Christophe Gisquet e2946e5c34 x86: sbrdsp: implement SSE qmf_deint_bfly ...

From 312 to 89/68 (sse/sse2) cycles on Arrandale and Win64.
Sandybridge: 68/47 cycles.

Having a loop counter is a 7 cycle gain.
Unrolling is another 7 cycle gain.
Working in reverse scan is another 6 cycles.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

2013-04-08 02:26:34 +02:00

..

ac3dsp_init.c

Merge commit 'b6649ab5037fb55f78c2606f3d23cea0867cdeaa'

2013-03-28 11:20:41 +01:00

ac3dsp.asm

…

cabac.h

…

cavsdsp.c

Merge commit 'a8b6015823e628047a45916404c00044c5e80415'

2013-03-13 14:18:53 +01:00

constants.c

dsputil: make selectable.

2013-03-12 19:56:58 +01:00

dct32.asm

…

deinterlace.asm

…

dirac_dwt.c

…

dirac_dwt.h

…

diracdsp_mmx.c

dirac: Only use MMX if MMX is available.

2013-01-28 20:00:55 +01:00

diracdsp_mmx.h

…

diracdsp_yasm.asm

…

dnxhdenc.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

dsputil_mmx.c

dsputil_mmx: remove unused variables

2013-03-13 15:07:46 +01:00

dsputil_mmx.h

Merge commit 'a8b6015823e628047a45916404c00044c5e80415'

2013-03-13 14:18:53 +01:00

dsputil_qns_template.c

avcodec/x86/dsputil_qns_template: use av_assert

2013-03-26 04:08:28 +01:00

dsputil_rnd_template.c

Move x86 half-pel assembly from dsputil to hpeldsp.

2013-03-13 03:59:23 +01:00

dsputil.asm

x86/dsputil.asm: make unaligned bswap actually work

2013-03-26 13:07:46 +01:00

dsputilenc_mmx.c

Merge remote-tracking branch 'qatar/master'

2013-02-09 14:45:45 +01:00

dsputilenc.asm

…

dwt_yasm.asm

…

fdct.c

Merge remote-tracking branch 'qatar/master'

2013-02-09 14:45:45 +01:00

fft_init.c

Merge commit 'c242bbd8b6939507a1a6fb64101b0553d92d303f'

2013-02-26 13:05:10 +01:00

fft.asm

…

fft.h

…

fmtconvert_init.c

Merge commit 'c242bbd8b6939507a1a6fb64101b0553d92d303f'

2013-02-26 13:05:10 +01:00

fmtconvert.asm

…

fpelbase.asm

Move x86 half-pel assembly from dsputil to hpeldsp.

2013-03-13 03:59:23 +01:00

h263_loopfilter.asm

x86/h263_loopfilter: Fix author attribution after code has been moved/splited around

2013-02-19 12:51:00 +01:00

h264_chromamc_10bit.asm

…

h264_chromamc.asm

dsputil: make selectable.

2013-03-12 19:56:58 +01:00

h264_deblock_10bit.asm

…

h264_deblock.asm

dsputil: make selectable.

2013-03-12 19:56:58 +01:00

h264_i386.h

…

h264_idct_10bit.asm

h264: integrate clear_blocks calls with IDCT.

2013-02-19 16:25:50 +01:00

h264_idct.asm

h264: integrate clear_blocks calls with IDCT.

2013-02-19 16:25:50 +01:00

h264_intrapred_10bit.asm

…

h264_intrapred_init.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

h264_intrapred.asm

…

h264_qpel_8bit.asm

Merge commit 'e8c52271c45ec27d783e74238dcfad0c2008731c'

2013-03-01 13:39:47 +01:00

h264_qpel_10bit.asm

…

h264_qpel.c

x86: Change a missed occurrance of int to ptrdiff_t for strides

2013-03-24 12:06:53 +02:00

h264_weight_10bit.asm

…

h264_weight.asm

…

h264chroma_init.c

h264chroma: Remove duplicate 9/10 bit functions

2013-02-19 22:33:19 +02:00

h264dsp_init.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

hpeldsp_avg_template.c

Move x86 half-pel assembly from dsputil to hpeldsp.

2013-03-13 03:59:23 +01:00

hpeldsp_init.c

Fix compilation with --disable-mmx.

2013-03-22 13:00:50 +01:00

hpeldsp_rnd_template.c

Move x86 half-pel assembly from dsputil to hpeldsp.

2013-03-13 03:59:23 +01:00

hpeldsp.asm

x86/qpel: move fullpel and l2 functions to separate file.

2013-03-09 17:25:30 +01:00

idct_mmx_xvid.c

…

idct_sse2_xvid.c

Merge commit 'c242bbd8b6939507a1a6fb64101b0553d92d303f'

2013-02-26 13:05:10 +01:00

idct_xvid.h

…

imdct36.asm

…

lpc.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

Makefile

Fix compilation with --enable-decoder=webp --disable-decoder=vp8

2013-03-30 08:25:44 +01:00

mathops.h

…

mlpdsp.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

motion_est.c

Merge commit 'c242bbd8b6939507a1a6fb64101b0553d92d303f'

2013-02-26 13:05:10 +01:00

mpeg4qpel.asm

Merge commit '304b806cb524fb040f8e09a241040f1af2cb820b'

2013-02-12 13:26:22 +01:00

mpegaudiodec.c

Merge commit '218aefce4472dc02ee3f12830a9a894bf7916da9'

2013-02-09 14:33:57 +01:00

mpegvideo.c

x86/mpegvideo: switch to av_assert

2013-03-22 22:57:23 +01:00

mpegvideoenc_template.c

…

mpegvideoenc.c

Merge remote-tracking branch 'qatar/master'

2013-02-09 14:45:45 +01:00

pngdsp_init.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

pngdsp.asm

…

proresdsp_init.c

…

proresdsp.asm

…

qpelbase.asm

Move x86 half-pel assembly from dsputil to hpeldsp.

2013-03-13 03:59:23 +01:00

rv34dsp_init.c

Remove unnecessary dsputil.h #includes

2013-02-26 00:51:34 +01:00

rv34dsp.asm

…

rv40dsp_init.c

Merge commit 'a8b6015823e628047a45916404c00044c5e80415'

2013-03-13 14:18:53 +01:00

rv40dsp.asm

…

sbrdsp_init.c

x86: sbrdsp: implement SSE qmf_deint_bfly

2013-04-08 02:26:34 +02:00

sbrdsp.asm

x86: sbrdsp: implement SSE qmf_deint_bfly

2013-04-08 02:26:34 +02:00

simple_idct.c

Merge commit 'c242bbd8b6939507a1a6fb64101b0553d92d303f'

2013-02-26 13:05:10 +01:00

snowdsp.c

…

v210-init.c

…

v210.asm

…

vc1dsp_init.c

dsputil: convert remaining functions to use ptrdiff_t strides

2013-03-12 18:26:42 +01:00

vc1dsp_mmx.c

Merge commit '3b2d0ec473b036bdd0a5bc0d896fd5292915f44d'

2013-03-27 11:32:45 +01:00

vc1dsp.asm

…

vc1dsp.h

…

videodsp_init.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

videodsp.asm

Merge commit 'b6649ab5037fb55f78c2606f3d23cea0867cdeaa'

2013-03-28 11:20:41 +01:00

vorbisdsp_init.c

Merge commit '620289a20e022b9c16c10d546ef86cc0bb77cc84'

2013-02-06 13:27:24 +01:00

vorbisdsp.asm

…

vp3dsp_init.c

vp3/x86: use full transpose for all IDCTs.

2013-03-12 22:54:10 +01:00

vp3dsp.asm

vp3: use hpeldsp instead of dsputil for half-pel functions.

2013-03-13 03:55:33 +01:00

vp8dsp_init.c

Merge commit 'b6649ab5037fb55f78c2606f3d23cea0867cdeaa'

2013-03-28 11:20:41 +01:00

vp8dsp.asm

dsputil: make selectable.

2013-03-12 19:56:58 +01:00

vp56_arith.h

…

vp56dsp_init.c

Merge commit 'c242bbd8b6939507a1a6fb64101b0553d92d303f'

2013-02-26 13:05:10 +01:00

vp56dsp.asm

…

w64xmmtest.c

x86: Remove win64 xmm clobbering wrappers for the now removed avcodec_encode_video function

2013-03-23 23:37:27 +02:00