Current Patches, Where to get them, How they affect speed/output

[JEEB]

1190 builds rebuilt, re-uploaded (all but the unmodified one, which had no patches, so (ry)

[Dark Shikari]

Originally Posted by JEEB

1190 builds rebuilt, re-uploaded (all but the unmodified one, which had no patches, so (ry)

Sorry, but your efforts were in vain!

[MythCreator]

x264 rev.1192 unpatched x86
download

built on Jul 28 2009, gcc: 4.4.0 (release) (TDM-GCC)

fprofiled

[juGGaKNot]

What setting does the --tune fastdecode have ?

ahh nvm

+ else if( !strcasecmp( optarg, "fastdecode" ) )

+ {

+ param->b_deblocking_filter = 0;

+ param->b_cabac = 0;

+ param->analyse.b_weighted_bipred = 0;

+ }

Originally Posted by MythCreator

x264 rev.1192 unpatched x86
download

built on Jul 28 2009, gcc: 4.4.0 (release) (TDM-GCC)

fprofiled

Try filefront/mediafire ?

Originally Posted by MythCreator

hmm..I'll upload to FileFront

I found the download, well hidden!

[Audionut]

http://mailman.videolan.org/pipermai...ly/006096.html

+ else if( !strcasecmp( optarg, "fastdecode" ) )
+ {
+ param->b_deblocking_filter = 0;
+ param->b_cabac = 0;
+ param->analyse.b_weighted_bipred = 0;
+ }

[MythCreator]

Originally Posted by juGGaKNot

Try filefront/mediafire ?

hmm..I'll upload to FileFront




x264 rev 1192 patched build

Download

built on Jul 28 2009, gcc: 4.4.0 (release) (TDM-GCC)

fprofiled

patched with:
x264_win_zone_parse_fix_05.diff
x264_hrd_pulldown.15_interlace.diff

[gator1102]

http://xeph.textcube.com/2/attach/x264.exe
x264 rev.1192 _ 32-bit binary with MP4 output

Intel C++ Compiler (icc) v11.1.038

Used patches:

x264_icc_03_win.diff
x264_win_zone_parse_fix_05.diff
x264_hrd_pulldown.15_interlace.diff
x264-r1183-threaded-slicetype-v17.diff

[techouse]

x264_x64_r1192_unpatched | MD5
GCC 4.4.0 20090524 (x64.core2.Komisar), unpatched, generic, fprofiled

________________________________________________________________________________

x264_x86_r1192_techouse | INFO
GCC 4.4.0 20090524 (x86.core2.Komisar), fprofiled, -march=core2

x264_x64_r1192_techouse | INFO
GCC 4.4.0 20090524 (x64.core2.Komisar), fprofiled, -march=core2

Patches used:

x264_hrd_pulldown.15_interlace.diff
x264_win_zone_parse_fix_05.diff

[techouse]

x264_x64_r1193_unpatched | MD5
GCC 4.4.0 20090524 (x64.core2.Komisar), unpatched, generic, fprofiled

________________________________________________________________________________

x264_x86_r1193_techouse | INFO
GCC 4.4.0 20090524 (x86.core2.Komisar), fprofiled, -march=core2

x264_x64_r1193_techouse | INFO
GCC 4.4.0 20090524 (x64.core2.Komisar), fprofiled, -march=core2

Patches used:

x264_hrd_pulldown.15_interlace.diff
x264_win_zone_parse_fix_05.diff

[JEEB]

x264 r1193 64bit unpatched:
download ; hash

• built on Jul 28 2009, gcc: 4.3.4 20090220 (prerelease) (x64.generic.Komisar)
• fprofiled, otherwise defaults

________________________________________________________________________________

x264 r1193 32bit
download ; release notes

• built on Jul 28 2009, gcc: 4.3.4 20090220 (prerelease) (x32.generic.Komisar)
• fprofiled, -march=i686

x264 r1193 64bit
download ; release notes

• built on Jul 28 2009, gcc: 4.3.4 20090220 (prerelease) (x64.generic.Komisar)
• fprofiled, -march=core2

patched with:

• x264_win_zone_parse_fix_05.diff
• x264_hrd_pulldown.15_interlace.diff

My help is probably unneeded, but since this is what my building script outputs nowadays, I'll be using this base I guess

[mopurist]

Originally Posted by Trahald

Hrd 15.

Sorry, maybe I misunderstand, but is this version supposed to fix the error described in post 2036?

I just recompiled x264 with this patch and I get the same failure. (Linux 64-bit, if it matters)

x264: encoder/set.c:644: x264_sei_picture_timing_write: Assertion `dpb_output_delay < pow( 2, sps->vui.nal_hrd_parameters.i_dpb_output_delay_length )' failed.

Thanks.

[Trahald]

whats your command line.

[imk]

x264-r1193M-imk-win.7z
win_build_info.txt

[mopurist]

Originally Posted by Trahald

whats your command line.

When I run with the same options XhmikosR described, it does work. I hadn't updated x264 for several months, and now that I have, a script I used previously without issue dies with the assertion failed.

If I trim the command used by the script down as far as the following, the assertion still fails:

/usr/bin/wine /home/mopurist/.wine/drive_c/Program\ Files/avs2yuv.exe /home/mopurist/test.avs -raw - | x264 --bitrate 7000 --level 4.1 --nal-hrd --vbv-maxrate 50000 --vbv-bufsize 50000 --output test.264 - 1440x1080 --fps 29.970

The actual command the script uses is:

/usr/bin/wine "${avs2yuvdir}/avs2yuv.exe" "${workdir}/${1}.avs" -raw -| /usr/bin/x264 --b-adapt 2 --pass 1 --psy-rd 1.0:0.2 --aq-strength 1.3 --bitrate $2 --stats "${workdir}/${1}.stats" --level 4.1 --nal-hrd --ref 4 --mixed-refs --bframes 4 --b-pyramid --no-fast-pskip --weightb --me umh --direct auto --keyint 300 --min-keyint 30 --vbv-maxrate 62500 --vbv-bufsize 62500 --mvrange 511 --aud --subme 7 --trellis 2 --analyse p8x8,b8x8,i4x4,i8x8 --8x8dct --threads auto --thread-input --deblock 0,1 --sar 4:3 --output "${twosixfourdir}/${1}_final.264" - 1440x1080 --fps 29.970

So, obviously I don't know what the assertion is testing, nor do I understand how it is determined or what the purpose of the change is, that now results in an assertion failed.

Any help in understanding and fixing the issue would be appreciated.

Thanks.

[ACoolie]

In a vanilla icc-compiled x264 r1193 build i get the following error around 300 frames into the second pass of an encode. CFLAGS and LDFLAGS are just "-openmp."

x264: encoder/analyse.c:1175: x264_mb_analyse_inter_p16x16: Assertion `a->l0.me16x16.mv[1] <= h->mb.mv_max_spel[1] || h->param.i_threads == 1' failed.

x264 -B 1264 -o 1.h264 -p2 --merange 24 --level 4.1 --fps 24000/1001 --sar 32/27 -b 12 -f -2,-2 --frames 52560 --psy-rd 1.3:0.15 --aq-strength 1.0 --aq-mode 2 --stats .stats - 708x348 --preset slower --no-fast-pskip --b-pyramid -m10
A gcc-compiled x264 works perfectly. If I set subme to 9 and use the icc build, it works as well.

[Dark Shikari]

.... I'm a complete idiot, it seems... adding more subme values and not increasing the size of the array... I'm shocked it worked at all.

Fix committed.

[Dark Shikari]

OK, now this is brilliant.

static const int subpel_iterations[][4] =
   {{0,0,0,0},
    {1,1,0,0},
    {0,1,1,0},
    {0,2,1,0},
    {0,2,1,1},
    {0,2,1,2},
    {0,0,2,2},
    {0,0,2,2},
    {0,0,4,10},
    {0,0,4,10}};

/* (x-1)%6 */
static const int mod6m1[8] = {5,0,1,2,3,4,5,0};

Accessing the array positions subpel_iterations[10][2] and subpel_iterations[10][3] gave "1,2", a reasonable enough hpel/qpel value, which resulted in me never noticing the problem.

Of course, on a compiler that didn't happen to arrange the data adjacently in memory...

[Trahald]

@mopurist
Hmm... i couldnt duplicate the issue, but i'll keep trying.

Except for wasting a few bits.. rev 13 of the patch is fine to use.

[techouse]

x264_x64_r1195_unpatched | MD5
GCC 4.4.0 20090524 (x64.core2.Komisar), unpatched, generic, fprofiled

________________________________________________________________________________

x264_x86_r1195_techouse | INFO
GCC 4.4.0 20090524 (x86.core2.Komisar), fprofiled, -march=core2

x264_x64_r1195_techouse | INFO
GCC 4.4.0 20090524 (x64.core2.Komisar), fprofiled, -march=core2

Patches used:

x264_hrd_pulldown.15_interlace.diff
x264_win_zone_parse_fix_05.diff

[MythCreator]

x264_x86_r1195_MythCreator

Download

GCC 4.4.1 released (MythCreator build), fprofiled

Patches used:

x264_hrd_pulldown.15_interlace.diff
x264_win_zone_parse_fix_05.diff