mplayer.dll original changing -O4 to -Os .

THE BEST !!!!!

[Mariano]

THE BEST !!!!!

Sorry, precipitation...

It's equivalent to fast mplayer 1.21

[wizboy11]

Sorry, precipitation...

It's equivalent to fast mplayer 1.21

lol well it should be
Thats pretty much all i did.

This whole "faster mplayer" was discovered purely by chance :p

[biasiman]

Guys I tried with fast Mplayer.dll(link above) to run a 1280x544 stereo 2 min video(harry potter and the order of the phenix trailer) at 1.35 zoom ratio with xbmc af April 15th. It gives me 57 dropped frames whereas
the original mplayer.dll(6.1Mb) only 12.

[elupus]

right, since xvid doesn't have any slowdown, that's very good. aac is abit harder to benchmark.

you can try this.
benchmark=1
novideo=1
ao=null

you won't see anything in xbmc while it's running i think, but it aught to work.

if that is reasonable, (no huge increase in decode time for aac), then i'd say we only have one thing to test left. mpeg2. that normally uses libmpeg2 instead of libavcodec.

If that is about as fast too, i'll change svn version to use -Os instead of -O4

gcc 4.2 would be interesting to test, but I haven't tested it at all yet. It aught to alot better at optimising.

ps. there a quite a few different decoders in libavcodec with different decoding paths, and technically they could all behave differently in this manner, but it feels alot better now

[wizboy11]

It'll have to wait though. Away today

[Mariano]

My results with compilation of mplayer from svn changing -O4 in OPTFLAGS:

- with -O4 --> good

- with -O2 --> better

- with -Os --> much better

Any other possibles flags?

you can try this.
benchmark=1
novideo=1
ao=null

Need time for it... and I don't understand the results of BENCHMARK in log of xbmc.

Perhaps, What is "ao"?

All compilations with gcc 3.4.5, impossible compilation with gcc 4.2.0 for now... (in windows xp sp2).

[Mariano]

BENCHMARKs for elupus.

Without results when optimize XBMC with VS .Net 2003

Comparation between "original" mplayer.dll and "modificated" mplayer.dll (-Os instead of -O4 in config.mak):

OK, ao=null --> no audio output AND vo=null --> no video output

Please, explain me these valors, I don't understand this benchmark...

x264 - original

BENCHMARKs: VC: 183.703s VO: 23.850s A: 0.000s Sys: 1.373s = 208.926s
BENCHMARK%: VC: 87.9277% VO: 11.4153% A: 0.0000% Sys: 0.6570% = 100.0000%


BENCHMARKs: VC: 183.001s VO: 23.805s A: 0.000s Sys: 1.449s = 208.255s
BENCHMARK%: VC: 87.8737% VO: 11.4305% A: 0.0000% Sys: 0.6957% = 100.0000%


BENCHMARKs: VC: 187.895s VO: 23.933s A: 0.000s Sys: 2.413s = 214.240s
BENCHMARK%: VC: 87.7028% VO: 11.1711% A: 0.0000% Sys: 1.1262% = 100.0000%


x264 - Os

BENCHMARKs: VC: 171.306s VO: 23.840s A: 0.000s Sys: 1.426s = 196.572s
BENCHMARK%: VC: 87.1468% VO: 12.1280% A: 0.0000% Sys: 0.7253% = 100.0000%


BENCHMARKs: VC: 171.805s VO: 23.824s A: 0.000s Sys: 1.413s = 197.041s
BENCHMARK%: VC: 87.1921% VO: 12.0907% A: 0.0000% Sys: 0.7172% = 100.0000%


BENCHMARKs: VC: 175.175s VO: 23.946s A: 0.000s Sys: 1.474s = 200.594s
BENCHMARK%: VC: 87.3280% VO: 11.9373% A: 0.0000% Sys: 0.7347% = 100.0000%


xvid - original

BENCHMARKs: VC: 165.841s VO: 8.572s A: 0.000s Sys: 3.406s = 177.818s
BENCHMARK%: VC: 93.2640% VO: 4.8204% A: 0.0000% Sys: 1.9156% = 100.0000%


BENCHMARKs: VC: 165.624s VO: 8.539s A: 0.000s Sys: 3.381s = 177.544s
BENCHMARK%: VC: 93.2859% VO: 4.8098% A: 0.0000% Sys: 1.9043% = 100.0000%


BENCHMARKs: VC: 165.946s VO: 8.551s A: 0.000s Sys: 3.589s = 178.086s
BENCHMARK%: VC: 93.1833% VO: 4.8016% A: 0.0000% Sys: 2.0151% = 100.0000%


xvid - Os

BENCHMARKs: VC: 158.265s VO: 8.586s A: 0.000s Sys: 3.596s = 170.447s
BENCHMARK%: VC: 92.8529% VO: 5.0374% A: 0.0000% Sys: 2.1097% = 100.0000%


BENCHMARKs: VC: 158.410s VO: 8.595s A: 0.000s Sys: 3.793s = 170.798s
BENCHMARK%: VC: 92.7469% VO: 5.0322% A: 0.0000% Sys: 2.2209% = 100.0000%


BENCHMARKs: VC: 158.119s VO: 8.597s A: 0.000s Sys: 3.850s = 170.567s
BENCHMARK%: VC: 92.7025% VO: 5.0404% A: 0.0000% Sys: 2.2571% = 100.0000%


he-aac - original

BENCHMARKs: VC: 0.000s VO: 0.000s A: 15.901s Sys: 218.757s = 234.657s
BENCHMARK%: VC: 0.0000% VO: 0.0000% A: 6.7761% Sys: 93.2239% = 100.0000%


BENCHMARKs: VC: 0.000s VO: 0.000s A: 15.343s Sys: 219.288s = 234.631s
BENCHMARK%: VC: 0.0000% VO: 0.0000% A: 6.5390% Sys: 93.4610% = 100.0000%


BENCHMARKs: VC: 0.000s VO: 0.000s A: 15.460s Sys: 219.194s = 234.655s
BENCHMARK%: VC: 0.0000% VO: 0.0000% A: 6.5886% Sys: 93.4114% = 100.0000%


he-aac - Os

BENCHMARKs: VC: 0.000s VO: 0.000s A: 16.736s Sys: 217.940s = 234.675s
BENCHMARK%: VC: 0.0000% VO: 0.0000% A: 7.1313% Sys: 92.8687% = 100.0000%


BENCHMARKs: VC: 0.000s VO: 0.000s A: 16.252s Sys: 218.409s = 234.661s
BENCHMARK%: VC: 0.0000% VO: 0.0000% A: 6.9256% Sys: 93.0744% = 100.0000%


BENCHMARKs: VC: 0.000s VO: 0.000s A: 16.378s Sys: 218.279s = 234.657s
BENCHMARK%: VC: 0.0000% VO: 0.0000% A: 6.9795% Sys: 93.0205% = 100.0000%

[wizboy11]

Thanks for the benchmarks

Hurray for aac audio not getting any bigger :p (it didn't get faster, but it didn't get slower either).

I'm gonna take a guess on what it means.
VC- probably means video codec or something like that. On the top row it means the amount of CPU time spent on it. On the bottom, the total percentage of CPU time.

VO- probably means something like video output. Same as above, top line is the amount of CPU time spend, bottom is percentage.

A- audio time (duh! )

Sys- amount of system CPU time. Probably stuff like cache and whatnot.

The percentage should always add up to 100%, but it's the time which dictates who won the benchmark.

The total time is the only thing thats really important, although it's nice to see the other things like the system time. That was the total time needed to play the file, shortest wins basically.

Thanks.

How do I force mplayer to decode mpeg2 files instead of libmpeg2?

[wizboy11]

I am attempting to compile mplayer with that latest GCC (it's mingw build, gcc 4.3.0, experimental).

Problem is whenever I go and run ./configurexbmc, it comes back telling me it can't fine inttypes.h (header file). Works fine with the 3.4.4.

What can I do to stop the error?

[spiff]

aac definitely got slower

[wizboy11]

aac definitely got slower

By how much, .05 seconds?

I know that the audio CPU time got slower, but it makes up for it with less system overhead, furthermore, it'll make up for it even more when paried with h264 video which is also noticeably faster (less drops too).

[wizboy11]

And would some mod please update the OP with the latest v1.21 faster mplayer?

I really don't see why enabling edit is bad.
Did something like happen that caused the lead people here to get traumatized from enabling edit?

See, wasted post. Could've just updated my last post which I just posted 10 seconds ago.

I know edit is never gonna get enabled. So, someone who "has" edit, please update OP?

[Ossiostborn]

Hi!
I compared dropped frames on the basis of warren_miller_m480p.mov.

Before: 1600-1800 dropped frames
Now: 1000-1100 dropped frames

Great discovery and work!

Alex

[pike]

@ wizboy11

got a bit fedup with the constant remarks about the lack of edit feature, you are now the sole member of a new usergroup I created, so you should be able to do that edit yourself now.

for an explanation why it's so, it's quite simple, if a user posts something like a personal attack on someone else, I don't want him or her to be able to "take it back" by editing his post afterwards. I trust you are not one of these people