Subme 11 - Implemented or not

[legal]

Hi, I just read this tread "https://forum.doom9.org/showthread.php?t=166808" and I want to ask if "subme 11" is implemented or not? I'm confused... some people say that is working and others say that is not working only consume processor cycles for nothing!

Best Regards.

[Groucho2004]

All you have to do is look at the documentation (x264 --fullhelp):
-m, --subme <integer> Subpixel motion estimation and mode decision [7]
- 0: fullpel only (not recommended)
- 1: SAD mode decision, one qpel iteration
- 2: SATD mode decision
- 3-5: Progressively more qpel
- 6: RD mode decision for I/P-frames
- 7: RD mode decision for all frames
- 8: RD refinement for I/P-frames
- 9: RD refinement for all frames
- 10: QP-RD - requires trellis=2, aq-mode>0
- 11: Full RD: disable all early terminations

Whether the additional CPU cycles/encoding time are worth it is up to you to decide.

In my humble opinion, it's a waste, I never use more than 9.

[legal]

Hi "Groucho2004", I know the option is there but if you read this thread "https://forum.doom9.org/showthread.php?t=166808" at some point they wrote this:

"About subme 11 and treillis 3 : According DS (it was a "long" time ago), treillis 3 will probably never be released, it was something very experimental, and had probably not given the expected results => subme 11 has actualy no real purpose. "

This is true or not? The subme 11 is there but it does not do anything (only consume cycles from processor for nothing)?

[Groucho2004]

I know the option is there but if you read this thread "https://forum.doom9.org/showthread.php?t=166808" at some point they wrote this:

"About subme 11 and treillis 3 : According DS (it was a "long" time ago), treillis 3 will probably never be released, it was something very experimental, and had probably not given the expected results => subme 11 has actualy no real purpose. "

This is true or not? The subme 11 is there but it does not do anything (only consume cycles from processor for nothing)?
I don't know what DS wrote a long time ago. However, the placebo preset is the only one using subme 11 which reveals its purpose: Burning CPU cycles and heating the room.

[LoRd_MuldeR]

As the description says "disable all early terminations".

The idea of an "early termination" is that you stop an optimization process at the point where further improvements are very unlikely. So, you can save a lot of CPU time for little or no loss in compression efficiency. Disabling the "early termination" means that you might get a small improvement in a few cases, but probably not enough to be relevant. At the same time, since the optimization process now is forced to always run to the very end, you always "burn" additional CPU cycles.

BTW: Number 11 (https://www.youtube.com/watch?v=4xgx4k83zzc) probably was not chosen coincidentally ;)

[legal]

I don't know what DS wrote a long time ago. However, the placebo preset is the only one using subme 11 which reveals its purpose: Burning CPU cycles and heating the room.

lol.......
But I want to know if it gives any advantage (even minimal) using it (subme 11) or not (like it gives 1% or 2% more quality to the encode)?
Running the encode with this option activated to heating the room is bad idea it takes too much time... ;)

[legal]

As the description says "disable all early terminations".

The idea of an "early termination" is that you stop an optimization process at the point where further improvements are very unlikely. So, you can save a lot of CPU time for little or no loss in compression efficiency. Disabling the "early termination" means that you might get a small improvement in a few cases, but probably not enough to be relevant. At the same time, since the optimization process now is forced to always run to the very end, you always "burn" additional CPU cycles.

BTW: Number 11 (https://www.youtube.com/watch?v=4xgx4k83zzc) probably was not chosen coincidentally ;)

Thanks "LoRd_MuldeR" for your answer... ;)

[jpsdr]

Yes, historical a long time ago, subme 11 was introduced only for the purpose of a new treillis mode 3, which was never released. Before it, placebo was with 10...
And when i've asked DS about it, it provided the answer you told in post #3. So, as this treillis mode 3 was never released, even if mode 11 "do something", it realy has no purpose actualy.

[mandarinka]

The demise of --me trellis was such a shame, although now we are switching over to x265, so maybe it ceased being important.

The patches/git should still be available somewhere though, in case somebody is interested in it.

There was also the K-means based --weightp 3. Too bad that one didn't materialize.

[jpsdr]

There was also the K-means based --weightp 3. Too bad that one didn't materialize.
This one is used in the 7mod release.
I've tried it, it seems to work in crf one pass mode, but it screws the ouput in vbr multi-pass (3 passes) mode (unless i've screwed up the patch implementation), so, i've removed it from my version.