daala.git
4 years agoAdding MAKE_SPARSE flag to intra_pred. style
Nathan E. Egge [Sat, 28 Sep 2013 00:01:48 +0000 (20:01 -0400)]
Adding MAKE_SPARSE flag to intra_pred.

4 years agoUpdating intra_pred to print betas after each step.
Nathan E. Egge [Wed, 25 Sep 2013 13:56:06 +0000 (09:56 -0400)]
Updating intra_pred to print betas after each step.

4 years agoMinor cleanup: fix parameter names
Jean-Marc Valin [Tue, 24 Sep 2013 05:07:31 +0000 (01:07 -0400)]
Minor cleanup: fix parameter names

4 years agoRefactored intradata.c to include OD_SATD_WEIGHTS.
Nathan E. Egge [Mon, 23 Sep 2013 17:50:26 +0000 (13:50 -0400)]
Refactored intradata.c to include OD_SATD_WEIGHTS.

4 years agoadapt: fix warning due to missing include
Tristan Matthews [Fri, 20 Sep 2013 17:46:22 +0000 (10:46 -0700)]
adapt: fix warning due to missing include

Signed-off-by: Ralph Giles <giles@mozilla.com>
4 years agoPrint correct block sizes for OD_SATD_WEIGHTS.
Gregory Maxwell [Tue, 17 Sep 2013 17:09:05 +0000 (10:09 -0700)]
Print correct block sizes for OD_SATD_WEIGHTS.

4 years agoDouble precision theta PVQ for upcoming experiments
Jean-Marc Valin [Tue, 17 Sep 2013 05:06:15 +0000 (01:06 -0400)]
Double precision theta PVQ for upcoming experiments

Not hooked up to anything yet

4 years agoAdded SSIM to RD-curve generation scripts.
Nathan E. Egge [Mon, 16 Sep 2013 16:54:06 +0000 (12:54 -0400)]
Added SSIM to RD-curve generation scripts.

4 years agoFixed typo in daala encoder_example.
Rl [Mon, 16 Sep 2013 16:24:58 +0000 (12:24 -0400)]
Fixed typo in daala encoder_example.

Signed-off-by: Nathan E. Egge <negge@dgql.org>
4 years agoMoving stack variables to the heap.
Nathan E. Egge [Fri, 6 Sep 2013 03:01:22 +0000 (23:01 -0400)]
Moving stack variables to the heap.

4 years agoAdding block mask to intra tools.
Nathan E. Egge [Wed, 4 Sep 2013 17:05:16 +0000 (13:05 -0400)]
Adding block mask to intra tools.

4 years agoPrefer 'return' to 'exit' in the test main(). r=ron
Ralph Giles [Wed, 21 Aug 2013 22:04:45 +0000 (15:04 -0700)]
Prefer 'return' to 'exit' in the test main(). r=ron

4 years agoSkip logging_test when logging isn't enabled. r=ron
Ralph Giles [Wed, 21 Aug 2013 17:24:18 +0000 (10:24 -0700)]
Skip logging_test when logging isn't enabled. r=ron

Makes a naive './configure && make check' pass out of the box.

4 years agoDouble-precision PVQ quantizer for upcoming experiments
Jean-Marc Valin [Sun, 18 Aug 2013 04:02:28 +0000 (00:02 -0400)]
Double-precision PVQ quantizer for upcoming experiments

4 years agoReplaces zig-zag with bands to makes it easier to mix scalar and pvq
Jean-Marc Valin [Fri, 16 Aug 2013 19:53:56 +0000 (15:53 -0400)]
Replaces zig-zag with bands to makes it easier to mix scalar and pvq

This does not change the SNR and has very small effect on the rate.

4 years agoAdds required declarations for using bands
Jean-Marc Valin [Sat, 10 Aug 2013 02:08:14 +0000 (22:08 -0400)]
Adds required declarations for using bands

4 years agoFixes the zigzag for sizes 8 and 16
Jean-Marc Valin [Fri, 9 Aug 2013 21:28:01 +0000 (17:28 -0400)]
Fixes the zigzag for sizes 8 and 16

Previous data was using the "reverse ordering", assuming that we were using
the lookup for the load when converting from raster to zigzag

4 years agodoc: Theoretical distortion for scalar quantization
Jean-Marc Valin [Wed, 7 Aug 2013 18:38:08 +0000 (14:38 -0400)]
doc: Theoretical distortion for scalar quantization

4 years agoAdding DROP_BY_MAG support to intra_pred.
Nathan E. Egge [Tue, 30 Jul 2013 14:56:39 +0000 (10:56 -0400)]
Adding DROP_BY_MAG support to intra_pred.

4 years agoAdding scripts to generate RD-curve graphs
Nathan E. Egge [Tue, 30 Jul 2013 14:26:41 +0000 (10:26 -0400)]
Adding scripts to generate RD-curve graphs

4 years agoClamp image value output for VP8 residue image
Monty Montgomery [Sat, 27 Jul 2013 09:19:05 +0000 (05:19 -0400)]
Clamp image value output for VP8 residue image

Clamp vp8 residue image output values to 0..255

4 years agoCorrect bugs in VP8 error residue image output
Monty Montgomery [Sat, 27 Jul 2013 08:39:50 +0000 (04:39 -0400)]
Correct bugs in VP8 error residue image output

X/Y on each block of output was swapped
center output on gray to match Daala output

4 years agoMore dubious theoretical results
Jean-Marc Valin [Wed, 24 Jul 2013 22:21:18 +0000 (18:21 -0400)]
More dubious theoretical results

4 years agoReworking the 2D moving average filter.
Gregory Maxwell [Wed, 10 Jul 2013 21:38:32 +0000 (14:38 -0700)]
Reworking the 2D moving average filter.

This changes the filter API to make it possible to use the filter in different
 contexts for different numbers of variables. It also gives the interface more
 intuitive names and avoids using struct arrays with half their members unused.

The functions themselves were rewritten to support runtime configured numbers
 of contexts. Some minor improvements were also made such as hoisting struct
 loads out of inner loops.

4 years agoBringing back quant_pvq_theta(), though disabled for now
Jean-Marc Valin [Tue, 9 Jul 2013 20:03:52 +0000 (16:03 -0400)]
Bringing back quant_pvq_theta(), though disabled for now

Squashed commit of the following:

commit 150119fac12701b428bd702e39ad27c011163457
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Tue Jul 9 15:59:53 2013 -0400

    pvq theta: using gain companding/expanding functions

commit decf0f9de945f4346413ec5982f817268ca9af6e
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Fri Jun 28 15:37:19 2013 -0400

    More incremental work

commit c84879585dca0be36fd75b84e0b0ed729e269792
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Wed Jun 26 12:59:22 2013 -0400

    Renaming

commit f7b40fa1db622d3b6902b6934b95b0dd0b32cb41
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Wed Jun 26 12:55:23 2013 -0400

    more integer quant_pvq_theta() work

commit 8d7a32164bd94770000edc342ed262e666bbf61d
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Wed Jun 26 03:40:03 2013 -0400

    Integerizing quant_pvq_theta() slowly

commit 79b8941a836be8ebf51ad6694b03e79bec37fd4b
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Wed Jun 26 01:36:59 2013 -0400

    change companded scale

commit 8be6b47f2e1bb9fae273d1eead8d6d1bb684a3ee
Author: Jean-Marc Valin <jmvalin@jmvalin.ca>
Date:   Tue Jul 9 14:56:37 2013 -0400

    Revices old theta-based PVQ code

4 years agoPVQ converts the rest to Q6
Jean-Marc Valin [Tue, 9 Jul 2013 03:27:06 +0000 (23:27 -0400)]
PVQ converts the rest to Q6

4 years agoPVQ: converts companded gains to CSCALE (Q6)
Jean-Marc Valin [Tue, 9 Jul 2013 03:07:41 +0000 (23:07 -0400)]
PVQ: converts companded gains to CSCALE (Q6)

4 years agoPVQ q now using CSCALE (Q6)
Jean-Marc Valin [Tue, 9 Jul 2013 01:09:24 +0000 (21:09 -0400)]
PVQ q now using CSCALE (Q6)

4 years agoChanges q to Q3 values
Jean-Marc Valin [Tue, 9 Jul 2013 00:27:22 +0000 (20:27 -0400)]
Changes q to Q3 values

4 years agoChange od_sqrt() output scaling to Q15.
Jean-Marc Valin [Mon, 8 Jul 2013 22:25:19 +0000 (18:25 -0400)]
Change od_sqrt() output scaling to Q15.

Increases accuracy of gains.

4 years agoReenable chroma from luma.
Gregory Maxwell [Mon, 1 Jul 2013 22:35:38 +0000 (15:35 -0700)]
Reenable chroma from luma.

This still needs some tuning as it's bouncing off the rails in some cases.

4 years agoFix DC PVQ rounding to prevent brightness/color shifts.
Gregory Maxwell [Mon, 1 Jul 2013 16:36:26 +0000 (09:36 -0700)]
Fix DC PVQ rounding to prevent brightness/color shifts.

This hurts R/D at sufficiently low rates due to there not being
much RDO. Though in Theora we didn't RDO the DC because it always
seemed to be a visual loss in any case, IIRC.

4 years agoKeep the lower precision MVs when subpel refine ties.
Gregory Maxwell [Mon, 1 Jul 2013 16:15:31 +0000 (09:15 -0700)]
Keep the lower precision MVs when subpel refine ties.

Also changes a couple of the debug level outputs to info.

4 years agoMake the parallel test mode of automake 1.13 useable
Ron [Sat, 29 Jun 2013 07:00:20 +0000 (16:30 +0930)]
Make the parallel test mode of automake 1.13 useable

4 years agointra prob doc: adds references, more on adaptation
Jean-Marc Valin [Sat, 29 Jun 2013 06:15:08 +0000 (02:15 -0400)]
intra prob doc: adds references, more on adaptation

4 years agointra doc fixes
Timothy B. Terriberry [Sat, 29 Jun 2013 05:32:51 +0000 (01:32 -0400)]
intra doc fixes

4 years agoSeparate the tests for the two different doc types
Ron [Fri, 28 Jun 2013 22:43:29 +0000 (08:13 +0930)]
Separate the tests for the two different doc types

There's no reason to not build any of them if the tools for only one
are available, they are independent at this stage.

This probably also needs to test for pdflatex, and could report what
will be enabled in a bit more detail in the final summary, but don't
get too carried with this until we know it's not all going to change.

Use && instead of 'test -a', the latter is a known portability trap.

4 years agoAdd functioning .pc files for daalaenc/dec
Ron [Fri, 28 Jun 2013 22:05:09 +0000 (07:35 +0930)]
Add functioning .pc files for daalaenc/dec

Based on a patch from Sebastian Dröge.

4 years agoIntra pred doc update
Jean-Marc Valin [Fri, 28 Jun 2013 19:38:11 +0000 (15:38 -0400)]
Intra pred doc update

4 years agoFixes intra trace.
Gregory Maxwell [Tue, 25 Jun 2013 23:39:10 +0000 (16:39 -0700)]
Fixes intra trace.

4 years agoUpdating tools to print updated intra predictors.
Nathan E. Egge [Tue, 25 Jun 2013 00:52:19 +0000 (20:52 -0400)]
Updating tools to print updated intra predictors.

Training tool now prints out intra predictors in packed format that is
 readable by od_intra_predNxN_mult functions.
Tools now use intra predictors in src/intradata.c instead of maintaining
 a separate copy.

4 years agoJmspeex' Journal of Dubious Theoretical Results
Jean-Marc Valin [Mon, 24 Jun 2013 18:21:02 +0000 (14:21 -0400)]
Jmspeex' Journal of Dubious Theoretical Results

4 years agoFixes for ecebf29e, fixes rounding avoids unnecessary normalization.
Gregory Maxwell [Mon, 24 Jun 2013 06:14:01 +0000 (23:14 -0700)]
Fixes for ecebf29e, fixes rounding avoids unnecessary normalization.

4 years agoNormalize mode probablities: lossless ~1.4% smaller.
Gregory Maxwell [Mon, 24 Jun 2013 04:23:59 +0000 (21:23 -0700)]
Normalize mode probablities: lossless ~1.4% smaller.

The mode probabilities use an unscaled CDF, the result was that the effective
 lambda depended on the CDF scale. This normalizes them in the search.

This also switches the cost scaling factors to Q8 integers and also makes the
 search lambda Q8.

4 years agoAdds an externally accessible lossless mode.
Gregory Maxwell [Sun, 23 Jun 2013 02:25:09 +0000 (19:25 -0700)]
Adds an externally accessible lossless mode.

This adds a scalar quantizer and a lossless mode (scalar + quantizer of 1)
 accessible by setting the quality to zero.

This is mostly useful for software testing, the lossless mode isn't
 very good as it intentionally tries to avoid executing different things
 than the lossy mode.

4 years agoSwitch back to the 1D Cg selected filters.
Gregory Maxwell [Sat, 22 Jun 2013 02:33:14 +0000 (19:33 -0700)]
Switch back to the 1D Cg selected filters.

The filters selected by the 2d coding gain appear to have problems
 with blocking artifacts and generally appear to perform worse
 on actual images.

4 years agoFix encoder_example's error output so it doesn't say it accepts RIFF WAV input.
Brendan Long [Sat, 22 Jun 2013 01:14:01 +0000 (19:14 -0600)]
Fix encoder_example's error output so it doesn't say it accepts RIFF WAV input.

Signed-off-by: Gregory Maxwell <greg@xiph.org>
4 years agoReformatting code to new style: state.c
Timothy B. Terriberry [Fri, 21 Jun 2013 22:40:13 +0000 (15:40 -0700)]
Reformatting code to new style: state.c

Minor follow-up corrections.

4 years agoFix MC visualization broken in 31a2e4d3.
Timothy B. Terriberry [Fri, 21 Jun 2013 22:38:53 +0000 (15:38 -0700)]
Fix MC visualization broken in 31a2e4d3.

4 years agoUse better libcheck references in the unix build. r=ron
Ralph Giles [Fri, 21 Jun 2013 21:16:35 +0000 (14:16 -0700)]
Use better libcheck references in the unix build. r=ron

Use pkg-config instead of just -lcheck to pick up dependencies
which may be specified in some versions. Also add -pthread for
the parallel tests, which pkg-config doesn't currently provide.

4 years agoMove ${LIBS} to the end of unix build example link lines. r=ron
Ralph Giles [Fri, 21 Jun 2013 18:04:25 +0000 (11:04 -0700)]
Move ${LIBS} to the end of unix build example link lines. r=ron

pvq.c in daalabase now requires -lm at least under gold.
The examples already have -lm from ${LIBS} but it was
listed before the libdaala*.a, which is important for static
linking. Moving it to the end resolves the ordering problem.

Unbreaks the build on Ubuntu.

4 years agoNon-bit-exact-disassembly reformatting changes to PVQ quant code
Jean-Marc Valin [Fri, 21 Jun 2013 20:01:54 +0000 (16:01 -0400)]
Non-bit-exact-disassembly reformatting changes to PVQ quant code

4 years agoFirst pass at reformatting PVQ quant (bit-exact disassembly)
Jean-Marc Valin [Fri, 21 Jun 2013 19:53:02 +0000 (15:53 -0400)]
First pass at reformatting PVQ quant (bit-exact disassembly)

4 years agoBe consistent with using parentheses for make variables
Ron [Fri, 21 Jun 2013 19:16:13 +0000 (04:46 +0930)]
Be consistent with using parentheses for make variables

4 years agoSome doc on the quantized RDO curve we could use in place of coding gain
Jean-Marc Valin [Fri, 21 Jun 2013 19:09:42 +0000 (15:09 -0400)]
Some doc on the quantized RDO curve we could use in place of coding gain

4 years agoAdd -pthread to CHECK_CFLAGS if check.pc does not.
Ron [Fri, 21 Jun 2013 18:05:07 +0000 (03:35 +0930)]
Add -pthread to CHECK_CFLAGS if check.pc does not.

Otherwise the tests fail to link under the same circumstances described
in commit 1f8a557c20b468e9ff87d24ef8f53116465c2fe1.

4 years agoCheck for libm and explicitly link it everywhere it is needed
Sebastian Dröge [Fri, 21 Jun 2013 15:01:01 +0000 (00:31 +0930)]
Check for libm and explicitly link it everywhere it is needed

The default for newer versions of ld is to enable --no-add-needed
(renamed to --no-copy-dt-needed-entries in later versions) which
means we need to explicitly link all the libraries we need, we
don't get them automatically from other dependencies anymore.

But that's not exactly the problem here.  The gold linker is even
more picky about this, and fails in some cases where GNU ld still
doesn't.  This patch fixes the build with gold.

4 years agoEnable logging in dump_video and player_example.
Timothy B. Terriberry [Fri, 21 Jun 2013 00:09:40 +0000 (17:09 -0700)]
Enable logging in dump_video and player_example.

4 years agoMake the logging initialization API public.
Timothy B. Terriberry [Fri, 21 Jun 2013 00:02:15 +0000 (17:02 -0700)]
Make the logging initialization API public.

If we have to call it from outside the library, it needs to be.

For the moment I'm not exposing the optional logging function, since
 it currently means exposing the list of logging facilities and
 levels.

4 years agoReformatting code to new style: state.c
Timothy B. Terriberry [Thu, 20 Jun 2013 23:51:24 +0000 (16:51 -0700)]
Reformatting code to new style: state.c

4 years agoReformatting code to new style: mcenc.c
Timothy B. Terriberry [Thu, 20 Jun 2013 23:50:59 +0000 (16:50 -0700)]
Reformatting code to new style: mcenc.c

Minor follow-up corrections.

4 years agoRemove another intentional negative shift.
Timothy B. Terriberry [Thu, 20 Jun 2013 23:25:21 +0000 (16:25 -0700)]
Remove another intentional negative shift.

Follow-up to 170a71e7.

4 years agoBased on gmaxwell's review
Jean-Marc Valin [Fri, 21 Jun 2013 01:04:02 +0000 (21:04 -0400)]
Based on gmaxwell's review

4 years agoDisables "faster" PVQ search that was producing obviously wrong results
Jean-Marc Valin [Thu, 20 Jun 2013 19:47:30 +0000 (15:47 -0400)]
Disables "faster" PVQ search that was producing obviously wrong results

4 years agoPVQ quant now turns off the gain offset on P-frames.
Jean-Marc Valin [Thu, 20 Jun 2013 06:15:48 +0000 (02:15 -0400)]
PVQ quant now turns off the gain offset on P-frames.

Slightly reduces the rate.

4 years agointra mode modelling doc changes
Jean-Marc Valin [Thu, 13 Jun 2013 18:25:37 +0000 (14:25 -0400)]
intra mode modelling doc changes

4 years agoFixing bug in 2D Cg computation.
Nathan E. Egge [Wed, 19 Jun 2013 22:22:04 +0000 (18:22 -0400)]
Fixing bug in 2D Cg computation.

4 years agoAdding flag to disable transforms in analysis/synthesis functions.
Nathan E. Egge [Wed, 19 Jun 2013 22:18:32 +0000 (18:18 -0400)]
Adding flag to disable transforms in analysis/synthesis functions.

4 years agoAdd a placeholder readme.
Ralph Giles [Thu, 13 Jun 2013 00:08:20 +0000 (17:08 -0700)]
Add a placeholder readme.

4 years agoPreparatory plugging to pacify parser parentheses paternalism.
Gregory Maxwell [Tue, 18 Jun 2013 00:56:20 +0000 (17:56 -0700)]
Preparatory plugging to pacify parser parentheses paternalism.

This silences silence all of the -Wparentheses warnings gcc 4.7 produces,
 except for state.c, tools/, tests/ectest.c, and tests/test_filters.c.

The warning is still disabled in configure.ac and unix/Makefile for now.

Objdump disassembly output on the whole tree is unchanged.

4 years agoRemove dead DPCM code.
Gregory Maxwell [Mon, 17 Jun 2013 23:20:03 +0000 (16:20 -0700)]
Remove dead DPCM code.

4 years agoPre/post-filter code now operates a SB at a time.
Nathan E. Egge [Mon, 17 Jun 2013 20:13:43 +0000 (16:13 -0400)]
Pre/post-filter code now operates a SB at a time.

Added od_apply_prefilter and od_apply_postfilter which filter an entire
 superblock in row order given an edge mask.
Added test_filters.c to test that pre and post filters correctly handle
 multiple block sizes.

4 years agoFix for prior coverage test change.
Gregory Maxwell [Wed, 12 Jun 2013 22:34:24 +0000 (15:34 -0700)]
Fix for prior coverage test change.

4 years agoCoverage test script improvements.
Gregory Maxwell [Wed, 12 Jun 2013 22:25:07 +0000 (15:25 -0700)]
Coverage test script improvements.

4 years agomerges the intra predictor x, y and neighbour id into the same table
Jean-Marc Valin [Fri, 7 Jun 2013 04:31:51 +0000 (00:31 -0400)]
merges the intra predictor x, y and neighbour id into the same table

4 years agoMissing od_ec_window cast in entdec.c.
Gregory Maxwell [Wed, 12 Jun 2013 17:31:14 +0000 (10:31 -0700)]
Missing od_ec_window cast in entdec.c.

Otherwise its int here and will shift too far and hit formally
 undefined behavior. Thanks to Tim for pointing out the obvious here.

4 years agoFixes the PVQ test to artificially overestimate OD_ADAPT_COUNT_Q8 by 256x
Jean-Marc Valin [Wed, 12 Jun 2013 07:41:54 +0000 (03:41 -0400)]
Fixes the PVQ test to artificially overestimate OD_ADAPT_COUNT_Q8 by 256x

4 years agoFixes more overflows on extreme values in the PVQ encoder.
Jean-Marc Valin [Wed, 12 Jun 2013 05:18:19 +0000 (01:18 -0400)]
Fixes more overflows on extreme values in the PVQ encoder.

4 years agoAdditional tests in test_coef_coder.c.
Gregory Maxwell [Wed, 12 Jun 2013 01:21:46 +0000 (18:21 -0700)]
Additional tests in test_coef_coder.c.

These tests trigger the bugs fixed in the prior commit as
 well as some signed integer overflows.

4 years agoFixes PVQ encoder corner cases
Jean-Marc Valin [Wed, 12 Jun 2013 01:15:33 +0000 (21:15 -0400)]
Fixes PVQ encoder corner cases

Mainly related to the max value in laplace_encode_special().

4 years agoRemaining formatting changes that affect the object code
Jean-Marc Valin [Tue, 11 Jun 2013 19:09:54 +0000 (15:09 -0400)]
Remaining formatting changes that affect the object code

4 years agoReformatting the pvq encoder code to the new coding style.
Jean-Marc Valin [Tue, 11 Jun 2013 19:38:23 +0000 (15:38 -0400)]
Reformatting the pvq encoder code to the new coding style.

Produces bit-exact objdump -d output

4 years agoUse the residual for block-size on non-keyframes.
Gregory Maxwell [Tue, 11 Jun 2013 18:00:47 +0000 (11:00 -0700)]
Use the residual for block-size on non-keyframes.

This results in using 16x16 all over the place and will need tuning.

Because the change moves around some code in order to perform the
 motion compensation prior to the block switching analysis diff gets
 a bit confused.

4 years agoConverts some of the PVQ quantizer to fixed-point.
Jean-Marc Valin [Tue, 11 Jun 2013 05:32:47 +0000 (01:32 -0400)]
Converts some of the PVQ quantizer to fixed-point.

The encoder no longer needs to call dequant_pvq() to avoid a desync.

4 years agoRemove intentional negative shifts.
Timothy B. Terriberry [Mon, 10 Jun 2013 19:31:57 +0000 (12:31 -0700)]
Remove intentional negative shifts.

We were doing them before the if test that ensured they were valid,
 but this is undefined behavior.

4 years agoCorrect unused parameter warnings
Monty Montgomery [Sun, 9 Jun 2013 06:30:51 +0000 (02:30 -0400)]
Correct unused parameter warnings

Correct several examples of legitimate unused parameter warnings where
functions were inadvertantly accessing a global rather than the passed
in copy of same structure.

4 years agoCorrect extraneous garbage in trans_gain after a pull merge
Monty Montgomery [Sun, 9 Jun 2013 06:07:20 +0000 (02:07 -0400)]
Correct extraneous garbage in trans_gain after a pull merge

Warning cleanyps in the middle of code being actively written mucked things up a little bit.

4 years agoMerge branch 'master' of git.xiph.org:/daala
Monty Montgomery [Sun, 9 Jun 2013 05:57:43 +0000 (01:57 -0400)]
Merge branch 'master' of git.xiph.org:/daala

Huh?  I didn't ask for a merge.

4 years agoCompleted trans_gain code with gain numbers
Monty Montgomery [Sun, 9 Jun 2013 05:54:12 +0000 (01:54 -0400)]
Completed trans_gain code with gain numbers

The trans_gain code is first-cut complete with nearly-complete gain
numbers for 1D and 2D across all data sets.

4 years agoFix some compiler warnings
Ron [Sun, 9 Jun 2013 01:11:09 +0000 (10:41 +0930)]
Fix some compiler warnings

Implicit declaration of abs() in block_size_enc.c.
Variable assigned but not used when logging is disabled in encode.c.
Unused parameter when asserts are disabled in ent{dec,enc}.c.

4 years agoEnable logging in `make debug` too
Ron [Sun, 9 Jun 2013 00:19:48 +0000 (09:49 +0930)]
Enable logging in `make debug` too

And split the defines out, partly for neatness and ease of future tweaking
but it also makes it possible to do `make debug DEBUG_OPTIONS=foo` to
override the extra options used on the fly.

4 years agoMore configure.ac housekeeping, no functional change
Ron [Sun, 9 Jun 2013 00:14:46 +0000 (09:44 +0930)]
More configure.ac housekeeping, no functional change

Note options that are completely unused here, that were brought in
from elsewhere, so they can be deleted later if they stay that way,
and unsplit related things that had drifted apart with other changes.

4 years agoAdd an --enable-logging configure option
Ron [Sun, 9 Jun 2013 00:12:51 +0000 (09:42 +0930)]
Add an --enable-logging configure option

4 years agoDon't force assertions and logging unconditionally
Ron [Sun, 9 Jun 2013 00:08:45 +0000 (09:38 +0930)]
Don't force assertions and logging unconditionally

This reverts commit 1767c40a6b150a17f66736a28520d8a1ed887bd5.
They can either be --enabled at configure time, or enabled on the fly
using `make debug` (once that's also fixed in a commit to follow).

4 years agoSwitch to using config.h for configuration options
Ron [Sun, 9 Jun 2013 00:00:00 +0000 (09:30 +0930)]
Switch to using config.h for configuration options

The main motivation for this is it ensures the files will actually be
rebuilt if the configuration options are changed.

Add a description to the OD_X86ASM define, since autoheader barfs and
dies when generating config.h.in without one.

4 years agoNote which tools created generated files
Ron [Sat, 8 Jun 2013 23:51:39 +0000 (09:21 +0930)]
Note which tools created generated files

Mostly because some of them will need to include config.h now, or may in
the future, so we want to make sure the right tool gets updated if needed.

4 years agoFix whitespace damage
Ron [Sat, 8 Jun 2013 23:46:33 +0000 (09:16 +0930)]
Fix whitespace damage

And a "nested" /* comment, that was actually unpaired.

4 years agoPreserve user supplied CFLAGS
Ron [Sat, 8 Jun 2013 18:39:21 +0000 (04:09 +0930)]
Preserve user supplied CFLAGS

By actually expanding them as a shell variable ...
Also put them at the end, so the user has the final say for overrides.

4 years agoDon't build the tools by default
Ron [Sat, 8 Jun 2013 17:56:59 +0000 (03:26 +0930)]
Don't build the tools by default

You can build them all with `make tools` or individually with
`make tools/foo`.  Using `make clean` will still remove them.

4 years agoFix build without assertions.
Gregory Maxwell [Sun, 9 Jun 2013 00:47:17 +0000 (17:47 -0700)]
Fix build without assertions.