Ari Lemmetti
14eae9ee86
Merge branch 'angular_pred_avx2'
2015-10-08 12:44:13 +03:00
Ari Lemmetti
5af7a42ebe
Enable AVX2 strategy. Add first version of optimizations.
2015-10-08 12:36:20 +03:00
Ari Lemmetti
f4fe3dca5e
Add AVX2 strategy. Copy generic implementation there.
2015-10-08 12:36:15 +03:00
Ari Lemmetti
54e8b346a3
Add intra strategy. Move angular prediction there.
2015-10-08 12:36:05 +03:00
Ari Lemmetti
6214e2b328
Fix small typos
2015-10-05 19:41:23 +03:00
Ari Lemmetti
c123b97fec
Remove option -fno-lto from strategies. LTO is no longer used anyway.
2015-10-05 19:34:56 +03:00
Ari Koivula
f055e4db74
Merge branch 'intra-pred-rewrite'
2015-10-05 06:26:53 +03:00
Ari Koivula
ff976e2afc
Arrange parameters in intra fancily
2015-10-05 06:23:14 +03:00
Ari Koivula
d83d57df1a
Fix function names in intra
...
Prefix non-static functions with kvz_intra_ and static with intra_.
2015-10-05 06:23:14 +03:00
Ari Koivula
30b4fa4247
Rename intra prediction to kvz_intra_predict
2015-10-05 06:23:14 +03:00
Ari Koivula
7280dbf429
Remove unnecessary function
...
This function used to be more complicated, but now it's so simple that
it's just obfuscating what's happening.
2015-10-05 06:23:05 +03:00
Ari Koivula
1221e4c7d2
Remove old intra prediction code.
2015-10-05 05:30:47 +03:00
Ari Koivula
23439557e6
Remove remaining usages of old intra prediction
2015-10-05 05:23:22 +03:00
Ari Koivula
ca3ba997aa
Switch to new intra pred in search_intra_chroma_rough.
2015-10-05 05:03:58 +03:00
Ari Koivula
eaff6e29d9
Switch to new intra pred in kvz_search_cu_intra
2015-10-05 04:00:42 +03:00
Ari Koivula
55d741e250
Switch to new intra pred in kvz_intra_recon_lcu_chroma
2015-10-05 04:00:42 +03:00
Ari Koivula
678a1dd1dd
Switch to new intra pred in kvz_intra_recon_lcu_luma
2015-10-05 02:29:02 +03:00
Ari Koivula
cd2f1797bf
Add reimplemented intra prediction code
...
Just along side for now to help with debugging.
The main difference with the new versions is that they take and output
width**2 blocks and two width*2+1 arrays of reference samples,
instead of the (2*width+8)**2 blocks the old ones do. This should make
the interface clearer and the memory footprint smaller.
Also commented the shit out of angular prediction, so hopefully Ari L.
will have an easier time with a SIMD implementation.
2015-10-05 02:29:02 +03:00
Ari Koivula
115756b9d7
Accept --rd=3 parameter
2015-10-05 02:28:56 +03:00
Ari Lemmetti
7a3dabf43e
Merge branch 'avx2-quant'
2015-10-02 16:31:33 +03:00
Ari Lemmetti
38106afa50
Add AVX2 version of quantization.
2015-10-02 16:18:52 +03:00
Ari Lemmetti
ef0ad292ef
Add quantization strategy.
2015-10-02 16:17:02 +03:00
Ari Koivula
41dd44f7cf
Fix warnings with -DNDEBUG
2015-10-02 15:13:07 +03:00
Ari Koivula
81f5ca76cb
Accept tile configurations with either dimension as one
2015-10-02 15:06:31 +03:00
Ari Lemmetti
a21c5b817f
Merge branch 'hadamard_avx2'
2015-10-02 14:31:17 +03:00
Ari Lemmetti
989cee1b04
Add 4x4 function as well
2015-10-01 22:14:56 +03:00
Ari Lemmetti
8b57b2bb1a
Refactor SATD to inline most of the function. Replace full horizontal add with shuffle and regular packed add.
2015-10-01 21:29:25 +03:00
Ari Lemmetti
55da2a9958
Add intrinsic version of SATD for 8x8 and larger blocks
2015-10-01 19:42:22 +03:00
Ari Lemmetti
d68fc4c41e
Add header for common utilities to use with strategies.
2015-10-01 19:40:35 +03:00
Arttu Ylä-Outinen
512e5bb25f
Bump version to 0.7.0
2015-09-30 15:20:57 +03:00
Arttu Ylä-Outinen
fdc31106a0
Merge branch 'frame-info'
2015-09-28 12:00:53 +03:00
Arttu Ylä-Outinen
8f404a3b6f
Add NAL unit type to frame_info.
2015-09-28 10:30:59 +03:00
Arttu Ylä-Outinen
1c898a2f4a
Prefix NAL unit type enum constants with KVZ_.
2015-09-28 10:30:58 +03:00
Arttu Ylä-Outinen
4e5c7fe6e8
Remove function kvz_encoder_compute_stats.
...
Changes main function to compute frame PSNR by calling
kvz_videoframe_compute_psnr directly with the source and reconstructed
pictures returned from encoder_encode.
2015-09-28 10:30:58 +03:00
Arttu Ylä-Outinen
efd361ee8e
Return the original picture from encoder_encode.
2015-09-28 10:30:58 +03:00
Arttu Ylä-Outinen
afd0d3eee0
Remove encoderstate dependency from cli module.
...
Changes function print_frame_info to use a kvz_frame_info struct to get
the data to be printed.
2015-09-28 10:30:58 +03:00
Arttu Ylä-Outinen
7edc1b0b1c
Add reference picture lists to kvz_frame_info.
2015-09-28 10:30:57 +03:00
Arttu Ylä-Outinen
d5dceb45f1
Factor out a function for building ref lists.
...
The code for building the reference picture lists was duplicated in
functions encoder_state_ref_sort and print_frame_info. This commit moves
it to a new function kvz_encoder_get_ref_lists. Also makes
encoder_ref_insertion_sort static since it is not used outside the
encoderstate module any more.
2015-09-28 10:30:57 +03:00
Arttu Ylä-Outinen
c856a6b598
Output frame info from encoder_encode.
...
Adds a new output parameter info_out to encoder_encode. It returns
a struct containing information about the encoded frame, including POC,
QP and slice type.
2015-09-28 10:30:57 +03:00
Arttu Ylä-Outinen
173b70b53f
Rename SLICE_* enum constants to KVZ_SLICE_*.
2015-09-28 10:30:56 +03:00
Ari Lemmetti
9051fbf593
Fix function names in speed tests
2015-09-18 17:15:59 +03:00
Ari Koivula
e9f1fdf6f2
Update the unit test lib to newest version
2015-09-18 15:38:55 +03:00
Ari Koivula
09baa6bd5e
Fix intra_sad and satd_tests
...
The tests were not being run, because the names have changed.
2015-09-18 15:26:37 +03:00
Ari Koivula
63ab4068be
Clean up the makefile a bit
...
Use the existing TARGET_CPU_ARCH and TARGET_CPU_BITS instead of filtering
ARCH over and over again.
Comment some of the more obscure parts.
2015-09-18 15:13:13 +03:00
Ari Koivula
eb12fe0d98
Re-enable disabled -m32 and -m64 flags.
2015-09-18 12:13:31 +03:00
Ari Koivula
9537b996e7
Make makefile work on arm
...
Only compile x86 specific optimizations for x86 and don't give
-m32 or -m64 on arm.
2015-09-18 00:23:49 +03:00
Ari Koivula
d76890bbff
Bump version to 0.6.1
2015-09-16 18:42:20 +03:00
Ari Koivula
1d5cfbdcc2
Remove unused variable.
2015-09-16 18:39:46 +03:00
Ari Koivula
513e80bcca
Fix bug causing unnecessary copying of memory
...
This bug caused a single tiles worth of lcu_info_t structs to be copied
unnecessarily for every LCU in the frame. This obviously caused huge
memory bandwidth issues when coding large frames without tiles. The
effect was minimized somewhat with a large number of tiles, because
only the current tile was copied.
From context it is clear that this piece of code was supposed to copy
a single tile or frame, once the frame was done, but because it was
placed in a function which is called for every LCU, it copied the data
for the LCU, but also lots of extra stuff.
The fix is to copy only the current LCU instead of the whole tile.
2015-09-16 18:23:44 +03:00
Marko Viitanen
d8b50d6951
Bump version to 0.6.0
2015-09-15 15:44:55 +03:00