CoreCodec/H.264 Codec "CoreAVC"

[SeeMoreDigital]

I've tested a couple of hi-def AVC within .TS streams...

The 332KB AVC.plg indicated that it did not support "interlaced" files. However, the 348KB AVC.plg had no such worries ;)

Playback was still a little jerky though... but that's more likely to be my "slow" PC :eek:

[BetaBoy]

yeah... the newer version I sent you had basic interlaced support for field only.

[SeeMoreDigital]

yeah... the newer version I sent you had basic interlaced support for field only.Thanks for the confirmation...


Cheers

[HookedOnTV]

Selur... yes, thx. I'll probally have one more test for you guys before we slap 1.0 on it. What we really need atm is some more people with 'true' 4 CPU's to test loads on each CPU.

Anyone?

I have 4 cpu's... but they aren't all in the same box ;)

[GhengisKhan]

I have 4 cpu's... but they aren't all in the same box ;)

Well, in that case I have 5... oops fogot my PDA - 6

[TEB]

Betaboy. I got a couple of HP-DL385 at work with 2xDualcore AMD opterons on.. unfortunatly these are running linux only.. Is there are possibility for a linux version in the nearby future? (i could run /dev/null for the rendering tho for testing threading capabilities of the decoder..)..

teb

[BetaBoy]

Our first of 5 linux versions will be ready for testing in about 3 weeks. So till then...

[708145]

Our first of 5 linux versions will be ready for testing in about 3 weeks. So till then...

great :D
finally testing on a non-obsolete OS :>

P.S.: I can offer testing on a quad 850 as well :)

bis besser,
Tobias

[MuTeK]

TCPMP Version 0.72RC1 Benchmark Results: Non-SMP / SMP

>>batman_begins_1080p.mov

Average Speed 228.68% 818.21%
Bench. Time 1:03.104 0:17.565
Bench. Frame Rate 54.83 196.17
Bench. Data Rate 19.7 Mbit/s 70.8 Mbit/s


>>bbc_1080p.mov

Average Speed 165.10% 598.16%
Bench. Time 0:56.481 0:15.512
Bench. Frame Rate 39.62 143.56
Bench. Data Rate 13.8 Mbit/s 50.3 Mbit/s


>>ice_age_2-tlrD_h1080p.mov

Average Speed 305.90% 1020.21%
Bench. Time 0:28.155 0:08.384
Bench. Frame Rate 73.34 244.60
Bench. Data Rate 32.2 Mbit/s 108.3 Mbit/s


>>king_kong_h1080p.mov

Average Speed 229.64% 811.37%
Bench. Time 1:05.401 0:18.443
Bench. Frame Rate 55.06 194.54
Bench. Data Rate 22.5 Mbit/s 79.8 Mbit/s


>>the_new_world_h1080p.mov

Average Speed 259.20% 910.21%
Bench. Time 0:57.333 0:16.257
Bench. Frame Rate 62.14 218.23
Bench. Data Rate 24.2 Mbit/s 85.6 Mbit/s


>>serenity_1080p.mov

Average Speed 266.91% 900.00%
Bench. Time 0:53.098 0:15.686
Bench. Frame Rate 64.01 215.85
Bench. Data Rate 21.8 Mbit/s 74.0 Mbit/s


>>the_ringer_h1080p.mov

Average Speed 238.18% 828.01%
Bench. Time 0:55.808 0:15.992
Bench. Frame Rate 57.11 198.52
Bench. Data Rate 23.3 Mbit/s 81.4 Mbit/s


>>transporter_2_h1080p.mov

Average Speed 239.72% 823.10%
Bench. Time 0:58.302 0:16.888
Bench. Frame Rate 57.48 197.35
Bench. Data Rate 21.5 Mbit/s 74.3 Mbit/s


>>x-men_3-pre_teaser_h1080p.mov

Average Speed 334.24% 1115.75%
Bench. Time 0:29.661 0:08.836
Bench. Frame Rate 80.14 267.51
Bench. Data Rate 34.0 Mbit/s 114.4 Mbit/s

[Gnerma]

And what hardware did you run those tests on MuTeK? Looks like a 4 device system.

[MuTeK]

And what hardware did you run those tests on MuTeK? Looks like a 4 device system.

Yes, Dual Opteron 270 :rolleyes:

[Farpenoodle]

Wow. Looks like it scales extremely well with more threads. Don't know how you guys did it. Looking forward to getting my hands on a dual-core system. :)

[SeeMoreDigital]

How successfully can you guys play the following AVC in .TS samples. Captured from Pro7 HD Channel using a DVB-S2 PCI card?

http://www.dvbmagic.de/download.php?nav=Download-Center&downloadid=26 (20.5MB)

http://www.dvbmagic.de/download.php?nav=Download-Center&downloadid=27 (at 18.4MB)


Cheers

[BetaBoy]

MuTeK... looks great. Your results mirror a few...

I guess... the bottom line is that it looks like there are no bugs being reported related to the SMP code... so it looks like a go.

[3ngel]

@Betaboy
In the retail release of CoreAVC DxFilter, there will be an option to obtain maximum quality (regadless of performance)? Something like "IEEE 64bit Float Decoding"? And deringing and deblocking options?

[IgorC]

That would be interesting to see.
But I think perfomance of H.264 decoders is different to ASP.
Deringing and deblocking are already here ( in encoder-decoder. Pre-postproscessed by inloop filter)
"IEEE 64bit Float Decoding" has misc. effect. Visual quality's impact is *much* less than 1%, but speed will gone down noticebly.

Nero/Ateme, Elecard, CoreAVC decoders have the same metric results. Standard restricitions? Not much to change?

[GodofaGap]

AVC decoding is bit-exact. Any decoder that is bug free and outputs to the same colorspace should give the *exact* same picture.

[tomos]

How successfully can you guys play the following AVC in .TS samples. Captured from Pro7 HD Channel using a DVB-S2 PCI card?

http://www.dvbmagic.de/download.php?nav=Download-Center&downloadid=26 (20.5MB)

http://www.dvbmagic.de/download.php?nav=Download-Center&downloadid=27 (at 18.4MB)


Cheers

plays fine on my 4400+ X2. 20-25% CPU usage

2nd file tho causes an alert - 'buffer size is too small for this media. try increase it in player settings for better playback'.

after that its fine tho :)

[SeeMoreDigital]

2nd file tho causes an alert - 'buffer size is too small for this media. try increase it in player settings for better playback'. Yes.... I get this alert when playing the second file too!

Does anybody else?

[Dreassica]

First plays fine, 2nd only audio output, no error msg though.

[pankov]

yeah... the newer version I sent you had basic interlaced support for field only.
I was just thinking if the decoder will support Interlaced content what kind of Deinterlaceing will it has ... if any. Also will it be possible to somehow use the GPU for deinterlacing as it's done with the NVidia's PureVideo MPEG2 decoder which has very good (may be better than all software I've seen) hardware deinterlacing (pixeladaptive / spatial-temporal de-interlacing).

I also want to ask if the current beta/test version includes a DirectShow filter because I use ZoomPlayer and would like to test it in reallife conditions.

[RBF]

Our first of 5 linux versions will be ready for testing in about 3 weeks. So till then...
Any plans for MacOS X (universal binary) versions?

[BetaBoy]

Any plans for MacOS X (universal binary) versions?
By the end of the year we hope to have 2 different OS X versions.

[BetaBoy]

I also want to ask if the current beta/test version includes a DirectShow filter because I use ZoomPlayer and would like to test it in reallife conditions.

No its strickly for TCPMP.

[pankov]

@BetaBoy

I'll be waiting for the official release then :)
any ETA?

any comment on deinterleasing and GPU support (hardware acceleration)?

[BetaBoy]

I'm hoping to announce the new launch date later today.

[SeeMoreDigital]

I'm hoping to announce the new launch date later today.Nice one ;)

[Romario]

BetaBoy, I know that your official CoreAVC will be very good. Can't wait to test it.

[SeeMoreDigital]

I forgot to ask before but what does the abbreviation, TCPMP stand for?


Cheers

[nm]

http://tcpmp.corecodec.org/about

[SeeMoreDigital]

Thanks mate....

"The Core Pocket Media Player".... doh!


Cheers

[niknik]

you probably aren't used to the "pocket" part. ;)

for anyone having a pocket PC, TCPMP is an essential tool. :)

[GhengisKhan]

Just for you guys to know Betaboy said in the corecodec forums (not exact quote): when v. 1.0 is released the name will change from TCPMP to Betaplayer and Coreplayer. Betaplayer will be the GPLed version and Coreplayer will be the Free and Demo/Pay version.
The difference between Coreplayer Free and Betaplayer will most noticeably be the GUI, and Betaplayer will be a "playground" for testing new releases/features.

For more info you can go to Corecodec.com.

How this tie in with CoreAVC? CoreAVC and Coreplayer are goiing to be released at the same time (if I understood correctly), and for maximum performance in CoreAVC you'll want to be using either Coreplayer, or Betaplayer (DS Filter is slower).

[BetaBoy]

Looks like we have a lil more work todo on the DS filter side of things all. I'll fill you in when I can.... /me grumbles at TS streams.

[BetaBoy]

Incoming email guys... CoreAVC SMP test #2 has been sent... it also now has PAFF interlaced content enabled... so give it a try and post feedback if you are seeing an improvement over SMP #1... thx

[tomos]

testing progressive content is not a problem, but interlaced....

any interlaced samples to download? have none handy to mess around with

[BetaBoy]

one of the earlier posts in this thread has some stuff you can download....

[tomos]

ah ok, didnt even realise they were interlaced. cool :)

1 thing i've noticed just now playing in mpc - my clips all seem out of sync - like the audio is playing about 250ms early

trying it in the first player and thats ok. cant seek forward as this version wont allow it.

the 2nd player seeks well and doesnt lose sync at all from what i've seen so far. jumped around movies of various lengths and no probs with audio/video sync yet.

the audio oddity is the only prob with the use of the codec in MPC so far. CPU usage is great

as far as comparing the 2 players :

player 1
TCPMP Version 0.72RC1 Benchmark Results

Average Speed 344.37%
Video Frames 4534
Audio Samples 9081600
Amount of Data 115860 KB

Bench. Time 0:54.913
Bench. Frame Rate 82.57
Bench. Sample Rate 165381
Bench. Data Rate 17.2 Mbit/s

Original Time 3:09.105
Original Frame Rate 23.98
Original Sample Rate 48000
Original Data Rate 5.0 Mbit/s

URL C:\sin city sync ttest.avi
Size 118640640
Platform Windows
OS Version 5.01
Clock speed 2143 Mhz
Video output Null Video 1920x1080 0bits Lookup
Video zoom 1920x1080 -> 0x0
Audio output Null Audio 48000Hz 32Bits 2Ch.

player 2

TCPMP Version 0.72RC1 Benchmark Results

Average Speed 352.37%
Video Frames 4541
Audio Samples 9081600
Amount of Data 115860 KB

Bench. Time 0:53.750
Bench. Frame Rate 84.48
Bench. Sample Rate 168972
Bench. Data Rate 17.6 Mbit/s

Original Time 3:09.397
Original Frame Rate 23.98
Original Sample Rate 48000
Original Data Rate 5.0 Mbit/s

URL C:\sin city sync ttest.avi
Size 118640640
Platform Windows
OS Version 5.01
Clock speed 2144 Mhz
Video output Null Video 1920x1080 0bits Lookup
Video zoom 1920x1080 -> 0x0
Audio output Null Audio 48000Hz 32Bits 2Ch.


not much diff between these as far as progressive content goes.

as for interlaced, tried the charlie's angels one and couldnt see any diff between the 2 with either player to be honest. both seemed spot on. couldnt get the video going in MpC. audio only.

as for the john cleese clip, the 1st player gives the buffer err as mentioned before but unlike before, the image now pixelates?

2nd player worked sweet, no errors, no pixellation. it just worked.

MpC wont show video. guessing i'm missing a splitter or something?

as for benchies. here's player 1 v 2 on the john cleese clip

player 1

TCPMP Version 0.72RC1 Benchmark Results

Average Speed 244.03%
Video Frames 207
Audio Samples 371200
Amount of Data 18793 KB

Bench. Time 0:03.168
Bench. Frame Rate 65.32
Bench. Sample Rate 117143
Bench. Data Rate 48.5 Mbit/s

Original Time 0:07.733
Original Sample Rate 48000
Original Data Rate 19.9 Mbit/s

URL c:\prosieben_hd_21155412022006.ts
Size 19323768
Platform Windows
OS Version 5.01
Clock speed 2152 Mhz
Video output Null Video 1920x1088 0bits Lookup
Video zoom 1920x1088 -> 0x0
Audio output Null Audio 48000Hz 32Bits 2Ch.


player 2

TCPMP Version 0.72RC1 Benchmark Results

Average Speed 258.37%
Video Frames 208
Audio Samples 378368
Amount of Data 18793 KB

Bench. Time 0:03.050
Bench. Frame Rate 68.18
Bench. Sample Rate 124016
Bench. Data Rate 50.4 Mbit/s

Original Time 0:07.882
Original Sample Rate 48000
Original Data Rate 19.5 Mbit/s

URL c:\prosieben_hd_21155412022006.ts
Size 19323768
Platform Windows
OS Version 5.01
Clock speed 2144 Mhz
Video output Null Video 1920x1088 0bits Lookup
Video zoom 1920x1088 -> 0x0
Audio output Null Audio 48000Hz 32Bits 2Ch.


Overall it looks like the 2nd player is a little faster but more stable too.

as far as Mpc v Player 2 with x264 in avi goes:

a test clip or rome i have uses between 50-75% CPU usage in Mpc while averaging out at roughly 50-55%

player 2 jumps to about 70% on 'complex' scenes but drops down to 33% and averages about 35-45% overall

also more importantly - the same clip is in sync in player 2 but slightly off in MpC with the newer codec. using time shift works but curious that it goes out of sync at all

forgot to add that this is on a stock 4400+ X2 :)

[PicardGK]

1 thing i've noticed just now playing in mpc - my clips all seem out of sync - like the audio is playing about 250ms
early
I just made some changes and compiled the new DShow filter, but it was not really tested yet :o

What type of files are out of sync: avi or mp4? Or both? Are apple trailers also effected? (I will too make some tests tomorrow)

[tomos]

will try some apple trailers later on. so far i;ve been using my own clips which are in avi

ok, just tried some movs

scanner darkly, xmen last stand both play in sync. so it looks like just avi's so far. cant use megui to make mp4's to test.

[BetaBoy]

Here is a wierd one... This Apple HD clip:
http://www.apple.com/trailers/imax/imaxdeepsea3d/hd/

Comparing to playback in QT... it seems to play slower in TCPMP... I have not tried the DS filter yet with it.

[PicardGK]

so it looks like just avi's so far.
There is a very good chance it's only avi. This is where I made some timing changes without verifing.

[tomos]

ah ok. if i find some mp4's i'll give those a go just to eliminate those too :)

[MuTeK]

MuTeK... looks great. Your results mirror a few...
I guess... the bottom line is that it looks like there are no bugs being reported related to the SMP code... so it looks like a go.

Yes, the results turned out to be very good, performance increases ~3,6 times on SMP version. At that, I experienced no decoding problems .

Incoming email guys... CoreAVC SMP test #2 has been sent... it also now has PAFF interlaced content enabled
Thank you, but it seems that I will have to search for PAFF streams. How the things are going with MBAFF? It’s much easier to find MBAFF streams.

[BlackSun]

Here is a wierd one... This Apple HD clip:
http://www.apple.com/trailers/imax/imaxdeepsea3d/hd/

Comparing to playback in QT... it seems to play slower in TCPMP... I have not tried the DS filter yet with it.

I don't have quicktime installed but it plays fine both with TCMP + CoreAVC or my own CorePlayer build.

[BetaBoy]

Load QT... play the file in it... then play it in TCPMP and you will see it plays about 10% slower.

[sasam]

Thanks for the mail Dan,
I ran a little speed test on my 2 desktop and 3 laptop 1core machines...

Here it is if anyone is interested:
http://img162.imageshack.us/img162/1851/coreavc6br.gif

PM = Pentium mobile
CM = Celeron mobile

All exept the old one can play HD Superman trailer, and thats pretty cool. ;)

[lexor]

Here is a wierd one... This Apple HD clip:
http://www.apple.com/trailers/imax/imaxdeepsea3d/hd/

Comparing to playback in QT... it seems to play slower in TCPMP... I have not tried the DS filter yet with it.
I had a similar issue with a lot of Apple's HD trailers (well slower playback, and more CPU usage). It has to do with .mov container, I repacked into mkv and it's all peachy.

btw I'm on the last alpha of DS filter publically available before you pulled the download (so 0.0.0.4 or something like that)

[BetaBoy]

Picard fixed the audio timing issue in SMP2 which I did not test it on.

BTW... SMP3 is heading out now guys... check your emails... this fixes the reported AVI timing issues.

[popper]

you probably aren't used to the "pocket" part. ;)

for anyone having a pocket PC, TCPMP is an essential tool. :)

great i can finally post here (whats with the LONG 5 days delay anyway, and could'nt even MP people)

so your saying if you have a PPC its essential :devil:

what about the people that have real PPC's as in G4/5 with altivec units ?.

will there be a linux PPC (G4)(with altivec ) version of Core for the likes of http://www.ppczone.org/forums/viewtopic.php?t=387 people
see http://www.ppczone.org/forums/index.php for all the linuxPPC platforms that exist today, theres a valid market there.

remember to have FUN :rolleyes: