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

[gav1577]

Thanks LoRd_MuldeR much appreciated but one thing i am unclear of is there cmd switch or is the patch automatically activated ?

[Atak_Snajpera]

Because people like to complain:
x264.859.modified.tenbillion (how many 859 builds do we have now???)

Patches:
hrd pulldown
fix win stdin
meprepass
gxRC
x264_progress
nr fix
gaussian cplxblur
2pass vbv
new RDO

Enjoy

wizboy11 could you compile with latest psy patch 0.22?

[TheRyuu]

Originally Posted by Atak_Snajpera

wizboy11 could you compile with latest psy patch 0.22?

I was already on it before you asked >_>

------------------------------------------

x264.861.modified.exe
(Alt Download Site)

Patches:
x264_progress.diff
x264.gaussian.cplxblur.01.diff
x264_hrd_pulldown.04_interlace.diff
x264_fix_win_stdin.diff
x264_me-prepass_DeathTheSheep.01.diff
x264_psyRDO_02.diff
gxRC.diff <---DTS 02 patch but aligned

avis input, mp4 output, pthreads activated upon configure. (statically linked pthreads)
Built with GCC 3.4.5, fprofiled.

Couple notes:
vbv has been integrated into git as well as the nr threads fix.

New option : --rdcmp
Usage:
--rdcmp psy (psy rdo)
--rdcmp ssd (regular)
Psy RDO is on by default. at subme levels 6 and 7. 7 + b-rdo is best.
Read psyRDO thread for more info regarding it.

me-prepass:
activated with --me-prepass
off by default

gxRC:
Activated at subme levels 6 and 7.

[Atak_Snajpera]

Thanks!

[survivant001]

Originally Posted by wizboy11

Couple notes:
vbv has been integrated into git as well as the nr threads fix.

that's mean that we wont' see vbv_0x.. patch anymore because they are in the common branch right ?

just to know because I'm only using 2 passes.

[TheRyuu]

Originally Posted by survivant001

that's mean that we wont' see vbv_0x.. patch anymore because they are in the common branch right ?

just to know because I'm only using 2 passes.

The patch has been integrated into the git so you won't see the patch in the list anymore because it's already there

[gav1577]

Nice 1 wizboy11

[TheRyuu]

x264.869.modified.exe
(alt DL link)

x264 Rev. 869 Modified Patch Build
Avis Input, mp4 output, pthreads config.

Built with GCC 3.4.5, make fprofiled, and statically linked pthreads. (means no pthreads.dll required in x264 directory)

Patches:
x264_gxRC.diff
x264.gaussian.cplxblur.01.diff
x264_hrd_pulldown.04_interlace.diff
x264_me-prepass_DeathTheSheep.01.diff
x264_progress.diff
x264_psyRDO_0.22.diff

The x264_fix_win_stdin.diff patch has been committed into the git so that has been left out.

Notes:
See the psyRDO thread on usage of it although I briefly explained it in an above post.
Option: --rdcmp
Usage:
--rdcmp psy (psy rdo)
--rdcmp ssd (regular)
Psy RDO is on by default. at subme levels 6 and 7. 7 + b-rdo is best.
Read psyRDO thread for more info regarding it.

me-prepass:
Use --me-prepass flag to activate, off by default
Runs a half-pel search at a slight speed cost to improve quality with any --me mode.

x264_gxRC.diff:
Activated at subme levels 6 and 7, basically increases quality at little to no speed reduction.

x264_progress.diff:
Adds a little progress thingy at the top of the x264 window. Neat feature.

[bokonon]

Thanks wizboy11

Originally Posted by wizboy11

x264.869.modified.exe
(alt DL link)

x264_gxRC.diff:
Activated at subme levels 6 and 7, basically increases quality at little to no speed reduction.

I've done a search for this and i get no results other than your posts in this thread. What does it really do/where can i read more?

[TheRyuu]

Originally Posted by bokonon

Thanks wizboy11

I've done a search for this and i get no results other than your posts in this thread. What does it really do/where can i read more?

gxRC actually isn't worth it after talking with some x264 devs. The speed/quality tradeoff just isn't worth it.

So it will be removed from later builds. It's in this thread, just look back if you want to learn a little about it I think.

[TheRyuu]

New build, basically just changing the patches (using less).

x264.869.modified.02.exe

Patches:
x264_progress.diff
x264.gaussian.cplxblur.01.diff
x264_hrd_pulldown.04_interlace.diff
x264_me-prepass_DeathTheSheep.01.diff
x264_psyRDO_02.diff

Notes:
Removed gxRC patch because there's a reason it wasn't committed into the git. (speed/quality trade off totally not worth it)

x264_me-prepass_DeathTheSheep.01.diff:
use the --me-prepass flag to activate the half pel search. Improves quality using any --me mode but isn't really worth it until you get up to --me esa.

x264.gaussian.cplxblur.01.diff:
improves 2 pass rate control

x264_psyRDO_02.diff:
see thread
Option: --rdcmp
Usage:
--rdcmp psy (psy rdo)
--rdcmp ssd (regular)
Psy RDO is on by default. at subme levels 6 and 7. 7 + b-rdo is best.
Read psyRDO thread for more info regarding it.

x264_progress.diff:
Adds a little progress thingy at the top of the x264 window. Neat feature.

[gav1577]

Hi how much of a difference can --me-prepass make? basically is there a noticeable difference because it slows things down quite a bit for me and i cant see any difference with it turned on or off. is it more effective for lower bitrates ?

[Dark Shikari]

Originally Posted by gav1577

Hi how much of a difference can --me-prepass make? basically is there a noticeable difference because it slows things down quite a bit for me and i cant see any difference with it turned on or off. is it more effective for lower bitrates ?

No it isn't very useful. The benefit is on par with --me umh -> esa or something like that.

[gav1577]

Thanks for the help Dark Shikari

[lexor]

//snip from git for gaussian blur patch
for(j=0; j<=cplxblur*2 && j<=i; j++){
    ratecontrol_entry_t *rcj = &rcc->entry[i-j];
    double gaussian_weight = weight * exp(-j*j/200.0);

Are we trusting gcc to take that double definition out of the loop? I know early 4.x's couldn't be always relied on to do that (though I encountered it with larger objects, not primitives). Even though if it stays in there performance shouldn't suffer measurably, still seems like bad form.

[Dark Shikari]

Originally Posted by lexor

//snip from git for gaussian blur patch
for(j=0; j<=cplxblur*2 && j<=i; j++){
    ratecontrol_entry_t *rcj = &rcc->entry[i-j];
    double gaussian_weight = weight * exp(-j*j/200.0);

Are we trusting gcc to take that double definition out of the loop? I know early 4.x's couldn't be always relied on to do that (though I encountered it with larger objects, not primitives). Even though if it stays in there performance shouldn't suffer measurably, still seems like bad form.

Its actually better for GCC to define a variable as close as possible to where its used. But it doesn't matter here; the main problem was a C99-ism in code a few lines later, and your issue as been resolved in the same patch, r871.

[MasterNobody]

Collection of patches that I use for building x264vfw (useful not only for VfW): bm_x264_patch_collection.r870.zip
List of independent patches:
x264_32x32samples_crash.r870.diff
x264_cosmetic.r870.diff
x264_debug_defines.r870.diff
x264_fix_stats_file_work.r870.diff
x264_frames_memoryleak.r870.diff
x264_multithreading_Nth_pass_ratecontrol.r870.diff
x264_psy_rdo.r870.diff
x264_thread_pool.r870.diff
x264_vaq2mod.03.r870.diff

Patches that depends of others:
bm_x264_thread_pool.r870.diff depends of x264_fix_stats_file_work.r870.diff, x264_frames_memoryleak.r870.diff, x264_multithreading_Nth_pass_ratecontrol.r870.diff
bm_x264_vfw2cli.r870.diff depends of x264_fix_stats_file_work.r870.diff

All in one patches:
bm_x264_cli_VAQ2mod_PsyRDO_fixed.r870.diff
bm_x264_exp_VAQ2mod_PsyRDO_fixed.r870.diff
bm_x264_normal_fixed.r870.diff

Description of the patches here.

CLI-version build with this patches here: x264_CLI_870bm_VAQ2mod_PsyRDO_fixed.zip
Build with GCC (GCC TDM-3 for MinGW) 4.3.0 and YASM 0.7.1.2093
Configuration:

Platform:   X86
System:     MINGW
avis input: yes
mp4 output: yes
pthread:    yes
gtk:        no
debug:      no
gprof:      no
PIC:        no
shared:     no
visualize:  no

[Razorholt]

Thanks a lot MasterNobody!

I don't see me-prepass patch listed here. What is your opinion on this patch? Apparently it has been "fixed" but I don't know what that means since I can't test it.

[TheRyuu]

Originally Posted by Razorholt

Thanks a lot MasterNobody!

I don't see me-prepass patch listed here. What is your opinion on this patch? Apparently it has been "fixed" but I don't know what that means since I can't test it.

me-prepass is only really worth it when using --me esa or higher. It's not even that useful, tesa is probably just better.

-------------------------------------------

x264.871.modified.exe

GCC 3.4.5 make fprofiled build.

Patches:
x264_hrd_pulldown.04_interlace.diff
x264_me-prepass_DeathTheSheep.01.diff
x264_psyRDO_02.diff
x264_progress.diff

More git commits means less patches. Guassian blur was committed so that's off the list now too.

x264_me-prepass_DeathTheSheep.01.diff:
use the --me-prepass flag to activate the half pel search. Improves quality using any --me mode but isn't really worth it until you get up to --me esa. Isn't really worth it since --me tesa will provide better motion searching. But since it has a flag and it's default is off I include it.

x264_psyRDO_02.diff:
see thread
Option: --rdcmp
Usage:
--rdcmp psy (psy rdo)
--rdcmp ssd (regular)
Psy RDO is on by default. at subme levels 6 and 7. 7 + b-rdo is best.
Read psyRDO thread for more info regarding it.

x264_progress.diff:
Adds a little progress thingy at the top of the x264 window. Neat feature.

[gav1577]

Thanks wizboy11