--rd documenation,..
As I understood it adding '--rdoq-level' made rdo quantization independent of the general '--rd' level so the documentation (http://x265.readthedocs.org/en/latest/cli.html?highlight=rd#cmdoption--rd) probably needs some adjustment, since old -rd 4&5 seem to be current --rdoq-level 1&2
Comments (6)
-
-
yes, currently there is no logic difference within the encoder between --rd 3 and --rd 4, just as there are still no internal differences between --rd 5 and --rd 6. It likely won't stay that way for long.
I don't think there's any reason to restrict --rdoq-level != 0 to --rd 4 to 6. It doesn't add any encoder complexity or overhead to allow it to be used in all presets. For use cases that really need psy-rd, psy-rdoq and rdoq-level=1 are often worth the trouble.
-
reporter Then why not: change: --rd to 0 sa8d mode and split decisions, intra w/ source pixels 1 recon generated (better intra), RDO merge/skip selection 2 RDO splits and merge/skip selection 3 RDO mode and split decisions, chroma residual used for sa8d 4 Adds RDO prediction decisions and adjust the presets accordingly, seeing that --rd 5&6 is the same internally for quite some time, is it really likely that additional rd-related options will be another rd level and not an additional option like '--rdoq-level' ?
-
This is what I had in mind, too. Sent a patch - not sure how disruptive changing existing settings is for our users.
-
There is some work already under-way to re-order analysis so splits are tried before ME or Intra at each depth (allowing many new early-outs); I already have ideas for how to differentiate --rd 3 & 4 and --rd 5 & 6 given the new logic flow
-
- changed status to resolved
closing, since the reported doc issue was fixed
- Log in to comment
Thanks, corrected the doc. Ideally, we would like to shrink valid rd-levels from 6 to 4. Not sure how disruptive that would be.