Paul Wilkins [Thu, 3 Oct 2013 13:20:20 +0000 (14:20 +0100)]
Further clean up of speed 4
Speed 4 still does not give a big gain over speed 3.
This just cleans it up a little from the last patch and comments
out features that do not seem to be giving much benefit.
Dmitry Kovalev [Fri, 4 Oct 2013 01:55:21 +0000 (18:55 -0700)]
Adding vp9_get_filter_kernel() function.
Moving INTERPOLATIONFILTERTYPE enum and subpix_fn_table struct to
vp9_filter.h. Adding convenient typedef for subpel kernels.
Function vp9_setup_interp_filters() besides setting xd->subpix.filter_x &
xd->subpix.filter_y has a side effect of also setting scale factors. This
is not required inside decode_modes_b() because scale factors have been
already set by set_ref() calls. That's why replacing
vp9_setup_interp_filters() call with newly created vp9_get_filter_kernel()
call. The behavior of vp9_setup_interp_filters() is unchanged (it
is used from the encoder).
Jingning Han [Fri, 27 Sep 2013 23:02:49 +0000 (16:02 -0700)]
Refactor inter mode rate-distortion search
This commit separates the rate-distortion optimization loop of
superblocks from that of sub8x8 blocks. This allows better design
rate-distortion optimization search loop for each setting. It also
removes the use of SPLITMV and I4X4_PRED therein.
No performance change in speed 0 settings. For bus@CIF at 2000kbps,
the speed 1 runtime goes from 48009ms to 43894ms (about 10% faster).
The overall compression performance on derf changed by -0.021%.
Speed 2 runtime goes from 27114ms to 28700ms (6% slower), while the
overall coding efficiency goes up by 1.629% for derf, 1.236% for yt.
Yunqing Wang [Thu, 3 Oct 2013 00:26:01 +0000 (17:26 -0700)]
Rewrite HORIZx4 and HORIZx8 in subpixel filter functions
In subpixel filters, prefetched source data, unrolled loops,
and interleaved instructions.
In HORIZx4, integrated the idea in Scott's CL (commit: d22a504d11a15dc3eab666859db0046b5a7d75c5), which was suggested by
Erik/Tamar from Intel. Further tweaking was done to combine row 0,
2, and row 1, 3 in registers to do more 2-row-in-1 operations until
the last add.
Paul Wilkins [Tue, 1 Oct 2013 15:57:18 +0000 (16:57 +0100)]
Speed setting review.
Substantial reworking of the speed vs quality trade offs for
speed 1 and 2.
In this patch I am attempting to freeze the "quality" meaning of
speeds 1 and 2 relative to speed 0 so that in future we can
better evaluate progress.
I am targeting :
Speed 1 quality ~-5% vs speed 0.
Speed 2 quality ~-10% vs speed 0
It is inevitable that quality will still fluctuate a little as we adjust
settings and add new features, but we will attempt to keep as
close as possible to these values. Above speed 2 things will remain
a bit more fluid for now.
In this patch speed 1 is approximately 4-5x as fast as speed 0. This
is similar to before but the quality hit is a lot less. Likewise speed 2
is approximately 2x as fast as speed 1 but is similar in quality to the
previous speed 1 configuration.
Also slight change to behavior of FLAG_EARLY_TERMINATE to insure
all reference frames get at least one rd test. Important for very low
variance regions.
WIP :- Added a new speed level with old speed 4 becoming speed 5.
Speed 3 and 4 tradeoffs still WIP
Dmitry Kovalev [Wed, 2 Oct 2013 21:13:33 +0000 (14:13 -0700)]
Moving all idct/iht functions in one place.
Moving functions from vp9_idct_blk to vp9_idct because these functions are
used from both encoder and decoder. Removing duplicated code from
vp9_encodemb.c and reusing existing functions.
Jingning Han [Sat, 28 Sep 2013 00:22:59 +0000 (17:22 -0700)]
BITSTREAM - CLARIFICATION OF MV SIZE RANGE
The codec should effectively run with motion vector of range (-2048, 2047)
in full pixels, for sequences of 1080p and below. Add assertions to clarify
this behavior.
Dmitry Kovalev [Tue, 1 Oct 2013 22:41:30 +0000 (15:41 -0700)]
Making decode_modes_b function more straightforward.
Moving out decode_tokens function calls and adding decode_blocks boolean
variable. We only have to decode if eobtotal > 0, i.e. we have at least one
non-zero coefficient. Also inlining and remove vp9_set_pred_flag_mbskip
function.
Matthew Heaney [Tue, 24 Sep 2013 22:13:29 +0000 (15:13 -0700)]
Fix linker warnings for bilinear filters
The declaration of the bilinear filters specified an alignment clause
in the implementation file but not in the header. This turned out
to be harmless, but it did cause linker warnings to be emitted when
building on Windows.
The (extern) declaration in the header was changed, to match the
declaration in the implementation.
Jingning Han [Mon, 30 Sep 2013 18:58:22 +0000 (11:58 -0700)]
Fix rectangular partition check in speed 1
Make encoder skip rectangular partition check in speed 1 and above,
when early termination was triggered in partition split.
Thanks Guillaume (gmartres@) for catching this issue.
This change makes bus_cif at 2000kbps speed 1 runtime goes down from
25612ms to 23438ms (about 9% speed-up), at the expense of -0.235%
performance down.
We don't need these functions anymore. The only one which was actually
used is vp9_add_constant_residual_32x32. Addition of
vp9_short_idct32x32_1_add eliminates this single usage. SSE2 optimized
version of vp9_short_idct32x32_1_add will be added in the next patch set,
right now it is only C implementation. Now we have all idct functions
implemented in a consistent manner.
projects / libvpx / log