Kyle Siefring ae35425ae6 Optimize convolve8 SSSE3 and AVX2 intrinsics ...

Changed the intrinsics to perform summation similiar to the way the assembly does.

The new code diverges from the assembly by preferring unsaturated additions.

Results for haswell

SSSE3
Horiz/Vert  Size  Speedup
Horiz       x4    ~32%
Horiz       x8    ~6%
Vert        x8    ~4%

AVX2
Horiz/Vert  Size  Speedup
Horiz       x16   ~16%
Vert        x16   ~14%

BUG=webm:1471

Change-Id: I7ad98ea688c904b1ba324adf8eb977873c8b8668

2017-10-24 10:39:48 -04:00

..

add_noise_sse2.asm

vpx: fix nasm build errors

2017-10-03 20:02:21 +00:00

avg_intrin_sse2.c

highbd x86: consolidate tran_low_t conversions

2017-02-06 10:43:26 -08:00

avg_pred_sse2.c

vpx_comp_avg_pred: sse2 optimization

2017-04-13 08:44:52 -07:00

avg_ssse3_x86_64.asm

bitdepth conversion: really use num elements

2017-02-16 15:02:48 +00:00

bitdepth_conversion_avx2.h

block error avx2: use tran_low_t

2017-02-16 12:39:02 -08:00

bitdepth_conversion_sse2.asm

quantize_fp highbd ssse3: use tran_low_t for coeff

2017-02-16 07:40:56 -08:00

bitdepth_conversion_sse2.h

correct bitdepth_conversion_sse2.h header guard

2017-02-16 12:43:33 -08:00

convolve_avx2.h

Optimize convolve8 SSSE3 and AVX2 intrinsics

2017-10-24 10:39:48 -04:00

convolve_ssse3.h

Optimize convolve8 SSSE3 and AVX2 intrinsics

2017-10-24 10:39:48 -04:00

convolve.h

Remove get_filter_base() and get_filter_offset() in convolve

2017-09-05 15:22:36 -07:00

deblock_sse2.asm

Fix segmentation fault caused by denoiser working with spatial SVC.

2017-02-21 09:38:28 -08:00

fwd_dct32x32_impl_avx2.h

Remove the unnecessary cast of (int16_t)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

fwd_dct32x32_impl_sse2.h

Remove the unnecessary cast of (int16_t)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

fwd_txfm_avx2.c

vpx_dsp: apply clang-format

2016-07-25 14:14:19 -07:00

fwd_txfm_impl_sse2.h

Remove the unnecessary cast of (int16_t)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

fwd_txfm_sse2.c

vpx_dsp: apply clang-format

2016-07-25 14:14:19 -07:00

fwd_txfm_sse2.h

vpx_dsp: apply clang-format

2016-07-25 14:14:19 -07:00

fwd_txfm_ssse3_x86_64.asm

Rework forward 8x8 2D-DCT ssse3 implementation

2017-01-10 12:50:55 -08:00

highbd_convolve_avx2.c

Remove get_filter_base() and get_filter_offset() in convolve

2017-09-05 15:22:36 -07:00

highbd_idct4x4_add_sse2.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct4x4_add_sse4.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct8x8_add_sse2.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct8x8_add_sse4.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct16x16_add_sse2.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct16x16_add_sse4.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct32x32_add_sse2.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_idct32x32_add_sse4.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_intrapred_intrin_sse2.c

vpxdsp: [x86] add highbd_d135_predictor functions

2017-09-29 08:56:38 -07:00

highbd_intrapred_intrin_ssse3.c

vpxdsp: [x86] add highbd_d135_predictor functions

2017-09-29 08:56:38 -07:00

highbd_intrapred_sse2.asm

Code clean of highbd_tm_predictor_32x32

2015-12-22 16:51:57 -08:00

highbd_inv_txfm_sse2.h

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_inv_txfm_sse4.h

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

highbd_loopfilter_sse2.c

Unify loopfilter function names

2016-09-29 16:25:42 -07:00

highbd_quantize_intrin_sse2.c

quantize: ignore skip_block in x86

2017-08-21 14:37:03 -07:00

highbd_sad4d_sse2.asm

Use newer x86inc.asm

2015-08-07 16:44:44 -07:00

highbd_sad_sse2.asm

Use newer x86inc.asm

2015-08-07 16:44:44 -07:00

highbd_subpel_variance_impl_sse2.asm

vpx: fix nasm build errors

2017-10-03 20:02:21 +00:00

highbd_variance_impl_sse2.asm

Move variance functions to vpx_dsp

2015-05-26 12:01:52 -07:00

highbd_variance_sse2.c

Resolve -Wshorten-64-to-32 in highbd variance.

2017-04-05 17:34:02 -07:00

intrapred_sse2.asm

vpx: fix nasm build errors

2017-10-03 20:02:21 +00:00

intrapred_ssse3.asm

Slow pshufb removal in 3 intra prediction functions.

2016-06-02 10:55:58 -07:00

inv_txfm_sse2.c

fix signed integer overflow of idct

2017-09-27 11:17:54 -07:00

inv_txfm_sse2.h

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

inv_txfm_ssse3.c

Remove the unnecessary upcasts of (int)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

inv_txfm_ssse3.h

Remove the unnecessary cast of (int16_t)cospi_{1...31}_64

2017-09-20 14:13:26 -07:00

inv_wht_sse2.asm

bitdepth conversion: really use num elements

2017-02-16 15:02:48 +00:00

loopfilter_avx2.c

Unify loopfilter function names

2016-09-29 16:25:42 -07:00

loopfilter_sse2.c

Unify loopfilter function names

2016-09-29 16:25:42 -07:00

mem_sse2.h

Add 4 to 3 scaling SSSE3 optimization

2017-10-16 15:42:42 -07:00

quantize_avx.c

Revert "Revert "quantize avx: copy 32x32 implementation""

2017-09-12 14:25:38 -07:00

quantize_sse2.c

quantize sse2: copy opts from ssse3

2017-08-22 13:01:44 -07:00

quantize_ssse3.c

quantize ssse3: copy implementation to intrinsics

2017-08-24 07:47:51 -07:00

sad4d_avx2.c

vpx_dsp: apply clang-format

2016-07-25 14:14:19 -07:00

sad4d_sse2.asm

Code clean of sad4xNx4D_sse

2015-12-17 17:43:46 -08:00

sad_avx2.c

vpx_dsp: apply clang-format

2016-07-25 14:14:19 -07:00

sad_sse2.asm

sad_sse2: fix sad4xN(_avg) on windows

2015-12-18 19:19:32 -08:00

sad_sse3.asm

Move shared SAD code to vpx_dsp

2015-05-06 16:58:20 -07:00

sad_sse4.asm

Move shared SAD code to vpx_dsp

2015-05-06 16:58:20 -07:00

sad_ssse3.asm

Move shared SAD code to vpx_dsp

2015-05-06 16:58:20 -07:00

ssim_opt_x86_64.asm

ssim: Replace unsigned long with uint32_t.

2015-08-07 11:48:31 -07:00

subpel_variance_sse2.asm

Code clean of sub_pixel_variance4xh -- 2

2016-05-24 04:44:05 -07:00

subtract_sse2.asm

Use newer x86inc.asm

2015-08-07 16:44:44 -07:00

sum_squares_sse2.c

vp9_rdopt: correct size to vpx_sum_squares_2d_i16

2017-03-22 12:04:33 -07:00

transpose_sse2.h

Add transpose_8bit_{4x4,8x8}() x86 optimization

2017-10-03 10:00:30 -07:00

txfm_common_sse2.h

Refactor highbd idct 4x4 and 8x8 x86 functions

2017-07-27 18:01:03 -07:00

variance_avx2.c

vpx_dsp: get32x32var_avx2() cleanup

2017-08-18 13:44:09 -07:00

variance_sse2.c

vpx_variance16x16_sse2: correct cast order

2017-07-25 16:45:40 -07:00

vpx_asm_stubs.c

Remove get_filter_base() and get_filter_offset() in convolve

2017-09-05 15:22:36 -07:00

vpx_convolve_copy_sse2.asm

Remove get_filter_base() and get_filter_offset() in convolve

2017-09-05 15:22:36 -07:00

vpx_high_subpixel_8t_sse2.asm

Code refactor on InterpKernel

2015-07-31 10:27:33 -07:00

vpx_high_subpixel_bilinear_sse2.asm

Code refactor on InterpKernel

2015-07-31 10:27:33 -07:00

vpx_subpixel_8t_intrin_avx2.c

Refactor x86/vpx_subpixel_8t_intrin_avx2.c

2017-10-17 11:57:40 -04:00

vpx_subpixel_8t_intrin_ssse3.c

Optimize convolve8 SSSE3 and AVX2 intrinsics

2017-10-24 10:39:48 -04:00

vpx_subpixel_8t_sse2.asm

Code refactor on InterpKernel

2015-07-31 10:27:33 -07:00

vpx_subpixel_8t_ssse3.asm

Refactor x86/vpx_subpixel_8t_intrin_ssse3.c

2017-10-03 13:02:05 -07:00

vpx_subpixel_bilinear_sse2.asm

Code refactor on InterpKernel

2015-07-31 10:27:33 -07:00

vpx_subpixel_bilinear_ssse3.asm

improve vpx_filter_block1d* based on replace paddsw+psrlw to pmulhrsw

2016-06-27 17:50:45 +00:00