Pauli Oikkonen
24d487f553
New weights for 12 <= QP <= 42
...
Trained using MSU ultrafast settings now
2020-10-29 15:20:27 +02:00
Pauli Oikkonen
3e1c6d84b8
Fix issues in fast coeff estimation
...
Allow weight table to start from nonzero QP, and round weights to Q8.8
instead of flooring them
2020-10-29 15:20:27 +02:00
Pauli Oikkonen
5f91bda762
Use newer data for fast coeff cost estimation
...
Same training dataset, but this time only buckets 0...3 were used to
approximate the function, no sign/cg width bucket.
2020-10-29 15:20:27 +02:00
Pauli Oikkonen
2abd733199
Use unsigned min() to correctly clip -32768
...
If a coeff happens to be -32768 (0x8000), its 16-bit abs() is also
0x8000. It should ultimately be clipped to 3, so interpret absolute
values as unsigned instead to make that happen.
2020-10-29 15:20:27 +02:00
Pauli Oikkonen
b93b90c0d7
Implement new fast coeff cost estimator in AVX2
2020-10-29 15:20:27 +02:00
Pauli Oikkonen
2f74a112b3
Try first lookup table based fast coeff estimation
2020-10-29 15:20:27 +02:00
siivonek
bc1206a4d3
Define qp_delta_min & max in global.h instead of calculating them locally.
2020-09-29 13:46:27 +02:00
siivonek
eeb2840505
Add examples folder and readme.md for examples. Add example roi file.
2020-09-28 17:15:12 +02:00
siivonek
d5fe6c3737
Remove xfail for VAQ from 10bit-test since it should be working now.
2020-09-22 20:15:23 +02:00
siivonek
0f3ef786b9
Modify delta QP range assert so it will work with any valid bit depth. Modify VAQ code so it will clip the QP to a proper range which is dependent on bit depth
2020-09-22 20:15:23 +02:00
siivonek
fe6f93a951
Fix delta QP range check assert. Add separate asserts based on bit depth.
2020-09-22 20:15:22 +02:00
Joose Sainio
8143ab971c
Merge branch 'stats-files'
...
# Conflicts:
# src/cfg.c
# src/cli.c
# src/kvazaar.h
2020-09-16 09:25:00 +03:00
Joose Sainio
1c06bd7f3d
Fix POC to be correct for all GOPs and Intra periods, fix issue with vaq
2020-09-14 14:25:48 +03:00
Sami Ahovainio
4d87fb2397
fixed potential out of bounds iteration
2020-09-10 12:59:39 +03:00
Sami Ahovainio
7e2bbcfb64
Merge remote-tracking branch 'remotes/origin/y4m' to master
2020-09-09 16:39:12 +03:00
Sami Ahovainio
245d0cd2cf
updated readme
2020-09-09 16:14:04 +03:00
Sami Ahovainio
5d521a2444
Added option to force yuv as file format and made the options and file endings case insensitive
2020-09-09 16:05:59 +03:00
Joose Sainio
984e5f72f1
update documents
2020-09-09 13:49:13 +03:00
Joose Sainio
3fb8b7ebc6
Add --stats-file-prefix option
...
When the option is defined with an option four files prefixlambda.txt,
prefixqp.txt, prefixdist.txt, and prefixbits.txt that have the corresponding
data for each ctu. This is a debug feature.
2020-09-09 12:35:47 +03:00
Sami Ahovainio
5b0d510647
updated readme
2020-09-07 17:07:58 +03:00
Sami Ahovainio
84cabd9c20
Fixed sign match
2020-09-07 15:39:31 +03:00
Sami Ahovainio
d691849594
Added frame header reading for both read and seek functions
2020-09-07 15:31:08 +03:00
Sami Ahovainio
cbcee67821
y4m start header parsing ready
2020-09-07 15:31:07 +03:00
Joose Sainio
c10b841e7c
Merge remote-tracking branch 'remotes/origin/fix-sao-parameter' into master
2020-09-07 13:10:36 +03:00
Joose Sainio
da09d49890
Remove optionality from --sao
...
SAO parameter was optional which caused that if one wants to pass argument
one needs to use "=" which is confusing since this is not required for any
other parameter
2020-09-07 12:35:40 +03:00
Pauli Oikkonen
2bda7c1e5f
Xfail lacking VAQ on 10-bit build
...
VAQ is currently not supported with 10-bit pixels, so xfail the test for
now.
2020-09-03 17:28:26 +03:00
Pauli Oikkonen
c062f12340
Add a test for the 10-bit build
2020-09-02 18:09:35 +03:00
Pauli Oikkonen
3f7f0d7ed7
Allow bit depth to be defined from the outside
...
For a 10-bit build, just use:
env CFLAGS="-DKVZ_BIT_DEPTH=10" ./configure && make clean && make
2020-09-02 17:55:22 +03:00
Pauli Oikkonen
780da4568a
Exclude 8-bit-only code from 10-bit builds and use uint8_t instead of kvz_pixel for code that assumes 8-bit pixels
2020-09-02 17:46:33 +03:00
Pauli Oikkonen
31ef4e4216
Fix ml functions to accept kvz_pixel*, not uint8_t*
2020-09-02 17:46:33 +03:00
Joose Sainio
faf5cc858d
Merge branch 'fix-lp-gop-rc'
2020-06-25 09:41:57 +03:00
Joose Sainio
138651ee85
Fix the bit and frame counts for calculating the gop allocation
...
Additionally dynamically adjust the smoothing window if there are rapid changes
2020-06-24 15:26:54 +03:00
Ari Lemmetti
f8ff6dd567
Merge pull request #262 from jbeich/truncate-freebsd
...
Unbreak build on FreeBSD
2020-06-22 18:08:01 +03:00
Ari Lemmetti
d1abf85229
Add MV constraint check to motion estimation start point
2020-06-01 23:51:38 +03:00
Jan Beich
1fa69c705d
Rename truncate() from 30ce461d98 to avoid conflict with POSIX version
...
strategies/avx2/dct-avx2.c:55:23: error: static declaration of 'truncate' follows non-static declaration
static INLINE __m256i truncate(__m256i v, __m256i debias, int32_t shift)
^
/usr/include/stdio.h:448:6: note: previous declaration is here
int truncate(const char *, __off_t);
^
2020-04-22 16:09:42 +00:00
Ari Lemmetti
9753820b3a
Update version to 2.0.0
2020-04-22 01:03:36 +03:00
Ari Lemmetti
40e81f3243
Update preset tables. Update docs.
2020-04-22 01:03:21 +03:00
siivonek
54f438a75c
Update VAQ help text. Update docs. Change some lingering tabs to spaces.
2020-04-20 16:52:07 +02:00
Pauli Oikkonen
35facb20e8
Do not build AVX2 with GCC on MinGW
...
There is apparently no way to build non-inline functions accepting or
returning AVX2 params with GCC on MinGW without mangling the source code
for all platforms, because it simply refuses to align stack to more than
16 bytes where needed. Drop the AVX2 optimizations on that platform to
have a working build.
AVX2 for MinGW builds is still supported with Clang, please use it
instead because it works.
2020-04-14 23:18:57 +03:00
Ari Lemmetti
f31dddc019
Bypass inverse quantization and inverse transform when trying early skip
2020-04-10 16:02:09 +03:00
Pauli Oikkonen
dfab3712c3
Tell people to use Clang instead of GCC on MinGW
2020-04-08 20:10:33 +03:00
Pauli Oikkonen
73abcf266b
Make distcheck a totally separate CI test
2020-04-08 19:27:12 +03:00
Pauli Oikkonen
fbdb1e2d15
Add correct path to sao_shared_generics.h in makefile
2020-04-08 19:27:12 +03:00
Pauli Oikkonen
652802e54e
Add distcheck to Gitlab CI
2020-04-08 19:27:12 +03:00
Pauli Oikkonen
8617530b13
Use _mm_store_epi64 instead of _mm_cvtsi128_si64
...
Fix 32-bit builds that tend to lack the cvt intrinsic. Hope it will be
optimized to a movq r64, xmm on modern platforms though
2020-04-07 23:51:54 +03:00
Pauli Oikkonen
a82966c0f5
Fix lacking _mm256_cvtss_f32 intrinsic on VS
...
Cast __m256 into __m128 first, the XMM variant of the intrinsic has been
around for a long enough time to be supported
2020-04-07 22:38:10 +03:00
Ari Lemmetti
4f96278162
Override angular_pred to generic in travis as well. Assuming false positive from valgrind.
2020-04-07 19:22:22 +03:00
Joose Sainio
c369ff8873
Fix a potential division by zero in a floating point operation
...
When C is calculated with K if the value of K is not clipped before in some
cases it is possible that K gets such a large negative value that bpp^K is
rounded to zero. In real-life cases this is extremely rare and clipping
beforhand has very little to no effect.
Also remove commented debug prints
2020-04-06 11:05:49 +03:00
Ari Lemmetti
901c25c0c8
Merge branch 'vaq'
2020-04-03 19:51:17 +03:00
Ari Lemmetti
51451be5ef
Handle cases where the number of pixels is not divisible by 32
2020-04-03 19:37:47 +03:00
