[jpsdr]

@maruchan : Problem with this file is totaly unrelated, and now with new fixes recently made in nighty builds :
- cache/fonts is always cleaned-up before each file play. So any file copied in it will be deleted when a file is played.
- Arial is correctly used even if not embemded in the mkv file.
These problems have been solved during the process of solving the ticket i've opened where i've provided a link in my previous post.

@bambi73 : Only this file among the around 7 files i used for testing correct behavior of subtitles failed. This file is very interesting for testing subtitles behavior. Unfortunately, cutting mkv always result in a total mess, and to be realy usefull, sample files are to be provided untouched, so no cut or rebuild.
The correct result, XBMCSetup-20111117-0e174c0-master and XBMCSetup-20111122-89fe561-master both produce is :


The result from your build is :

[sereny]

@jpsdr: Is that the 10bit or 8bit version of the file? If 10 bit, does it run smoothly on your PC? I'm having a few issues with it - that is, about 50% framedrops, while my quadcore still has lots of headroom left.

My problems might not be related to this specific build or 10bit though, as I also get massive framedrops (around 50% again) on very high bitrate (> 50mbit) 8bit files, I'll test more later today. With DXVA2 decoding everything is perfectly smooth though (on 8bit).

[jpsdr]

This file is 8bits. Nevertheless, DXVA2 works only on 8 bits. I haven't checked dropout with bamby73 version, but as i remember play is smooth. I have an i7@870 under Windows XP SP3 (so no DXVA2) but i7@870 is powerfull enough for 1080p + subtitles with lot of features.

[sereny]

@bambi73: Do you use ffmpeg-mt in your build, or can you enable multithreading-flags somehow? It would appear that this is the cause of my dropped frames in very high bitrate videos.

[bambi73]

FFmpeg-mt branch was merged into FFmpeg trunk at the Spring and as far as I understand it it should work "out of the box" (without necessity to change any code). I handled only interface changes between FFmpeg and XBMC and corrected small problems with Hi10P playback. I changed nothing in how XBMC uses FFmpeg library, so there is chance that it's not optimal for FFmpeg-mt now. It'll be task for main developers for migration to evaluate current code and do necessary changes (if any), not for me. I have not enough knowledge for such task

Fyi, FFmpeg guys released new tag with a lot of corrections 2 days ago, so I'll try to assemble and compile it again and we will see if something improve. It all what I can do now with all problems.

[sereny]

Thanks for your insights.
Also, don't feel pressured into doing things you don't have time to do , however, I'll gladly help with testing.

[CrystalP]

bambi73, yes I was asking you to publish the code for GPL compliance.

[Death-Axe]

As I recall, anime release groups have adopted new ways of encoding months before people could play them many times in the past.

No idea why they do this, you would think that if they go to the trouble to make things available for you to watch, that they would use a format people can play. Or is it done for the reason that they only want select groups of people to watch their releases? like some elitist nonsense?

So to get this right: dxva2 can't be used with these hi10 files? so ontop of lack of software player support, my htpc (asrock 330bd) won't be able to play these back at all?

Are these groups only allowing heir stuff to be played on high power machines then? it really does sound like elitist bollocks to me.

I take it you need the same £1000+ machine to re encode these too?

[alexrose1uk]

You don't NEED a high power machine to run them at all, just something that wasn't dated 4+ years ago. I tested a 720p 10bit file on an X2 3800+, one of the oldest and slowest dual core chips on the market *released 2005* and it worked fine. The other week I bought a £70 CPU/Mobo/Cooler kit new from a large sales chain (so that should tell you something about how dated the parts were) as a quick upgrade present for a family member, this also handles 10bit 1080p absolutely fine; the vast majority of anything anywhere near modern should be able to play 10bit ok.

Your Asrock will struggle, simple because its coupled with a CPU that was never, ever aimed at any semi-demanding work at all, it was simply designed to be a low power, netbook/portable chip. Atom's aren't very good at decoding standard 8bit h264 video either, it's just a case they were lucky enough hardware came along they could bolt on to do the job.

It's not like 8bit has gone away either, there's usually someone doing a release or a reencode.

[Death-Axe]

I haven't had any troubles at all with 8 bit playback on my machine. Have yet to test 10 bit due to lack of software support.

[bambi73]

Another week, another build. But this time with alpha version of Ordered chapters support

http://www.megaupload.com/?d=4IZU49RC

Please be aware that this is alpha build (first somehow working version) so you can expect a lot of problems.
Anyway I would appreciate any help with testing because there are so much possibilities how to assemble Ordered chapters in Matroska container that I can't test them all by myself.
When you are reporting your problems, please describe what exactly you did, what file are you playing and attach debug log.

@CrystalIP: I'll post sources/patches soon

[jpsdr]

@bambi73 : Have you been able to check the subtitles problem i've reported ?
I'll test your new version this evening, and report subtitles test and ordered chapters, even if for the last one, the only sample i have is the one i've provided in the ticket i've linked before.

[bambi73]

Not yet, I have it in my list but with lower prio. Anyway I'm not sure how much I'll be able to help, because I didn't changed anything close to subtitles engine. But I'll check it when I have some free time
About new build, it's based on 1+ month old XBMC sources, so don't expect more than from last Hi10P build. It's only for testing purposes.

[sereny]

I just tested my collection against the new release and sent my results/debuglogs to you via PM, bambi. I've tested 9 files with ordered chapters, 5 play fine, 2 have some errors, 2 won't play at all. Things look quite promising.

@Death-Axe:
Fansubbing has always been pushing the limit. You can call them elitists (and by god, yes, often there is heavy autism at work, but you can impossibly deny the pure technical brilliance of many of the fansubbing works.)

Without fansubbing we'd probably still be on mp4/ASP in .avi for everything.


About the 10bit release of coalgirls' bakemono: This is the sickest release ever. Due to the heavy subtitling, my q6600 is maxed out to 90% on all 4 cores, using mpc-hc with the heavily optimized xy-vsfilter.

I am pointing this out because these (animated!) subtitles massively interfere with the whole performance/framedrop-issue, which could skew the perception of bambis hi10p playback-implementation. In jpsdrs screenshots, only the top one is correct, and also the one which is massively CPU-intensive. The broken one at the bottom happens if animation is disabled and does not need as much CPU.

[jpsdr]

Bakemono coalgirl i've provided is 8 bits !!!!!. I use this file for subtitles testing purpose only. This file is very good to check subtitles implemented features and timing accuracy.

Watching almost only anime, i know sometimes subtitles feature are cpu intensive, this is why i've choosen to build my multimedia PC with an i7@870. With this, this file plays perfectly fine on xbmc, and CPU only, because i've Windows XP SP3, so no DVXA2.

Yes, anime fansubers are always seeking for the best of the best, without any consideration for any compliancy. Since more than a year, a lot of 1080p releases were at level5.1, breaking compliancy with a lot hardware players which are only level4.1 (except Boxee wich is pure SW decoding), but now with the 10bits it's over. They absolutely don't care about compliancy with HW players or peoples who have HW players. On the other hands, they are doing this freely and on their good will, so, in a way, they have the rigth to do what they want...