Doom10 Forum: Digital Video Discussion

No clue about 'generally accepted download site', but these to work:
http://komisar.gin.by/index.html
http://x264.fushizen.eu/

okay, solved -> http://forum.doom9.org/showpost.php?p=1630670&postcount=10

You are probably right.

afaik Mac OS starting with 10.6 directly ships with the OpenCL, but no clue about what needs to be loaded,...

when calling ls -R * on /System/Library/Frameworks/OpenCL.framework/ I get:

CodeResources   OpenCL

Frameworks:

Headers:
cl.h      cl_ext.h   cl_gl.h      cl_gl_ext.h   cl_platform.h   opencl.h

Libraries:
AMDil.dylib         PTX.dylib         libCLVMCPUPlugin.dylib      libcl2module.dylib      libclparser.dylib
ImageFormats         libCLVMAMDilPlugin.dylib   libCLVMPTXPlugin.dylib      libcldcpuengine.dylib

Libraries/ImageFormats:
float_rgba.dylib   half_rgbx.dylib      sint32_rgba.dylib   snorm8_rgba.dylib   uint8_rgba.dylib   unorm8_a.dylib      unorm8_i.dylib
float_rgbx.dylib   sfixed14_rgba.dylib   sint8_rgba.dylib   uint16_rgba.dylib   unorm16_rgba.dylib   unorm8_argb.dylib   unorm8_rgba.dylib
half_rgba.dylib      sint16_rgba.dylib   snorm16_rgba.dylib   uint32_rgba.dylib   unorm16_rgbx.dylib   unorm8_bgra.dylib   unorm8_rgbx.dylib

Resources:
Info.plist      cl_conversions.h   cl_kernel.h      cl_kernel_shared.h   runtime.amdil.bc   runtime.i386.bc      runtime.ptx.bc      runtime.x86_64.bc   version.plist

Versions:
A   Current

Versions/A:
CodeResources   Frameworks   Headers      Libraries   OpenCL      Resources   _CodeSignature

Versions/A/Frameworks:

Versions/A/Headers:
cl.h      cl_ext.h   cl_gl.h      cl_gl_ext.h   cl_platform.h   opencl.h

Versions/A/Libraries:
AMDil.dylib         PTX.dylib         libCLVMCPUPlugin.dylib      libcl2module.dylib      libclparser.dylib
ImageFormats         libCLVMAMDilPlugin.dylib   libCLVMPTXPlugin.dylib      libcldcpuengine.dylib

Versions/A/Libraries/ImageFormats:
float_rgba.dylib   half_rgbx.dylib      sint32_rgba.dylib   snorm8_rgba.dylib   uint8_rgba.dylib   unorm8_a.dylib      unorm8_i.dylib
float_rgbx.dylib   sfixed14_rgba.dylib   sint8_rgba.dylib   uint16_rgba.dylib   unorm16_rgba.dylib   unorm8_argb.dylib   unorm8_rgba.dylib
half_rgba.dylib      sint16_rgba.dylib   snorm16_rgba.dylib   uint32_rgba.dylib   unorm16_rgbx.dylib   unorm8_bgra.dylib   unorm8_rgbx.dylib

Versions/A/Resources:
Info.plist      cl_conversions.h   cl_kernel.h      cl_kernel_shared.h   runtime.amdil.bc   runtime.i386.bc      runtime.ptx.bc      runtime.x86_64.bc   version.plist

Versions/A/_CodeSignature:
CodeResources

may be this helps.

Cu Selur

Ps.: btw. seems like the 'notification'-feature of the forum isn't working.

using:
./configure --disable-lavf --enable-staticI get:
platform:      X86_64
system:        MACOSX
cli:           yes
libx264:       internal
shared:        no
static:        yes
asm:           yes
interlaced:    yes
avs:           avxsynth
lavf:          no
ffms:          no
gpac:          no
gpl:           yes
thread:        posix
opencl:        yes
filters:       resize crop select_every
debug:         no
gprof:         no
strip:         no
PIC:           no
visualize:     no
bit depth:     8
chroma format: all

You can run 'make' or 'make fprofiled' now.
it fails with: http://pastebin.com/G1sjBEj6
if I use --disable-opencl compilation works fine.
Is there a way to compile x264 with opencl support on Mac OS X 10.6.8 or does it require a newer Mac OS X version?

Cu Selur

What do you command line calls look like?
And yes decoding with ffmpeg and piping to x264 for encoding can work fine.

1. ffmpeg can use x264 and there should be no difference between ffmpegs and x264 output if ffmpeg is using x264 as encoder; I'm not even sure if libav still contains a functional H.264 encoder on it's on or if it has switched to x264 completely,...
2. I prefer qaac because it's output sounds better to me, but there are probably some more objective audio codec comparisons out there (iirc the hydrogenaudio community did some aac listening tests where both qaac and nero's aac encoder were part of)

--cabac --ref 16 --deblock -1:-1 --partitions i4x4,i8x8,p8x8,b8x8 --me umh --subme 6 --psy-rd 1.0:0.0 --merange 16 --trellis 2 --8x8dct --cqm jvt --deadzone-inter 21 --deadzone-intra 11 --qblur -4 --threads 6 --bframes 8 --b-pyramid normal --b-adapt 2 --b-bias 0 --direct spatial --keyint 250 --min-keyint 25 --scenecut 40 --crf 18.0 --qcomp 0.60 --qpmin 10 --qpmax 51 --qpstep 4 --ipratio 1.40 --pbratio 1.30 --aq-mode 1 --aq-strength 1.00should be a conversion the encoding settings (aside from zones=202334,207820,q=70) to the normal x264 command line syntax.
Not sure if it helps unless MeGui has some sort of "import x264 cl settings" option.

About the concatenating: I would demux both streams and then append them through mkvmerge/mmg. (I would not try to concatenate the raw H.264 streams.)

I agree that in theory such a stream can be created, but I doubt that you can encounter a natural occurring stream which, if you aim for a specific file size, will create the same output through 2pass and 1pass encoding. (assuming you do not find a way to cripple the 2pass settings in a way that it is basically a 1pass encoding)

Is there some reasonable scenario to do that?

sure, if you capture it 1st just to have it, but later need it at a specific size.

Normally you use:
1. crf if you don't need a specific file size
2. 2pass if you need a specific file size

Cu Selur

if you also want to reencode the audio, you could:
1. load the audio source in Avisynth
2. AudioDub(video,audio)
3. trim the audio alongside the video

otherwise you will probably have to calculate the cut times and cut the audio with the audio editor of your choice.

Psy RDO measures, in addition to normal PSNR-wise distortion, the difference in complexity between the reconstructed frame (what the video will look like on decoding) and the source frame. It values complexity values as close to that of the original frame as possible. Thus, it strongly biases against blurring, and for that matter, any significant loss of detail. Even on non-grainy sources, this generally results in a reduction of banding, increased sharpness, and better retention of fine details. The speed cost is relatively low, though Psy RDO requires regular RD to be enabled.

source: http://x264dev.multimedia.cx/archives/37

http://forum.doom9.org/showthread.php?t=138293 might also be interesting to read.

I would try a more simple call, e.g.:
x264 --crf 19 --vf crop:0,132,0,132 --input-res 1920x1080 --preset slow --tune film  --fps 24000/1001 -o Looper.2012.mkv Looper.mkv
other than that 'No clue' (assuming your system has no heat problems etc.)

[h264 @ 0x2b0d580] number of reference frames (0+5) exceeds max (4; probably corrupt input), discarding one
Segmentation fault fps, 5173.71 kb/s

That sound like your input is corrupt (or lav can't decode it properly) at one point. -> tried another source?
Since you mentioned you convert blu-ray content, I hope you used a decent ripper to copy the content to your hdd, since x264 won't be able to circumvent any encryption / copy protection.
Btw. Why are you using eac3to to demux the video before encoding and not feed the mkv (makemkv is a good tool to do Blu-ray/DVD to mkv and aside from the WIndows version it was a Linux and a Mac version aside directly to x264?

"--fps 24000/1001p"
are you sure the 'p'- at the end is right? (never say this syntax except in mkvmerge)