Strange crashes

[InTrancer]

i have this script
LoadPlugin("D:\Program Files\dgavcdec\DGAVCDecode.dll")
AVCSource("H:\Share\ef - a tale of melodies. 1(BD)\BD1\BDMV\STREAM\s01.dga", deblock=false)
SetMTMode(2,6)

Crop(4, 4, -4, -4)

mt("lanczos4resize(1920,last.height)", threads=4)
mt("lanczos4resize(last.width,1080)", threads=4, splitvertical=true)

AddBorders(0, 0, 0, 8)
and this encode settings
program --crf 20.0 --ref 16 --mixed-refs --no-fast-pskip --bframes 16 --b-adapt 2 --b-pyramid --weightb --direct auto --subme 9 --trellis 1 --psy-rd 0.0:0 --partitions all --8x8dct --me umh --merange 32 --threads auto --thread-input --aud --progress --no-dct-decimate --output "output" "input"
And after ~50 frames x264 crashes
if edit my script to
LoadPlugin("D:\Program Files\dgavcdec\DGAVCDecode.dll")
AVCSource("H:\Share\ef - a tale of melodies. 1(BD)\BD1\BDMV\STREAM\s01.dga", deblock=false)
all work fine, but i need crop ...

[vucloutr]

i guess you have not enough RAM. (just a wild guess ;D)

maybe check it with taskmanager..

[InTrancer]

~3.25Gb of RAM on Windows XP SP3

[kemuri-_9]

what happens if you put in

SetMemoryMax(512)

?

[InTrancer]

SetMemoryMax(256)

Help me

[ggf31416]

check your RAM with memtest86+ http://www.memtest.org/

[talen9]

What's the revision of x264 you're using?

There was a regression in the revisions from r1066 (included) to r1074 (fixed here), IIRC, this is the log of the fix:

With some combinations of video width and other settings, the scratch buffer was slightly too small.

This caused heap corruption on some systems.

I don't know for sure that you're being hit by this ... but the fact is that 1920x1080 is not a mod16 resolution (I think x264 pads it internally to 1920x1088).

Try upgrading x264 to the last rev (r1080 as of now).

[InTrancer]

ggf31416
~9 hours with 0 errors

talen9
1077 and 1080, now 1080 crashes often in the end of encode (example of script in the bottom of post)
i encode always with mod16 resolution
LoadPlugin("D:\Program Files\DGIndex\DGDecode.dll")
mpeg2source("H:\**\s12.d2v", cpu=0)
tfm(d2v="H:\***\s12.d2v", PP=6, mode=5, display=false, MI=80, slow=2)
tdecimate(mode=1)
SetMTMode(2,6)
Crop(0,2,-0,0)
mt("lanczos4resize(848,last.height)", threads=4)
mt("lanczos4resize(last.width,480)", threads=4, splitvertical=true)

[Aktan]

Are you encoding to MP4 container? There might be a bug in muxing to MP4 (unconfirmed) but if you are outputing to MP4, try outputing to MKV instead


I love that anime!

[InTrancer]

in mkv

[Snowknight26]

Run a debug build of x264 and get a backtrace with gdb?

[Sagekilla]

On a side note, is there a reason you're using a resizer inside of MT()? All the resizers built in are so fast it's absurd to even need to thread it.

[InTrancer]

Snowknight26
where i can download it?

[LoRd_MuldeR]

Snowknight26
where i can download it?

You can make a debug build of x264 yourself easily. Just pass the "--enable-debug" switch to the configure script.

x264 sources:
http://git.videolan.org/gitweb.cgi?p=x264.git;a=tree;

MinGW GCC:
http://www.tdragon.net/recentgcc/

GNU Source-Level Debugger:
http://sourceforge.net/project/showfiles.php?group_id=2435&package_id=20507

[kemuri-_9]

Snowknight26
where i can download it?

x264_debug.exe (http://kemuri9.net/dev/x264/x264_debug.exe)

follow LoRd_MuldeR's link for GDB...

[InTrancer]

about first post
[New thread 10180.0x227c].57 fps, 607.98 kb/s, eta 6:00:06
[New thread 10180.0x17f8]

Program received signal SIGSEGV, Segmentation fault.
x264_malloc (i_size=9142272) at common/common.c:714
714 common/common.c: No such file or directory.
in common/common.c

[Dark Shikari]

That means malloc failed.

We really should add a test for that in x264_malloc as otherwise it really makes CHECKED_MALLOC pointless on platforms that need the emulated memalign... of course, x264 can't do anything except die if malloc fails.