CoreCodec/H.264 Codec "CoreAVC"

[squid_80]

LOL, when has Core ever met a release date they set? This has been a tradition going back to 1.0 back in April 2006, why break it now?

When was a definite release date given?

[BetaBoy]

My mistake - I thought there was only one bug in CoreAVC and HMS, not simply CoreAVC. You mentioned in one of your previous posts you were only aware of one bug in the 2.0 codebase, but I skipped over the part where you mentioned that you don't work in tandem with Haali. Bummer.
Well we found out that the Motion Vector limit we had has a direct effect on WeightP, so its a 2-for-1 bug fix and requires we rewrite a nice chunk of code. The good thing about this though is as I stated, a) It removes the MV limitation, and now b) Will take into account some future x264 WeightP changes, which we already know are coming sometime next year.

While its obvious we have a vested interest in Haali's work with the splitter, he works at his own pace but has always listened to what needs to be fixed/changed... and if you ask many of the testers (*cough*Jarod) the latest versions have fixed almost every known issue (outside of VC1), but I'm sure more needs to be done.

[betaking]

Well we found out that the Motion Vector limit we had has a direct effect on WeightP, so its a 2-for-1 bug fix and requires we rewrite a nice chunk of code. The good thing about this though is as I stated, a) It removes the MV limitation, and now b) Will take into account some future x264 WeightP changes, which we already know are coming sometime next year.

While its obvious we have a vested interest in Haali's work with the splitter, he works at his own pace but has always listened to what needs to be fixed/changed... and if you ask many of the testers (*cough*Jarod) the latest versions have fixed almost every known issue (outside of VC1), but I'm sure more needs to be done.

:)today i test Haali Media Splitter 1.9.355.21! on windows7 wmp12 can use haali play avi files but test ts/tp file not work! use mpc-hc use haali play some ts file encoder by vc1 not see video but sound is ok! and some vob files can see video but no sound! sound encoder by lpcm! can repot it to haali ?thanks!

[BetaBoy]

A big hurdle has been overcome. CoreAVC 2.0 now supports Non-Compliant Motion Vector's.... and it gets even better, WeightP is now fully supported ;-) (this also includes support for future x264 weightp changes coming in 2010). We are going through QA now and tweaking the code as we go. No timeline on the release, let's see how QA goes.... but current results puts us on top again, but there is still room for improvement ;-)

Thanx all for your patience.

[Mixer73]

Bring it on Dan!

[hydra3333]

Yep, do a good job in QA and then please release it as soon as you reasonably can !

[link2009]

When 2.0 gets released and I see a decent improvement for playback, I will re-purchase it to support the project.

[Romario]

BetaBoy,please, I have one question for you. :)

Can you, please, tell me from which site I can download latest Development version of Halli media splitter, which you will use in CoreAVC 2.0 ?

thanks. I want to test it on Windows 7.

[BetaBoy]

Technically its the same that's on Haali's site right now.... all we do is add our special sauce (non-hacking) to make it work on Windows 7.

[honai]

While its obvious we have a vested interest in Haali's work with the splitter, he works at his own pace but has always listened to what needs to be fixed/changed... and if you ask many of the testers (*cough*Jarod) the latest versions have fixed almost every known issue (outside of VC1), but I'm sure more needs to be done.

Can you, please, tell me from which site I can download latest Development version of Halli media splitter, which you will use in CoreAVC 2.0 ?

thanks. I want to test it on Windows 7.

Technically its the same that's on Haali's site right now.... all we do is add our special sauce (non-hacking) to make it work on Windows 7.

Now you got me really confused. So does the latest version at your hands realize bug-fixes not implemented in the latest downloadable splitter (dated 11/01/2009), or not?

[BetaBoy]

Now you got me really confused. So does the latest version at your hands realize bug-fixes not implemented in the latest downloadable splitter (dated 11/01/2009), or not?
Have you tested the new version Haali posted on his site for any bugs?

[ChronoCross]

Have you tested the new version Haali posted on his site for any bugs?

The one on Haali's site is from January.

[BetaBoy]

I thought Haali had posted the new version already. Pinging him now.

[ajp_anton]

A little offt, but a "11/01/2009" date can be either Nov 1st or Jan 11th, depending on which one of the weird formats you use.
2009/01/11 = no confusion.

[LoRd_MuldeR]

I'd prefer the ISO 8601 format. For example:
2009-12-16

And the latest version on Haali's site definitely dates back to 2009-01-11 ;)

[THX-UltraII]

great news that the new Haali Renderer will be up soon!

[Carpo]

http://haali.net/mkv/mkx.y.9.exe - thought that was the latest unofficial version about

[Keiyakusha]

THX-UltraII
Renderer? I guess you mean splitter? Or there is some changes to the renderer too?

[THX-UltraII]

THX-UltraII
Renderer? I guess you mean splitter? Or there is some changes to the renderer too?

sorry, mistyped

[hydra3333]

Technically its the same that's on Haali's site right now.... all we do is add our special sauce (non-hacking) to make it work on Windows 7.

Hmm, is this http://www.videohelp.com/tools/Preferred_Filter_Tweaker your work in getting codecs to work under Win7 ?

[Jeff Flowerday]

Hmm, is this http://www.videohelp.com/tools/Preferred_Filter_Tweaker your work in getting codecs to work under Win7 ?

:confused:

http://forum.doom9.org/showthread.php?t=146910

[hydra3333]

Ah, it's something different. Thanks.

[honai]

I thought Haali had posted the new version already. Pinging him now.

I understand you are on a busy schedule, but are there any followups on this?

[STaRGaZeR]

You can get the (lastest?) Haali beta with CCCP beta. Deselect everything but Haali and you're done. It supports TrueHD in MKV and detects Blu-ray stereo LPCM tracks ;)

[BetaBoy]

I understand you are on a busy schedule, but are there any followups on this?

Haali will post it on his site when we launch 2.0.

[tal.aloni]

You can get the (lastest?) Haali beta with CCCP beta.

the revision included is Haali 1.9.355.21 (2009-11-14)
you might as well grab it from his site (sorry):
-removed-

nice progress haali, thanks!
however, this release crash with the m2ts samples I posted earlier (TrueHD / EAC3), and also crash with another Blu-Ray I have (TrueHD).

Note:
I have not yet investigated the origin of the crash, there's a slight chance it's unrelated to the splitter.

Edit:
The crash is related to ffdshow, later build resolve the crash, but the playback is still very choppy with TrueHD, I'm not sure if it's a bug with the splitter or the decoder.

[BetaBoy]

tal.aloni.... Excuse me if I ask but, has Haali given you the ok to post that? As far as we are concerned we have respected Haali's work until we have the OK to distribute it or he posts a public link to it (which I do not think has occurred).

[tal.aloni]

I have removed the link, I don't want to offend anybody,
however, it's already embedded into many codec packs.
the file name was published already by a few codec-pack authors.

[BetaBoy]

tal.aloni... Thx it's appreciated. It seems some of the testers decided it was ok to give out the links ;-/ . . . there is a reason its only posted on Haali's site when its ready.

[Keiyakusha]

Not necessarily. It's not hard to guess that link using capabilities of any download manager with support downloading by mask.

[BetaBoy]

True.... but in this case I already have confirmation that the testers in fact did give out download links. No biggie now.... we just don't want reports coming back on versions that maybe old and fixed (or not ;-).

[Carpo]

so is there anything wrong with that build ;)

[BetaBoy]

It dosen't really matter at this point as Haali will be posting the update soon enough.

[BetaBoy]

Well the time has come... We are about to release CoreAVC 2.0 Professional Edition for Windows. Current customers will be notified first... Purchasers in the past 60 days get the upgrade included.... post 60 will get a unique code to upgrade. Emails are going out now for; 1) The new portal login account (all users get this), and 2) The upgrade email.

I'll post more details next.

[BetaBoy]

After 9 months of work, CoreCodec is proud to release CoreAVC 2.0 Professional Edition.

More details below.... but lets get to what's new:

CoreAVC H.264 Video Codec - Version 2.0.0.0 (20091218)
- ADD: Initial support for Windows 7
- ADD: Support for Windows Media Center (in .MKV)
- ADD: Custom fourcc to match the Haali splitter
- ADD: Support for Non-Compliant Motion Vectors (MV)
- ADD: mmx optimizations
- ADD: mmx2 optimizations
- ADD: sse optimizations
- ADD: sse2 optimizations
- ADD: sse3 optimizations
- ADD: ssse3 optimizations (almost 70)
- ADD: sse4 optimizations
- ADD: faster handling of pure-DC chroma blocks
- ADD: new x86 cpu caps
- ADD: Multi-dupe weightp (+Future x264 changes)
- ADD: 64bit support
- FIX: Fix potential failure in CABAC MVD decoding.
- FIX: Proper alignment intrinsics for MSVC and GCC.
- FIX: AVC SEI+userdata fix
- FIX: Focus bug in Windows Media Center that disabled CUDA
- FIX: Better frame re-ordering on difficult streams
- CHG: Supports 16 CPU Cores (was 4 cores max)
- CHG: Support for QuadHD resolutions(4096×4096)
- CHG: New dc_add
- CHG: Faster CAVLC
- CHG: Faster CABAC
- CHG: Faster Deblocking (Massive deblocking overhaul)
- CHG: SSSE3: Faster Motion Compensation (20% faster on Core 2 Conroe and more for Penryn)
- CHG: New CoreNumber for 2.0
- CHG: Initial support for CoreAccount. Linking purchase+account
- CHG: Integrated the Haali splitter into the installer
- CHG: New installer supports both 32/64 bit Operating Systems
- CHG: Filter compiled against ICC (2% speed increase for modern Intel Processors)
- CHG: Updated IDCT to support ARM NEON Cortex A8 Support
- CHG: Updated Blitter YUV/RGB for ARM NEON Cortex A8 Support
- OEM: Initial ARM NEON Cortex A8 Support (iPhone, Touch, Linux, Windows Mobile)
- OEM: SDK NVIDIA CUDA 2.3 support
- OEM: Removed the windows direct3d requirement for CUDA (it can now be native)
- OEM: Updated GStreamer plug-in


Haali Media Splitter (20091112)
- ADD: Official 64bit support
- ADD: Official support for Windows 7
- ADD: Custom fourcc option for windows 7
- ADD: Silent install flags for each install option
- CHG: Disabled explorer thumbnail support (off by default)

Haali Media Splitter supports the following command line options:
/S - silent install without any UI
/MKVONLY - register only Matroska components
/AVI=[yes|no] - register AVI support
/MP4=[yes|no] - register MP4 support
/OGG=[yes|no] - register OGG/OGM support
/TS=[yes|no] - register MPEG TS support
/PS=[yes|no] - register MPEG PS support
/WMP=[yes|no] - register WMP to play in Windows 7


Bulletpoints on CoreAVC 2.0
-------------------------------
- First on Windows 7..... Both CoreAVC and Haali's splitter now use a custom fourcc so that when you play an MKV video it uses Directshow instead of MediaFoundation in WMP and MC. No hacks and its a simple solution to gets ppl out of the pitfalls of MF. Note that this is ONLY for .mkv and does not affect any other container.
- You will also see CoreAVC now supports both 32/64bit, but like Haali we have opted to keep both .AX's in the same folder for simplicity's sake.
- We would also like to thank Clsid for his help with some of the changes we made to the installer for all OS's but mostly Win7.

[clsid]

Haali Media Splitter (20091112)Based on the date I assume this is version 1.9.355.21. What will happen with the bugs that have been reported for that version? Will there be another release in the near future?

[BetaBoy]

If I am not mistaken the date in the readme is wrong. The build # is: 1.9.355.21 that's what counts (as you mention). But otherwise yes, Haali said when he gets time (motivation) he will work on the additional bugs reported.

[Cyber-Mav]

criky thats a big changelog listing a lot of speed boosts, hopefully someone will post some test results here showing how coreavc 2.0 fairs against divx, diavc and against coreavc 1.9.5

[the_corona]

Awesome, lots of speed optimizations!

Any benchmarks you can share? Which proc generation saw the biggest speed bump (probably the latest which ironically needs it the least, lol)

[Disabled]

One Question: Do you sell a full featured High Profile* Decoder this time? Or is it again possible that in 2 years x264 will use a feature previously not used and everyone has to buy a new licence?
(This of course excludes new features that are added to the specs after the purchase, as you don't need to promise on those.)

*normal High profile, not High10 or with 4:4:4 support

[Dark Shikari]

Awesome, lots of speed optimizations!

Any benchmarks you can share? Which proc generation saw the biggest speed bump (probably the latest which ironically needs it the least, lol)
The speed bump is pretty big, though note that the following test was done on a Core i7 with HT, and CoreAVC 1.9.5 was capped to 4 threads max, so the real boost is not as large as the graph suggests. Obviously, CoreAVC 1.9.5 was still faster than ffdshow-mt on a per-core basis; it's just that 8 threads of ffdshow-mt beat out 4 threads of CoreAVC 1.9.5.

http://i50.tinypic.com/b7ett3.png

CoreAVC is now slightly faster than DivX on 1080p HD content. The weightp example is somewhat invalid, as DivX does not decode weightp bit-exact. The results of the 2160p test were a bit odd though: I'm going to guess that DivX is faster on really low bitrate content (the 2160p test case was 4 megabit).

Note "CoreAVC 2.0 Intel" is using ICC, so you can see the benefit of that.

[Cyber-Mav]

yikes looks like divx is still damn quick. would be good to see how diavc compares with the above. i guess coreavc is only good for those who want cuda support, cant see people paying good money for a decoder thats hardly any quicker and in half the cases above slower than a free alternative. ??

[Dark Shikari]

yikes looks like divx is still damn quick. would be good to see how diavc compares with the above. i guess coreavc is only good for those who want cuda support, cant see people paying good money for a decoder thats hardly any quicker and in half the cases above slower than a free alternative. ??Well DivX still doesn't handle weightp correctly, albeit the artifacts are generally not too noticeable. Also note that CoreAVC likely has a larger advantage on older CPUs, which is where you most need the speed of a fast decoder to begin with. I don't have an older CPU to test it on though; I'd be curious about the results.

Hopefully in 2.1 the speed will be even higher ;)

[Cyber-Mav]

need some tests done on older hardware such as athlon xp. wonder if there is some inprovements for atom cpu's too.

shikari any idea if divx will sort out the weight-p issues you mentioned in future divx updates?

[Dark Shikari]

shikari any idea if divx will sort out the weight-p issues you mentioned in future divx updates?Dunno. Slap them until they do. It's a trivial deblocking bug that's very common among decoders, and fortunately one whose effects are probably not noticeable in real situations. It still should get fixed though.

Do note that CoreAVC had to sacrifice a small amount of speed (not much, but measurable) to fix the MV range and weightp issues.

[the_corona]

yikes looks like divx is still damn quick. would be good to see how diavc compares with the above. i guess coreavc is only good for those who want cuda support, cant see people paying good money for a decoder thats hardly any quicker and in half the cases above slower than a free alternative. ??

I agree, that's kinda dissapointing.....hmmm

But from what I remember DivX did comparitably good on high core/thread counts (probably better multithreading), so I'm guessing the graphs would look different on a 2 core machine. At least I'm hoping.....

[Dark Shikari]

I agree, that's kinda dissapointing.....hmmm

But from what I remember DivX did comparitably good on high core/thread counts (probably better multithreading), so I'm guessing the graphs would look different on a 2 core machine. At least I'm hoping.....Likely true as well. We're probably going to look into some multithreading tweaking with lots of cores, though to be honest one hardly needs CoreAVC on a system as beefy as the one I benchmarked on.

[the_corona]

Likely true as well. We're probably going to look into some multithreading tweaking with lots of cores, though to be honest one hardly needs CoreAVC on a system as beefy as the one I benchmarked on.

Wait...."we're" as in "we" as in you and CoreAVC? Are you on their team now?

Does that mean CoreAVC got their optimizations from yours truly?

Edit: I guess you are since the ICC version indicates you have access to the source code.

[Dark Shikari]

Wait...."we're" as in "we" as in you and CoreAVC? Are you on their team now?

Does that mean CoreAVC got their optimizations from yours truly?Yes, all the x86 asm since 1.9.5 has been written by me, including some from x264 that I wrote as well ;)

A bit has gone in the other direction too; x264's cacheline-split chroma MC was originally written for CoreAVC.

[Snowknight26]

Any word on when we can expect HMS as a standalone download?