Log in

View Full Version : CoreCodec/H.264 Codec "CoreAVC"


Pages : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 [101] 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144

Jay Bee
13th October 2009, 07:19
What can I do to get smooth playback of interlaced H264 content (50Hz 1080i - EurosportHD / BBC HD)?


Cyberlink decoder. Free with PDVD demo.

CiNcH
13th October 2009, 08:53
blubberbirne... What do you mean by interlaced? If it works fine now... and there is no diff in that regards to 2.0.
What can I do to get smooth playback of interlaced H264 content (50Hz 1080i - EurosportHD / BBC HD)?
Problem is the bad standard renderer support (e.g. VMR) of CoreAVC (especially with interlaced content). I have reported it several times and the interest in it was pretty low. I will check back with 2.0 and see whether the problem is addressed there.. (even though I am not too confident about that)

All I have learnd is that CoreAVC works well with Haali Renderer which DVB apps do not make use of (hint was given by a CoreAVC dev so I think they are pretty much aware..).

hajj_3
25th October 2009, 12:34
I don't suppose the new CoreAVC is any closer to release now? Win7 has been out for a few days, hope its out soon:)

BetaBoy
25th October 2009, 14:11
We are working on the 2.0 installer now for the release. No time frame... but QA will be next.

leeperry
25th October 2009, 18:21
We are working on the 2.0 installer now for the release. No time frame... but QA will be next.
including HMS 2.0? still open for beta-testing if you want http://img514.imageshack.us/img514/8138/icecreamt.gif

JohnnyFu
27th October 2009, 23:08
Betaboy, is there a changelog for 2.0?

BetaBoy
28th October 2009, 00:11
As soon as it is released sure.... but the focus with the initial 2.0 is:

- Updated Haali splitter
- Merged 32bit and 64bit versions
- Windows 7 Compatible Installer
- CPU instruction set additions (massive)
- Updated CUDA SDK support
- New Installer
- Bug fixes

Plus about 30+ more changes/additions/fixes... but its a well deserved 2.0 Milestone for sure.

leeperry
28th October 2009, 03:57
if you could somehow decrease the CUDA initialization time, as discussed in this thread..this would be most fantastic: http://forum.doom9.org/showpost.php?p=1328499&postcount=21

JimmyZ
28th October 2009, 18:09
As soon as it is released sure.... but the focus with the initial 2.0 is:

- Updated Haali splitter
- Merged 32bit and 64bit versions
- Windows 7 Compatible Installer
- CPU instruction set additions (massive)
- Updated CUDA SDK support
- New Installer
- Bug fixes

Plus about 30+ more changes/additions/fixes... but its a well deserved 2.0 Milestone for sure.

so i guess OpenCL support is not coming...

nm
28th October 2009, 18:24
so i guess OpenCL support is not coming...

Unlike CUDA, OpenCL doesn't have a video API for accessing hardware decoders. Writing a new decoder that runs on the stream processors would be a huge task and probably not worth it anyway.

dimitrik
29th October 2009, 12:36
As soon as it is released sure.... but the focus with the initial 2.0 is:

- Updated Haali splitter
- Merged 32bit and 64bit versions
- Windows 7 Compatible Installer
- CPU instruction set additions (massive)
- Updated CUDA SDK support
- New Installer
- Bug fixes

Plus about 30+ more changes/additions/fixes... but its a well deserved 2.0 Milestone for sure.

Does "Windows 7 Compatible Installer" mean that v2.0 will Windows Media Foundation compatible i.e. it will be able to function inside Win7 Media Center?

Or am I hoping for too much?

Keiyakusha
3rd November 2009, 03:41
BetaBoy
Sorry for asking here, I just can't wait...
Is there some changes to Haali's renderer? I remember you posted something about that.

EDIT: Also this is a bit too long for creating an installer... is there any problems suddenly appeared?

G_M_C
3rd November 2009, 17:28
After reading this in the "x264 development" thread;

the issues i recall being stated surrounding weightp being broken (haven't seen any confirmation of fixes) were

A) 2pass encoding does not work to a miss constructed check under some situations which i don't recall offhand (x264 errors out and doesn't start processing, so you'll know it if you come across it).
B) weightp 2 triggers an incredibly large number of scenecuts (when scenecut is active) which will cause a quality drop.

on a side note, this email (http://mailman.videolan.org/pipermail/x264-devel/2009-November/006502.html) is relevant.
though it was stated as 'few days' in above email, pengvado/akupenguin is now on vacation for a while
and weightp will not commit until it has his full approval.

CoreAVC broken with "weight-p" in default mode. D'oh! Hope they get the fix out very soon :D

I'll ask the obvious question: When will version 2.0 of CoreAVC see daylight ?

BetaBoy
3rd November 2009, 17:36
We are aiming for the end of this week/early next week so it does not impact x264 encodes.

G_M_C
3rd November 2009, 17:36
We are aiming for the end of this week/early next week so it does not impact x264 encodes.

Thx for your quick reply :)

BetaBoy
3rd November 2009, 17:42
BetaBoy
Sorry for asking here, I just can't wait...
Is there some changes to Haali's renderer? I remember you posted something about that.

EDIT: Also this is a bit too long for creating an installer... is there any problems suddenly appeared?

Well I am waiting on Haali for the full changelog.... but the biggest changes are the removal of the explorer integration and the silent install options. for example Haali's splitter is now integrated into the CoreAVC installer as an additional install component.

The later change for taking a higher priority over MediaFoundation in both the splitter and CoreAVC will come after the 2.0 release for sure. We have argued against fighting MS and don't want to bastardize the OS with any registry hacks... but instead have opted for a simple solution that just gives Directshow the higher priority.

clsid
3rd November 2009, 18:00
Could you provide some more details about how you have given DirectShow a higher priority?

BetaBoy
3rd November 2009, 21:44
I'd prefer to just get CoreAVC 2.0 out... but I will PM you with the now method Vs. the later method.

Disabled
4th November 2009, 11:37
I'll ask the obvious question: When will version 2.0 of CoreAVC see daylight ?

I thought the obvious question was: What are you doing with buyers of the 1.x Version? As far as I know the 2.x series is not free for 1.x buyers, but if 1.x still does not implement the full specs...?

buzzqw
4th November 2009, 11:44
i hope for the 1.9.5 buyer that license upgrade will be free of cost
or.. al least with a significative reduction..

BHH

Cyber-Mav
4th November 2009, 13:23
betaboy will you be doing some benchmarks to show the difference in speed between coreavc 1.9.5, divx latest version and coreavc 2.0. i would like to know if coreavc 2.0 manages to best divx this time.

BetaBoy
4th November 2009, 14:56
betaboy will you be doing some benchmarks to show the difference in speed between coreavc 1.9.5, divx latest version and coreavc 2.0. i would like to know if coreavc 2.0 manages to best divx this time.

To be clear.... CoreAVC's 1.x CORE beats DivX everytime... its only because of CPU instruction set optimizations that they were faster in 'some' cases.

CoreAVC 2.0 beats DivX in every aspect now from our testing. But that's my 'opinion'.... I'll leave it for you all to tell me if its a fact.

BetaBoy
4th November 2009, 15:04
Disabled.... 1.x is EOL and we will not be releasing a new version. 1.x is also fully spec compliant against the features that are included.... there are many features not supported in both CoreAVC and x264 that are being added, and we are tying to match our feature set to that of x264's as they release them. But this will only be in 2.0 and above versions.

Disabled
4th November 2009, 16:02
So "the most advanced directshow H.264 Video decoder in the industry" with "H.264 High profile support" will never be able to decode some high profile features... but you never cared about what you were falsely advertising anyway.

clsid
4th November 2009, 16:08
As long as those 'missing' features are not actually getting used in real life, I do not see much harm in not supporting such obscure features.

BetaBoy
4th November 2009, 16:19
Disabled... this is what, about the 10x you have attacked us in this thread? I thought we were past this OT crap. Mods notified...

STaRGaZeR
4th November 2009, 16:34
Why do you report him if he's right?

Disabled
4th November 2009, 16:41
The harm is done, when you think you bought a decoder for high profile files, while you bought a decoder for files that use 'some' high profile features. You are right, there is no harm done - until you need those features and are not added for free.

And Betaboy, you never could stand criticism about your business practices. I'm sorry to not be the customer who swallows everything. Be happy that with 2.0 I'm not your customer anymore - I'm not.

BetaBoy
4th November 2009, 16:47
The AVC specs today are not the same spec they were 6 months ago no less 2-3 years ago. As far as changes... you are really talking about encoding... and if its encoding, you are talking about x264, and we have been working with Jason and the rest of the devs to ensure our feature set matches that of x264's capabilities.

STaRGaZeR... this has been discussed about 100+ pages back... we are fully open to what is supported and always have been.

Disabled
4th November 2009, 18:38
Not that I know a lot about h264 internals, but something that seems like the specs from 05/2003 (http://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-H.264-200305-S!!PDF-E&type=items) states that baseline profile shall not have weighted_pred_flag but main profile can. I don't know much of what that means, but that flag is exactly what --weightp sets with a gsoc git build of x264. The specs from 2005 (http://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-H.264-200503-S!!PDF-E&type=items) also do not restrict high profile to have that flag.
I may have misinterpreted some of those specs, so tell me if I'm wrong.

ChronoCross
4th November 2009, 19:13
Disabled.... 1.x is EOL and we will not be releasing a new version. 1.x is also fully spec compliant against the features that are included.... there are many features not supported in both CoreAVC and x264 that are being added, and we are tying to match our feature set to that of x264's as they release them. But this will only be in 2.0 and above versions.

So long as your not going to rape me cost wise to update versions this is fine. However for purely clarification purposes you should probably have a full list of all AVC features currently in the spec in a table that indicates what you currently support and what you plan to support in each revision of coreavc. Disabled does have a point in that you do say you support baseline, main, and high with no asterisk that says "only features we deem are in use in the real world are implemented"

BetaBoy
4th November 2009, 19:21
I'll look into a chart... but we do have a list of non-supported features here: http://bit.ly/F1IiS . All of which will soon be moved over to our wiki on CoreCodec.com .

Forteen88
4th November 2009, 19:51
"16 reference frame limitation"
It's awesome that CoreAVC supports up to 16 refs with CUDA, but could you please make the same support for ATI-graphicscards?

EDIT: @popper, yeah I mean hardware assisted decoding.

Dark Shikari
4th November 2009, 19:58
CoreAVC supports weightp just fine. What it doesn't support is duplicating a frame in the reference list with a different weight more than once.

BetaBoy
4th November 2009, 20:25
DS.... Thx... I thought that was assumed with the x264 changes coming.

LoRd_MuldeR
4th November 2009, 22:11
CoreAVC supports weightp just fine. What it doesn't support is duplicating a frame in the reference list with a different weight more than once.

Does libavcodec handle that case? :confused:

Dark Shikari
4th November 2009, 22:28
Does libavcodec handle that case? :confused:Of course it does.

LoRd_MuldeR
4th November 2009, 23:26
Of course it does.

Of course? I still remember there was a time when CoreAVC did support Predictive Lossless (as used by x264), but libavcodec didn't do so :p

But it's good to know that libavcodec (and with it dozens of OSS applications) is 100% ready for weight-p :)

As nobody complained about incomplete weight-p support in CoreAVC until now, is x264 really the first H.264 encoder to make use of that feature ???

Dark Shikari
4th November 2009, 23:51
Of course? I still remember there was a time when CoreAVC did support Predictive Lossless (as used by x264), but libavcodec didn't do so :p

But it's good to know that libavcodec (and with it dozens of OSS applications) is 100% ready for weight-p :)

As nobody complained about incomplete weight-p support in CoreAVC until now, is x264 really the first H.264 encoder to make use of that feature ???No. Ateme's broadcast encoders are known to use multi-dupe weightp (don't call it weightp, regular weightp is used everywhere).

Cyber-Mav
5th November 2009, 01:22
To be clear.... CoreAVC's 1.x CORE beats DivX everytime... its only because of CPU instruction set optimizations that they were faster in 'some' cases.

you cant blame divx for adding cpu instruction set support which coreavc doesnt have implemented yet. my testing showed divx to be faster on the cpu i tested on, cpu had up to ssse3 support.

betaboy do you have a rough idea on what sort of percentage speed difference there is between coreavc 2.0 and 1.9.5?

also iv read that nvidias new cards have purevideo4 support and can do gpu based decoding of mpeg4 video such as xvid files now. is that something that will be added to the cuda support in coreavc2.0? or is coreavc focused on purely h264 decoding?

squid_80
5th November 2009, 01:30
CoreAVC = H264 only. If it handled xvid/divx files it would probably be called CoreASP.

LoRd_MuldeR
5th November 2009, 10:30
i hope for the 1.9.5 buyer that license upgrade will be free of cost
or.. al least with a significative reduction..

Any comment on this? Or did I miss it ???

BetaBoy
5th November 2009, 16:05
I did post it dozens of pages back.. but purchases within the past 60 days of the initial release date includes the 2.0 update. For each of our other 1.x user they will be sent a email with a custom link for a discounted 2.0 purchase.

buzzqw
5th November 2009, 16:12
so only earlier buyer of 1.9.5 wil got 2.0 for free ?

BHH

LoRd_MuldeR
5th November 2009, 16:16
so only earlier buyer of 1.9.5 wil got 2.0 for free ?

As far as I understand, he means the initial relaese date of CoreAVC 2.0.

So if you bought CoreAVC 1.x right before CoreAVC 2.0 was released (the limit is 60 days), you get updated to 2.0 for free. Otherwise you need to pay ;)

What about licenses you were given for free by the CoreCodec team? Will I have to pay this time? :D

buzzqw
5th November 2009, 17:14
bummer.. since there is no eta for 2.0 how i will know that 60 days is kicking in ?

BHH

BetaBoy
5th November 2009, 17:34
purchases within the past 60 days of the initial release date

So 60 days prior to the 2.0 release date... so if we release it on Monday.... users who purchased it 60 days prior, will get the 2.0 update included.

Rumbah
5th November 2009, 18:22
And there will be no fix for 1.9.5 to play back the new x264 weighted p mode in CoreAVC although the results are spec compliant? So I have to pay again to play back a perfectly fine stream?

ranpha
5th November 2009, 18:51
A question of mine is still not answered: Will CoreAVC 2.0 still support Windows XP?

Shinigami-Sama
5th November 2009, 19:47
A question of mine is still not answered: Will CoreAVC 2.0 still support Windows XP?

its a direct show filter... of course it'll work in XP...