wrapper/FFmpeg
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
83571 commits 40 branches 405 tags 230 MiB
Commit graph

2204 commits

Author SHA1 Message Date
James Darnley
5336887867 avcodec/h264: sse2, avx h luma mbaff deblock/loop filter 
x86-64 only

Yorkfield:
- sse2: ~2.17x (434 vs. 200 cycles)

Nehalem:
- sse2: ~2.94x (409 vs. 139 cycles)

Skylake:
- sse2: ~3.10x (370 vs. 119 cycles)
- avx:  ~3.29x (370 vs. 112 cycles)
 2017-02-18 20:26:52 +01:00
James Darnley
e18bc2114f avcodec/h264: add named parameters to x86 function 2017-02-18 20:26:50 +01:00
James Darnley
9d815b7424 avcodec/x86: deduplicate PASS8ROWS macro 2017-02-18 20:26:49 +01:00
James Almer
c8467abbad x86/rv34dsp: add ff_rv34_idct_dc_add_sse2 
Also disable ff_rv34_idct_dc_add_mmx on x86_64 as the presence of sse2
is guaranteed in such builds.

Signed-off-by: James Almer <jamrial@gmail.com>
 2017-02-02 17:51:21 -03:00
James Almer
ab5c4d006d x86/vp8dsp: add ff_vp8_idct_dc_add_sse2 
Also disable ff_vp8_idct_dc_add_mmx on x86_64 as the presence of sse2
is guaranteed in such builds.

Signed-off-by: James Almer <jamrial@gmail.com>
 2017-02-02 17:18:58 -03:00
Michael Niedermayer
536ac72f46 Revert "Merge commit '0a39c9ac0b'" 
The assumption this is based on is wrong, the code is not always run with bitexact flags

This reverts commit a956164e1e, reversing
changes made to f6005907fd.

Approved-by: James Almer <jamrial@gmail.com>
 2017-02-01 02:01:07 +01:00
James Almer
ba5d089381 Merge commit 'd06dfaa5cb' 
* commit 'd06dfaa5cb':
  x86: huffyuv: Use EXTERNAL_SSSE3_FAST convenience macro where appropriate

Merged-by: James Almer <jamrial@gmail.com>
 2017-01-31 15:36:49 -03:00
James Almer
ac774cfa57 Merge commit '4efab89332' 
* commit '4efab89332':
  x86: Use *_FAST/*_SLOW CPU feature detection macros where appropriate

Merged-by: James Almer <jamrial@gmail.com>
 2017-01-31 15:08:19 -03:00
James Almer
a956164e1e Merge commit '0a39c9ac0b' 
* commit '0a39c9ac0b':
  x86: hpeldsp: Don't check for bitexact flag when initializing VP3-specific code

Merged-by: James Almer <jamrial@gmail.com>
 2017-01-31 14:59:29 -03:00
James Almer
f6005907fd Merge commit '95c1df929b' 
* commit '95c1df929b':
  x86: hpeldsp: Drop unused function parameters

Merged-by: James Almer <jamrial@gmail.com>
 2017-01-31 14:56:11 -03:00
James Almer
4d0e89ce27 Merge commit 'c3e83ad3b7' 
* commit 'c3e83ad3b7':
  x86: hpeldsp: Use EXTERNAL_SSE2_FAST where appropriate

Merged-by: James Almer <jamrial@gmail.com>
 2017-01-31 14:53:27 -03:00
James Almer
ca8a3978e5 Merge commit '1dfc3cf89d' 
* commit '1dfc3cf89d':
  x86: hpeldsp: Split off VP3-specific bits into a separate file

Merged-by: James Almer <jamrial@gmail.com>
 2017-01-31 14:49:29 -03:00
Clément Bœsch
7c300a8ed4 lavc/hevc: remove a few random spaces to reduce diff with libav 2017-01-31 17:02:24 +01:00
Clément Bœsch
78d16eb452 Merge commit 'fca3c3b619' 
* commit 'fca3c3b619':
  hevc: Add AVX2 DC IDCT

Mostly noop as we already have that code.

In the ASM, code is merged with the exception of SECTION which is kept
uppercase for consistency with the rest of the codebase.

Still in the ASM, the prototype comment is fixed to honor the '_' added
from the original commit.

idct_dc_proto() is dropped as it's not used anymore here.

Merged-by: Clément Bœsch <cboesch@gopro.com>
 2017-01-31 16:53:37 +01:00
Clément Bœsch
d0e132bab6 Merge commit '1bd890ad17' 
* commit '1bd890ad17':
  hevc: Separate adding residual to prediction from IDCT

This commit should be a noop but isn't because of the following renames:

- transform_add  → add_residual
- transform_skip → dequant
- idct_4x4_luma  → transform_4x4_luma

Merged-by: Clément Bœsch <cboesch@gopro.com>
 2017-01-31 15:31:34 +01:00
James Almer
6d4c9f2ade lossless_videodsp: rename add_hfyu_left_pred_int16 to add_left_pred_int16 
Signed-off-by: James Almer <jamrial@gmail.com>
 2017-01-12 22:53:05 -03:00
James Almer
47f212329e huffyuvdsp: move functions only used by huffyuv from lossless_videodsp 
Signed-off-by: James Almer <jamrial@gmail.com>
 2017-01-12 22:53:05 -03:00
James Almer
cf9ef83960 huffyuvencdsp: move shared functions to a new lossless_videoencdsp context 
Signed-off-by: James Almer <jamrial@gmail.com>
 2017-01-12 22:53:04 -03:00
James Almer
30c1f27299 huffyuvencdsp: move functions only used by huffyuv from lossless_videodsp 
Signed-off-by: James Almer <jamrial@gmail.com>
 2017-01-12 22:53:04 -03:00
James Almer
5ac1dd8e23 lossless_videodsp: move shared functions from huffyuvdsp 
Several codecs other than huffyuv use them.

Signed-off-by: James Almer <jamrial@gmail.com>
 2017-01-12 22:53:04 -03:00
Michael Niedermayer
aa95292043 avcodec/x86/vc1dsp_mc: Fix build with NASM 2.09.10 
make fate passes

Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
 2017-01-02 22:37:55 +01:00
John Comeau
d06518752b avcodec/x86/imdct36: fix building with nasm 2.11.05 
fixes `operation size not specified` errors as described here:
http://stackoverflow.com/questions/36854583/compiling-ffmpeg-for-kali-linux-2

I rebuilt again with yasm and made sure it didn't break that.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
 2017-01-02 20:44:16 +01:00
Paul B Mahol
6d09d6edbc avcodec/magicyuv: add 10 bit support 
Signed-off-by: Paul B Mahol <onemda@gmail.com>
 2016-12-20 13:32:15 +01:00
James Darnley
acdd2d805d avcodec/h264: resolve assert being triggered when stack is not aligned 
32-bit msvc.
 2016-12-07 22:32:19 +01:00
James Darnley
728651df06 avcodec/h264: mmx2, sse2, avx 10-bit 4:2:2 h chroma deblock/loop filter 
Yorkfield:
 - mmx2: 2.53x (504 vs. 199 cycles)
 - sse2: 3.83x (504 vs. 131 cycles)

Nehalem:
 - mmx2: 2.42x (365 vs. 151 cycles)
 - sse2: 3.56x (365 vs. 103 cycles)

Skylake:
 - mmx2: 1.81x (308 vs. 170 cycles)
 - sse2: 2.84x (308 vs. 108 cycles)
 - avx:  2.93x (308 vs. 105 cycles)
 2016-12-07 00:29:13 +01:00
James Darnley
add21d0bb3 avcodec/h264: mmx2, sse2, avx 10-bit h chroma deblock/loop filter 
Yorkfield:
 - mmx2: 2.45x (279 vs. 114 cycles)
 - sse2: 3.36x (279 vs.  83 cycles)

Nehalem:
 - mmx2: 2.10x (192 vs.  92 cycles)
 - sse2: 2.84x (192 vs.  68 cycles)

Skylake:
 - mmx2: 1.75x (170 vs.  97 cycles)
 - sse2: 2.47x (170 vs.  69 cycles)
 - avx:  2.47x (170 vs.  69 cycles)
 2016-12-07 00:29:13 +01:00
James Darnley
58ca2ef62e whitespace changes after last commit 2016-12-07 00:29:13 +01:00
James Darnley
f33714a694 avcodec/h264: clean up and expand x86 function definitions 2016-12-07 00:29:13 +01:00
James Darnley
13d71c28cc avcodec/h264: sse2 and avx 4:2:2 idct add8 10-bit functions 
Yorkfield:
 - sse2:
   - complex: 4.13x faster (1514 vs. 367 cycles)
   - simple:  4.38x faster (1836 vs. 419 cycles)

Skylake:
 - sse2:
   - complex: 3.61x faster ( 936 vs. 260 cycles)
   - simple:  3.97x faster (1126 vs. 284 cycles)
 - avx (versus sse2):
   - complex: 1.07x faster (260 vs. 244 cycles)
   - simple:  1.03x faster (284 vs. 274 cycles)
 2016-11-30 22:58:28 +01:00
James Darnley
1dae7ffa0b avcodec/h264: mmx 4:2:2 idct add8 function 
2.87 times faster (1830 vs. 638 cycles)
 2016-11-30 22:58:27 +01:00
James Darnley
815ea8c6cc avcodec/h264: mmxext 4:2:2 chroma intra deblock/loop filter 
2.1 times faster (401 vs. 194 cycles)
 2016-11-30 22:58:27 +01:00
James Almer
2de1c79b61 x86/vp9itxfm: add missing AVX2 guards 
Fixes compilation with Yasm 1.1.0 and older.

Signed-off-by: James Almer <jamrial@gmail.com>
 2016-11-18 17:01:11 -03:00
Ronald S. Bultje
83a139e3d8 vp9: add avx2 iadst16 implementations. 
Also a small cosmetic change to the avx2 idct16 version to make it
explicit that one of the arguments to the write-out macros is unused
for >=avx2 (it uses pmovzxbw instead of punpcklbw).
 2016-11-15 11:01:36 -05:00
Hendrik Leppkes
db854c6c4a Merge commit '4a081f224e' 
* commit '4a081f224e':
  libavcodec: fix constness in clobber test avcodec_open2() wrappers

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>
 2016-11-13 17:30:33 +01:00
Andreas Cadhalpun
c8a6eb58d7 doc: fix spelling errors 
Thanks to Mathieu Malaterre <malat@debian.org> for reporting the
Que/Queue typo. (https://bugs.debian.org/839542)

Reviewed-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>
 2016-10-21 23:58:47 +02:00
Rostislav Pehlivanov
d2ae5f77c6 aacenc: add SIMD optimizations for abs_pow34 and quantization 
Performance improvements:

quant_bands:
with:     681 decicycles in quant_bands, 8388453 runs,    155 skips
without: 1190 decicycles in quant_bands, 8388386 runs,    222 skips
Around 42% for the function

Twoloop coder:

abs_pow34:
with/without: 7.82s/8.17s
Around 4% for the entire encoder

Both:
with/without: 7.15s/8.17s
Around 12% for the entire encoder

Fast coder:

abs_pow34:
with/without: 3.40s/3.77s
Around 10% for the entire encoder

Both:
with/without: 3.02s/3.77s
Around 20% faster for the entire encoder

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Tested-by: Michael Niedermayer <michael@niedermayer.cc>
Reviewed-by: James Almer <jamrial@gmail.com>
 2016-10-18 21:41:18 +01:00
James Almer
42111e8543 avcodec: fix arguments on xmm/neon clobber test wrappers 
Signed-off-by: James Almer <jamrial@gmail.com>
 2016-10-02 02:15:47 -03:00
James Almer
449f263f9f avcodec: add missing xmm/neon clobber test wrappers for the new encode API 
Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
 2016-10-01 14:08:50 -03:00
Hendrik Leppkes
5ae0ad001a x86/h264_weight: use appropriate register size for weight parameters 
Fixes trac 5579

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Acked-by: Michael Niedermayer <michael@niedermayer.cc>
 2016-09-23 16:40:57 +02:00
Michael Niedermayer
bc26fe8927 avcodec/h264: Use ptrdiff_t for (bi)weight functions 
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
 2016-09-23 04:10:44 +02:00
James Almer
d950279cbf avcodec/ttadsp: cosmetics 
Clean some header includes and use the same naming scheme as
in ttaencdsp

Signed-off-by: James Almer <jamrial@gmail.com>
 2016-08-06 18:27:01 -03:00
James Almer
efc9d5c4bc x86/ttaenc: add ff_ttaenc_filter_process_{ssse3,sse4} 
Signed-off-by: James Almer <jamrial@gmail.com>
 2016-08-02 15:48:04 -03:00
Clément Bœsch
15b26e88cb Merge commit '9df889a5f1' 
* commit '9df889a5f1':
  h264: rename h264.[ch] to h264dec.[ch]

Merged-by: Clément Bœsch <u@pkh.me>
 2016-07-29 11:01:36 +02:00
Ronald S. Bultje
a4edaa0270 vp9: add mxext versions of the single-block (w=8,npx=8) h/v loopfilters. 
Each takes about 0.1% of runtime in my profiles, and they didn't have
any SIMD yet so far (we only had simd for npx=16 double-block versions).
 2016-07-26 15:59:07 -04:00
Ronald S. Bultje
7ca422bb1b vp9: add mxext versions of the single-block (w=4,npx=8) h/v loopfilters. 
Each takes about 0.5% of runtime in my profiles, and they didn't have
any SIMD yet so far (we only had simd for npx=16 double-block versions).
 2016-07-26 15:59:07 -04:00
Ronald S. Bultje
726501a34e vp9: add 32x32 idct AVX2 implementation. 
About 1.8x speedup compared to AVX version for full IDCT. Other
sub-IDCT scenarios also see speedups. Full --bench output for
idct_32x32_add_{bpp}_${subidct}_${opt} (50k cycles):

nop: 16.5
vp9_inv_dct_dct_32x32_add_8_1_c: 2284.4
vp9_inv_dct_dct_32x32_add_8_1_sse2: 145.0
vp9_inv_dct_dct_32x32_add_8_1_ssse3: 137.4
vp9_inv_dct_dct_32x32_add_8_1_avx: 137.1
vp9_inv_dct_dct_32x32_add_8_1_avx2: 73.2
vp9_inv_dct_dct_32x32_add_8_2_c: 14680.8
vp9_inv_dct_dct_32x32_add_8_2_sse2: 2617.2
vp9_inv_dct_dct_32x32_add_8_2_ssse3: 982.9
vp9_inv_dct_dct_32x32_add_8_2_avx: 958.5
vp9_inv_dct_dct_32x32_add_8_2_avx2: 704.2
vp9_inv_dct_dct_32x32_add_8_4_c: 14443.1
vp9_inv_dct_dct_32x32_add_8_4_sse2: 2717.1
vp9_inv_dct_dct_32x32_add_8_4_ssse3: 965.7
vp9_inv_dct_dct_32x32_add_8_4_avx: 1000.7
vp9_inv_dct_dct_32x32_add_8_4_avx2: 717.1
vp9_inv_dct_dct_32x32_add_8_8_c: 14436.4
vp9_inv_dct_dct_32x32_add_8_8_sse2: 2671.8
vp9_inv_dct_dct_32x32_add_8_8_ssse3: 1038.5
vp9_inv_dct_dct_32x32_add_8_8_avx: 983.0
vp9_inv_dct_dct_32x32_add_8_8_avx2: 729.4
vp9_inv_dct_dct_32x32_add_8_16_c: 14614.7
vp9_inv_dct_dct_32x32_add_8_16_sse2: 2701.7
vp9_inv_dct_dct_32x32_add_8_16_ssse3: 1334.4
vp9_inv_dct_dct_32x32_add_8_16_avx: 1276.7
vp9_inv_dct_dct_32x32_add_8_16_avx2: 719.5
vp9_inv_dct_dct_32x32_add_8_32_c: 14363.6
vp9_inv_dct_dct_32x32_add_8_32_sse2: 2575.6
vp9_inv_dct_dct_32x32_add_8_32_ssse3: 2633.9
vp9_inv_dct_dct_32x32_add_8_32_avx: 2539.6
vp9_inv_dct_dct_32x32_add_8_32_avx2: 1395.0
 2016-07-26 15:59:07 -04:00
James Almer
7a15cf42ee x86/diracdsp: make ff_put_signed_rect_clamped_10_sse4 work on x86_32 
Reviewed-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
 2016-07-20 13:43:38 -03:00
Diego Biurrun
d06dfaa5cb x86: huffyuv: Use EXTERNAL_SSSE3_FAST convenience macro where appropriate 2016-07-20 18:43:28 +02:00
Diego Biurrun
4efab89332 x86: Use *_FAST/*_SLOW CPU feature detection macros where appropriate 2016-07-20 18:43:28 +02:00
Diego Biurrun
0a39c9ac0b x86: hpeldsp: Don't check for bitexact flag when initializing VP3-specific code 
That code is only ever initialized with that flag set.
 2016-07-20 18:37:45 +02:00