Log of /branches/dev-api-4/xvidcore/src/utils/mbtransquant.c
Parent Directory
Revision
1230 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Sun Nov 30 16:13:16 2003 UTC (20 years, 6 months ago) by
edgomez
File length: 41630 byte(s)
Diff to
previous 1224
,
to
selected 384
* Thread safe MPEG4 quantization functions.
Cleaned up version of patch provided by Michael
- fixed compiling problems on gcc
- added const qualifiers every where it was possible to help C compiler
optimization.
- added the mpeg_quant_matrices param to all ASM function prototype
in comments (even if it's not used, that shows we do it deliberatly)
- forces m[intra][0][0] = 8, otherwise XviD could write invalid streams.
* Added real CRC computing in xvid_bench.c
Revision
1224 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Mon Nov 24 22:06:19 2003 UTC (20 years, 6 months ago) by
edgomez
File length: 41015 byte(s)
Diff to
previous 1223
,
to
selected 384
Trellis was treating big levels exactly the same way as lower ones.
In some cases, trellis was doing wild optimizations favoring a 0
because the distortion introduced by that big coeff change was
acceptable. But visually this could result in some nasty blocks with wrong
chroma information or similar brutal changes in other planes as well.
Skal added big levels handling where trellis just tries to minimize
the cost varying the run value only. No level modification is done
anymore.
Revision
1223 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Sun Nov 23 17:01:08 2003 UTC (20 years, 6 months ago) by
edgomez
File length: 40591 byte(s)
Diff to
previous 1174
,
to
selected 384
* Added trellis support for MPEg quantization type.
* Changed RD fixed point precision, should help avoiding overflow
(see the constant TL_SHIFT)
NB: we still have some problems when trellis optimizes DC for big DC
values.
Revision
982 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Thu Apr 10 13:05:54 2003 UTC (21 years, 2 months ago) by
edgomez
File length: 15289 byte(s)
Diff to
previous 965
,
to
selected 384
2003-04-10 GMT Edouard Gomez <ed.gomez@free.fr> patch-14
Summary:
Removed all ABS() macros.
Revision:
xvidcore--devapi4--1.0--patch-14
All ABS macros have been replace with their stdlib.h/math.h
equivalent. This gives a 33% overall speedup for the plain C encoder,
while the ia32 one seems to suffer a small speed loss. However this
speed loss is very small and it seems it depends on the CPU type as
the abs/fabs usage is impacting badly on sad functions but impacting
well on interpolate functions ... weird inst'it ?
modified files:
src/bitstream/mbcoding.c src/global.h src/image/image.c
src/motion/motion_comp.c src/motion/motion_est.c
src/motion/sad.c src/prediction/mbprediction.c
src/utils/mbtransquant.c src/xvid.c
Revision
920 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Sat Mar 15 14:32:56 2003 UTC (21 years, 3 months ago) by
suxen_drol
File length: 22304 byte(s)
Diff to
previous 914
,
to
selected 384
enc_encode CLOSE_GOP fixed, further cleanup
XVID_REQORIGINAL plugin support
plugin examples: xvid_plugin_psnr, xvid_plugin_dump
Revision
914 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Mon Mar 10 00:38:49 2003 UTC (21 years, 3 months ago) by
edgomez
File length: 22297 byte(s)
Diff to
previous 890
,
to
selected 384
2003-03-10 00:36:15 GMT Edouard Gomez <ed.gomez@free.fr> patch-4
Summary:
Adds extended stats support even for bframes in xvidcore.
Revision:
xvidcore--devapi4--1.0--patch-4
This patch enables core extended stats support even for bframes. It
modifies the way the MBTransQuantBVOP function does its work. It used
to not dequant, idct the MB because bframes are never used as reference
frames. However if we want to compute stats, then we must perform these
inverse transformations.
modified files:
src/encoder.c src/utils/mbfunctions.h src/utils/mbtransquant.c
Revision
78 -
(
view)
(
download)
(
annotate)
-
[select for diffs]
Modified
Thu Mar 28 20:57:25 2002 UTC (22 years, 2 months ago) by
edgomez
Original Path:
trunk/xvidcore/src/utils/mbtransquant.c
File length: 13014 byte(s)
Diff to
previous 69
,
to
selected 384
Big patch, use it with care (tests done with GNU/Linux) :
- Trivial warnings in encoder.c line 282 ret1 = ret 2 = 0;
- Lot of "New line at end of file missing" warnings fixed (gcc3)
- All 2d arrays have been turned into 1d arrays
To force alignment of matrices on stack, use DECLARE_ALIGNED_MATRIX
(read portab.h)
Win32 users should give feedback.
This form allows you to request diffs between any two revisions of this file.
For each of the two "sides" of the diff,
enter a numeric revision.