1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00
FFmpeg/libavcodec/aarch64/h264qpel_neon.S

1468 lines
52 KiB
ArmAsm
Raw Normal View History

/*
* Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
* Copyright (c) 2013 Janne Grunau <janne-libav@jannau.net>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/aarch64/asm.S"
#include "neon.S"
/* H.264 qpel MC */
.macro lowpass_const r
movz \r, #20, lsl #16
movk \r, #5
mov v6.s[0], \r
.endm
//trashes v0-v5
.macro lowpass_8 r0, r1, r2, r3, d0, d1, narrow=1
ext v2.8b, \r0\().8b, \r1\().8b, #2
ext v3.8b, \r0\().8b, \r1\().8b, #3
uaddl v2.8h, v2.8b, v3.8b
ext v4.8b, \r0\().8b, \r1\().8b, #1
ext v5.8b, \r0\().8b, \r1\().8b, #4
uaddl v4.8h, v4.8b, v5.8b
ext v1.8b, \r0\().8b, \r1\().8b, #5
uaddl \d0\().8h, \r0\().8b, v1.8b
ext v0.8b, \r2\().8b, \r3\().8b, #2
mla \d0\().8h, v2.8h, v6.h[1]
ext v1.8b, \r2\().8b, \r3\().8b, #3
uaddl v0.8h, v0.8b, v1.8b
ext v1.8b, \r2\().8b, \r3\().8b, #1
mls \d0\().8h, v4.8h, v6.h[0]
ext v3.8b, \r2\().8b, \r3\().8b, #4
uaddl v1.8h, v1.8b, v3.8b
ext v2.8b, \r2\().8b, \r3\().8b, #5
uaddl \d1\().8h, \r2\().8b, v2.8b
mla \d1\().8h, v0.8h, v6.h[1]
mls \d1\().8h, v1.8h, v6.h[0]
.if \narrow
sqrshrun \d0\().8b, \d0\().8h, #5
sqrshrun \d1\().8b, \d1\().8h, #5
.endif
.endm
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
//trashes v0-v4
.macro lowpass_8_v r0, r1, r2, r3, r4, r5, r6, d0, d1, narrow=1
uaddl v2.8h, \r2\().8b, \r3\().8b
uaddl v0.8h, \r3\().8b, \r4\().8b
uaddl v4.8h, \r1\().8b, \r4\().8b
uaddl v1.8h, \r2\().8b, \r5\().8b
uaddl \d0\().8h, \r0\().8b, \r5\().8b
uaddl \d1\().8h, \r1\().8b, \r6\().8b
mla \d0\().8h, v2.8h, v6.h[1]
mls \d0\().8h, v4.8h, v6.h[0]
mla \d1\().8h, v0.8h, v6.h[1]
mls \d1\().8h, v1.8h, v6.h[0]
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
.if \narrow
sqrshrun \d0\().8b, \d0\().8h, #5
sqrshrun \d1\().8b, \d1\().8h, #5
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
.endif
.endm
//trashes v0-v5, v7, v30-v31
.macro lowpass_8H r0, r1
ext v0.16b, \r0\().16b, \r0\().16b, #2
ext v1.16b, \r0\().16b, \r0\().16b, #3
uaddl v0.8h, v0.8b, v1.8b
ext v2.16b, \r0\().16b, \r0\().16b, #1
ext v3.16b, \r0\().16b, \r0\().16b, #4
uaddl v2.8h, v2.8b, v3.8b
ext v30.16b, \r0\().16b, \r0\().16b, #5
uaddl \r0\().8h, \r0\().8b, v30.8b
ext v4.16b, \r1\().16b, \r1\().16b, #2
mla \r0\().8h, v0.8h, v6.h[1]
ext v5.16b, \r1\().16b, \r1\().16b, #3
uaddl v4.8h, v4.8b, v5.8b
ext v7.16b, \r1\().16b, \r1\().16b, #1
mls \r0\().8h, v2.8h, v6.h[0]
ext v0.16b, \r1\().16b, \r1\().16b, #4
uaddl v7.8h, v7.8b, v0.8b
ext v31.16b, \r1\().16b, \r1\().16b, #5
uaddl \r1\().8h, \r1\().8b, v31.8b
mla \r1\().8h, v4.8h, v6.h[1]
mls \r1\().8h, v7.8h, v6.h[0]
.endm
// trashes v2-v5, v30
.macro lowpass_8_1 r0, r1, d0, narrow=1
ext v2.8b, \r0\().8b, \r1\().8b, #2
ext v3.8b, \r0\().8b, \r1\().8b, #3
uaddl v2.8h, v2.8b, v3.8b
ext v4.8b, \r0\().8b, \r1\().8b, #1
ext v5.8b, \r0\().8b, \r1\().8b, #4
uaddl v4.8h, v4.8b, v5.8b
ext v30.8b, \r0\().8b, \r1\().8b, #5
uaddl \d0\().8h, \r0\().8b, v30.8b
mla \d0\().8h, v2.8h, v6.h[1]
mls \d0\().8h, v4.8h, v6.h[0]
.if \narrow
sqrshrun \d0\().8b, \d0\().8h, #5
.endif
.endm
// trashed v0-v7
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
.macro lowpass_8.16 r0, r1, r2, r3, r4, r5
saddl v5.4s, \r2\().4h, \r3\().4h
saddl2 v1.4s, \r2\().8h, \r3\().8h
saddl v6.4s, \r1\().4h, \r4\().4h
saddl2 v2.4s, \r1\().8h, \r4\().8h
saddl v0.4s, \r0\().4h, \r5\().4h
saddl2 v4.4s, \r0\().8h, \r5\().8h
shl v3.4s, v5.4s, #4
shl v5.4s, v5.4s, #2
shl v7.4s, v6.4s, #2
add v5.4s, v5.4s, v3.4s
add v6.4s, v6.4s, v7.4s
shl v3.4s, v1.4s, #4
shl v1.4s, v1.4s, #2
shl v7.4s, v2.4s, #2
add v1.4s, v1.4s, v3.4s
add v2.4s, v2.4s, v7.4s
add v5.4s, v5.4s, v0.4s
sub v5.4s, v5.4s, v6.4s
add v1.4s, v1.4s, v4.4s
sub v1.4s, v1.4s, v2.4s
rshrn v5.4h, v5.4s, #10
rshrn2 v5.8h, v1.4s, #10
sqxtun \r0\().8b, v5.8h
.endm
function put_h264_qpel16_h_lowpass_neon_packed
mov x4, x30
mov x12, #16
mov x3, #8
bl put_h264_qpel8_h_lowpass_neon
sub x1, x1, x2, lsl #4
add x1, x1, #8
mov x12, #16
mov x30, x4
b put_h264_qpel8_h_lowpass_neon
endfunc
.macro h264_qpel_h_lowpass type
function \type\()_h264_qpel16_h_lowpass_neon
mov x13, x30
mov x12, #16
bl \type\()_h264_qpel8_h_lowpass_neon
sub x0, x0, x3, lsl #4
sub x1, x1, x2, lsl #4
add x0, x0, #8
add x1, x1, #8
mov x12, #16
mov x30, x13
endfunc
function \type\()_h264_qpel8_h_lowpass_neon
1: ld1 {v28.8b, v29.8b}, [x1], x2
ld1 {v16.8b, v17.8b}, [x1], x2
subs x12, x12, #2
lowpass_8 v28, v29, v16, v17, v28, v16
.ifc \type,avg
ld1 {v2.8b}, [x0], x3
ld1 {v3.8b}, [x0]
urhadd v28.8b, v28.8b, v2.8b
urhadd v16.8b, v16.8b, v3.8b
sub x0, x0, x3
.endif
st1 {v28.8b}, [x0], x3
st1 {v16.8b}, [x0], x3
b.ne 1b
ret
endfunc
.endm
h264_qpel_h_lowpass put
h264_qpel_h_lowpass avg
.macro h264_qpel_h_lowpass_l2 type
function \type\()_h264_qpel16_h_lowpass_l2_neon
mov x13, x30
mov x12, #16
bl \type\()_h264_qpel8_h_lowpass_l2_neon
sub x0, x0, x2, lsl #4
sub x1, x1, x2, lsl #4
sub x3, x3, x2, lsl #4
add x0, x0, #8
add x1, x1, #8
add x3, x3, #8
mov x12, #16
mov x30, x13
endfunc
function \type\()_h264_qpel8_h_lowpass_l2_neon
1: ld1 {v26.8b, v27.8b}, [x1], x2
ld1 {v16.8b, v17.8b}, [x1], x2
ld1 {v28.8b}, [x3], x2
ld1 {v29.8b}, [x3], x2
subs x12, x12, #2
lowpass_8 v26, v27, v16, v17, v26, v27
urhadd v26.8b, v26.8b, v28.8b
urhadd v27.8b, v27.8b, v29.8b
.ifc \type,avg
ld1 {v2.8b}, [x0], x2
ld1 {v3.8b}, [x0]
urhadd v26.8b, v26.8b, v2.8b
urhadd v27.8b, v27.8b, v3.8b
sub x0, x0, x2
.endif
st1 {v26.8b}, [x0], x2
st1 {v27.8b}, [x0], x2
b.ne 1b
ret
endfunc
.endm
h264_qpel_h_lowpass_l2 put
h264_qpel_h_lowpass_l2 avg
function put_h264_qpel16_v_lowpass_neon_packed
mov x4, x30
mov x2, #8
bl put_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #2
bl put_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #8
bl put_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #2
mov x30, x4
b put_h264_qpel8_v_lowpass_neon
endfunc
.macro h264_qpel_v_lowpass type
function \type\()_h264_qpel16_v_lowpass_neon
mov x4, x30
bl \type\()_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #2
bl \type\()_h264_qpel8_v_lowpass_neon
sub x0, x0, x2, lsl #4
add x0, x0, #8
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #8
bl \type\()_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #2
mov x30, x4
endfunc
function \type\()_h264_qpel8_v_lowpass_neon
ld1 {v16.8b}, [x1], x3
ld1 {v17.8b}, [x1], x3
ld1 {v18.8b}, [x1], x3
ld1 {v19.8b}, [x1], x3
ld1 {v20.8b}, [x1], x3
ld1 {v21.8b}, [x1], x3
ld1 {v22.8b}, [x1], x3
ld1 {v23.8b}, [x1], x3
ld1 {v24.8b}, [x1], x3
ld1 {v25.8b}, [x1], x3
ld1 {v26.8b}, [x1], x3
ld1 {v27.8b}, [x1], x3
ld1 {v28.8b}, [x1]
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17
lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19
lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21
lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23
.ifc \type,avg
ld1 {v24.8b}, [x0], x2
ld1 {v25.8b}, [x0], x2
ld1 {v26.8b}, [x0], x2
urhadd v16.8b, v16.8b, v24.8b
ld1 {v27.8b}, [x0], x2
urhadd v17.8b, v17.8b, v25.8b
ld1 {v28.8b}, [x0], x2
urhadd v18.8b, v18.8b, v26.8b
ld1 {v29.8b}, [x0], x2
urhadd v19.8b, v19.8b, v27.8b
ld1 {v30.8b}, [x0], x2
urhadd v20.8b, v20.8b, v28.8b
ld1 {v31.8b}, [x0], x2
urhadd v21.8b, v21.8b, v29.8b
urhadd v22.8b, v22.8b, v30.8b
urhadd v23.8b, v23.8b, v31.8b
sub x0, x0, x2, lsl #3
.endif
st1 {v16.8b}, [x0], x2
st1 {v17.8b}, [x0], x2
st1 {v18.8b}, [x0], x2
st1 {v19.8b}, [x0], x2
st1 {v20.8b}, [x0], x2
st1 {v21.8b}, [x0], x2
st1 {v22.8b}, [x0], x2
st1 {v23.8b}, [x0], x2
ret
endfunc
.endm
h264_qpel_v_lowpass put
h264_qpel_v_lowpass avg
.macro h264_qpel_v_lowpass_l2 type
function \type\()_h264_qpel16_v_lowpass_l2_neon
mov x4, x30
bl \type\()_h264_qpel8_v_lowpass_l2_neon
sub x1, x1, x3, lsl #2
bl \type\()_h264_qpel8_v_lowpass_l2_neon
sub x0, x0, x3, lsl #4
sub x12, x12, x2, lsl #4
add x0, x0, #8
add x12, x12, #8
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #8
bl \type\()_h264_qpel8_v_lowpass_l2_neon
sub x1, x1, x3, lsl #2
mov x30, x4
endfunc
function \type\()_h264_qpel8_v_lowpass_l2_neon
ld1 {v16.8b}, [x1], x3
ld1 {v17.8b}, [x1], x3
ld1 {v18.8b}, [x1], x3
ld1 {v19.8b}, [x1], x3
ld1 {v20.8b}, [x1], x3
ld1 {v21.8b}, [x1], x3
ld1 {v22.8b}, [x1], x3
ld1 {v23.8b}, [x1], x3
ld1 {v24.8b}, [x1], x3
ld1 {v25.8b}, [x1], x3
ld1 {v26.8b}, [x1], x3
ld1 {v27.8b}, [x1], x3
ld1 {v28.8b}, [x1]
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17
lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19
lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21
lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23
ld1 {v24.8b}, [x12], x2
ld1 {v25.8b}, [x12], x2
ld1 {v26.8b}, [x12], x2
ld1 {v27.8b}, [x12], x2
ld1 {v28.8b}, [x12], x2
urhadd v16.8b, v24.8b, v16.8b
urhadd v17.8b, v25.8b, v17.8b
ld1 {v29.8b}, [x12], x2
urhadd v18.8b, v26.8b, v18.8b
urhadd v19.8b, v27.8b, v19.8b
ld1 {v30.8b}, [x12], x2
urhadd v20.8b, v28.8b, v20.8b
urhadd v21.8b, v29.8b, v21.8b
ld1 {v31.8b}, [x12], x2
urhadd v22.8b, v30.8b, v22.8b
urhadd v23.8b, v31.8b, v23.8b
.ifc \type,avg
ld1 {v24.8b}, [x0], x3
ld1 {v25.8b}, [x0], x3
ld1 {v26.8b}, [x0], x3
urhadd v16.8b, v16.8b, v24.8b
ld1 {v27.8b}, [x0], x3
urhadd v17.8b, v17.8b, v25.8b
ld1 {v28.8b}, [x0], x3
urhadd v18.8b, v18.8b, v26.8b
ld1 {v29.8b}, [x0], x3
urhadd v19.8b, v19.8b, v27.8b
ld1 {v30.8b}, [x0], x3
urhadd v20.8b, v20.8b, v28.8b
ld1 {v31.8b}, [x0], x3
urhadd v21.8b, v21.8b, v29.8b
urhadd v22.8b, v22.8b, v30.8b
urhadd v23.8b, v23.8b, v31.8b
sub x0, x0, x3, lsl #3
.endif
st1 {v16.8b}, [x0], x3
st1 {v17.8b}, [x0], x3
st1 {v18.8b}, [x0], x3
st1 {v19.8b}, [x0], x3
st1 {v20.8b}, [x0], x3
st1 {v21.8b}, [x0], x3
st1 {v22.8b}, [x0], x3
st1 {v23.8b}, [x0], x3
ret
endfunc
.endm
h264_qpel_v_lowpass_l2 put
h264_qpel_v_lowpass_l2 avg
function put_h264_qpel8_hv_lowpass_neon_top
lowpass_const w12
ld1 {v16.8h}, [x1], x3
ld1 {v17.8h}, [x1], x3
ld1 {v18.8h}, [x1], x3
ld1 {v19.8h}, [x1], x3
ld1 {v20.8h}, [x1], x3
ld1 {v21.8h}, [x1], x3
ld1 {v22.8h}, [x1], x3
ld1 {v23.8h}, [x1], x3
ld1 {v24.8h}, [x1], x3
ld1 {v25.8h}, [x1], x3
ld1 {v26.8h}, [x1], x3
ld1 {v27.8h}, [x1], x3
ld1 {v28.8h}, [x1]
lowpass_8H v16, v17
lowpass_8H v18, v19
lowpass_8H v20, v21
lowpass_8H v22, v23
lowpass_8H v24, v25
lowpass_8H v26, v27
lowpass_8H v28, v29
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
lowpass_8.16 v16, v17, v18, v19, v20, v21
lowpass_8.16 v17, v18, v19, v20, v21, v22
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
lowpass_8.16 v18, v19, v20, v21, v22, v23
lowpass_8.16 v19, v20, v21, v22, v23, v24
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
lowpass_8.16 v20, v21, v22, v23, v24, v25
lowpass_8.16 v21, v22, v23, v24, v25, v26
aarch64: h264qpel: Do vertical filtering without transposing This gives rather big speedups on these functions: Before: put_h264_qpel_8_mc01_8_neon: 241.0 131.5 138.7 put_h264_qpel_8_mc02_8_neon: 214.7 121.2 127.5 put_h264_qpel_8_mc03_8_neon: 242.5 131.2 135.7 put_h264_qpel_8_mc11_8_neon: 421.2 218.7 251.0 put_h264_qpel_8_mc12_8_neon: 878.0 509.5 537.5 put_h264_qpel_8_mc13_8_neon: 423.7 217.0 252.0 put_h264_qpel_8_mc21_8_neon: 858.2 479.5 514.0 put_h264_qpel_8_mc22_8_neon: 649.7 385.2 403.0 put_h264_qpel_8_mc23_8_neon: 860.2 476.5 517.7 put_h264_qpel_8_mc31_8_neon: 437.2 219.5 252.5 put_h264_qpel_8_mc32_8_neon: 892.5 510.5 546.0 put_h264_qpel_8_mc33_8_neon: 438.2 218.5 257.0 put_h264_qpel_16_mc01_8_neon: 944.2 509.7 546.7 put_h264_qpel_16_mc02_8_neon: 878.7 469.5 509.7 put_h264_qpel_16_mc03_8_neon: 945.7 510.7 557.0 put_h264_qpel_16_mc11_8_neon: 1663.2 858.5 979.5 put_h264_qpel_16_mc12_8_neon: 3510.2 2027.7 2112.7 put_h264_qpel_16_mc13_8_neon: 1664.7 857.5 980.5 put_h264_qpel_16_mc21_8_neon: 3366.2 1928.5 2030.5 put_h264_qpel_16_mc22_8_neon: 2584.7 1514.7 1590.2 put_h264_qpel_16_mc23_8_neon: 3367.7 1927.7 2035.0 put_h264_qpel_16_mc31_8_neon: 1716.7 849.7 997.0 put_h264_qpel_16_mc32_8_neon: 3564.0 2044.2 3835.2 put_h264_qpel_16_mc33_8_neon: 1717.7 863.0 989.5 After: put_h264_qpel_8_mc01_8_neon: 136.0 73.7 76.0 put_h264_qpel_8_mc02_8_neon: 108.7 65.0 64.0 put_h264_qpel_8_mc03_8_neon: 137.5 72.7 73.0 put_h264_qpel_8_mc11_8_neon: 316.2 159.0 188.5 put_h264_qpel_8_mc12_8_neon: 653.0 375.5 384.7 put_h264_qpel_8_mc13_8_neon: 318.7 165.5 189.5 put_h264_qpel_8_mc21_8_neon: 739.2 385.7 432.5 put_h264_qpel_8_mc22_8_neon: 530.7 295.5 309.5 put_h264_qpel_8_mc23_8_neon: 741.2 393.7 421.0 put_h264_qpel_8_mc31_8_neon: 332.2 162.5 190.0 put_h264_qpel_8_mc32_8_neon: 667.5 378.2 390.5 put_h264_qpel_8_mc33_8_neon: 332.7 166.5 195.5 put_h264_qpel_16_mc01_8_neon: 524.2 285.2 294.0 put_h264_qpel_16_mc02_8_neon: 454.7 252.2 250.2 put_h264_qpel_16_mc03_8_neon: 525.7 286.0 283.0 put_h264_qpel_16_mc11_8_neon: 1243.2 630.7 726.7 put_h264_qpel_16_mc12_8_neon: 2610.2 1479.7 1481.2 put_h264_qpel_16_mc13_8_neon: 1250.5 631.7 727.7 put_h264_qpel_16_mc21_8_neon: 2890.2 1571.2 1679.7 put_h264_qpel_16_mc22_8_neon: 2108.7 1177.5 1223.5 put_h264_qpel_16_mc23_8_neon: 2891.7 1578.7 1667.7 put_h264_qpel_16_mc31_8_neon: 1296.7 630.5 752.5 put_h264_qpel_16_mc32_8_neon: 2664.0 1483.2 1503.5 put_h264_qpel_16_mc33_8_neon: 1297.7 632.5 747.2 I.e. overall a 20%-60% reduction in runtime of these functions. Signed-off-by: Martin Storsjö <martin@martin.st>
2021-09-03 12:56:05 +02:00
lowpass_8.16 v22, v23, v24, v25, v26, v27
lowpass_8.16 v23, v24, v25, v26, v27, v28
ret
endfunc
.macro h264_qpel8_hv_lowpass type
function \type\()_h264_qpel8_hv_lowpass_neon
mov x10, x30
bl put_h264_qpel8_hv_lowpass_neon_top
.ifc \type,avg
ld1 {v0.8b}, [x0], x2
ld1 {v1.8b}, [x0], x2
ld1 {v2.8b}, [x0], x2
urhadd v16.8b, v16.8b, v0.8b
ld1 {v3.8b}, [x0], x2
urhadd v17.8b, v17.8b, v1.8b
ld1 {v4.8b}, [x0], x2
urhadd v18.8b, v18.8b, v2.8b
ld1 {v5.8b}, [x0], x2
urhadd v19.8b, v19.8b, v3.8b
ld1 {v6.8b}, [x0], x2
urhadd v20.8b, v20.8b, v4.8b
ld1 {v7.8b}, [x0], x2
urhadd v21.8b, v21.8b, v5.8b
urhadd v22.8b, v22.8b, v6.8b
urhadd v23.8b, v23.8b, v7.8b
sub x0, x0, x2, lsl #3
.endif
st1 {v16.8b}, [x0], x2
st1 {v17.8b}, [x0], x2
st1 {v18.8b}, [x0], x2
st1 {v19.8b}, [x0], x2
st1 {v20.8b}, [x0], x2
st1 {v21.8b}, [x0], x2
st1 {v22.8b}, [x0], x2
st1 {v23.8b}, [x0], x2
ret x10
endfunc
.endm
h264_qpel8_hv_lowpass put
h264_qpel8_hv_lowpass avg
.macro h264_qpel8_hv_lowpass_l2 type
function \type\()_h264_qpel8_hv_lowpass_l2_neon
mov x10, x30
bl put_h264_qpel8_hv_lowpass_neon_top
ld1 {v0.8b, v1.8b}, [x2], #16
ld1 {v2.8b, v3.8b}, [x2], #16
urhadd v0.8b, v0.8b, v16.8b
urhadd v1.8b, v1.8b, v17.8b
ld1 {v4.8b, v5.8b}, [x2], #16
urhadd v2.8b, v2.8b, v18.8b
urhadd v3.8b, v3.8b, v19.8b
ld1 {v6.8b, v7.8b}, [x2], #16
urhadd v4.8b, v4.8b, v20.8b
urhadd v5.8b, v5.8b, v21.8b
urhadd v6.8b, v6.8b, v22.8b
urhadd v7.8b, v7.8b, v23.8b
.ifc \type,avg
ld1 {v16.8b}, [x0], x3
ld1 {v17.8b}, [x0], x3
ld1 {v18.8b}, [x0], x3
urhadd v0.8b, v0.8b, v16.8b
ld1 {v19.8b}, [x0], x3
urhadd v1.8b, v1.8b, v17.8b
ld1 {v20.8b}, [x0], x3
urhadd v2.8b, v2.8b, v18.8b
ld1 {v21.8b}, [x0], x3
urhadd v3.8b, v3.8b, v19.8b
ld1 {v22.8b}, [x0], x3
urhadd v4.8b, v4.8b, v20.8b
ld1 {v23.8b}, [x0], x3
urhadd v5.8b, v5.8b, v21.8b
urhadd v6.8b, v6.8b, v22.8b
urhadd v7.8b, v7.8b, v23.8b
sub x0, x0, x3, lsl #3
.endif
st1 {v0.8b}, [x0], x3
st1 {v1.8b}, [x0], x3
st1 {v2.8b}, [x0], x3
st1 {v3.8b}, [x0], x3
st1 {v4.8b}, [x0], x3
st1 {v5.8b}, [x0], x3
st1 {v6.8b}, [x0], x3
st1 {v7.8b}, [x0], x3
ret x10
endfunc
.endm
h264_qpel8_hv_lowpass_l2 put
h264_qpel8_hv_lowpass_l2 avg
.macro h264_qpel16_hv type
function \type\()_h264_qpel16_hv_lowpass_neon
mov x13, x30
bl \type\()_h264_qpel8_hv_lowpass_neon
sub x1, x1, x3, lsl #2
bl \type\()_h264_qpel8_hv_lowpass_neon
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #8
sub x0, x0, x2, lsl #4
add x0, x0, #8
bl \type\()_h264_qpel8_hv_lowpass_neon
sub x1, x1, x3, lsl #2
mov x30, x13
b \type\()_h264_qpel8_hv_lowpass_neon
endfunc
function \type\()_h264_qpel16_hv_lowpass_l2_neon
mov x13, x30
sub x2, x4, #256
bl \type\()_h264_qpel8_hv_lowpass_l2_neon
sub x1, x1, x3, lsl #2
bl \type\()_h264_qpel8_hv_lowpass_l2_neon
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #8
sub x0, x0, x3, lsl #4
add x0, x0, #8
bl \type\()_h264_qpel8_hv_lowpass_l2_neon
sub x1, x1, x3, lsl #2
mov x30, x13
b \type\()_h264_qpel8_hv_lowpass_l2_neon
endfunc
.endm
h264_qpel16_hv put
h264_qpel16_hv avg
.macro h264_qpel8 type
function ff_\type\()_h264_qpel8_mc10_neon, export=1
lowpass_const w3
mov x3, x1
sub x1, x1, #2
mov x12, #8
b \type\()_h264_qpel8_h_lowpass_l2_neon
endfunc
function ff_\type\()_h264_qpel8_mc20_neon, export=1
lowpass_const w3
sub x1, x1, #2
mov x3, x2
mov x12, #8
b \type\()_h264_qpel8_h_lowpass_neon
endfunc
function ff_\type\()_h264_qpel8_mc30_neon, export=1
lowpass_const w3
add x3, x1, #1
sub x1, x1, #2
mov x12, #8
b \type\()_h264_qpel8_h_lowpass_l2_neon
endfunc
function ff_\type\()_h264_qpel8_mc01_neon, export=1
mov x14, x30
mov x12, x1
\type\()_h264_qpel8_mc01:
lowpass_const w3
mov x3, x2
sub x1, x1, x2, lsl #1
bl \type\()_h264_qpel8_v_lowpass_l2_neon
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc11_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel8_mc11:
lowpass_const w3
mov x11, sp
sub sp, sp, #64
mov x0, sp
sub x1, x1, #2
mov x3, #8
mov x12, #8
bl put_h264_qpel8_h_lowpass_neon
mov x0, x8
mov x3, x2
mov x12, sp
sub x1, x9, x2, lsl #1
mov x2, #8
bl \type\()_h264_qpel8_v_lowpass_l2_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc21_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel8_mc21:
lowpass_const w3
mov x11, sp
sub sp, sp, #(8*8+16*12)
sub x1, x1, #2
mov x3, #8
mov x0, sp
mov x12, #8
bl put_h264_qpel8_h_lowpass_neon
mov x4, x0
mov x0, x8
sub x1, x9, x2, lsl #1
sub x1, x1, #2
mov x3, x2
sub x2, x4, #64
bl \type\()_h264_qpel8_hv_lowpass_l2_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc31_neon, export=1
add x1, x1, #1
mov x14, x30
mov x8, x0
mov x9, x1
sub x1, x1, #1
b \type\()_h264_qpel8_mc11
endfunc
function ff_\type\()_h264_qpel8_mc02_neon, export=1
mov x14, x30
lowpass_const w3
sub x1, x1, x2, lsl #1
mov x3, x2
bl \type\()_h264_qpel8_v_lowpass_neon
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc12_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel8_mc12:
lowpass_const w3
mov x11, sp
sub sp, sp, #(8*8+16*12)
sub x1, x1, x2, lsl #1
mov x3, x2
mov x2, #8
mov x0, sp
bl put_h264_qpel8_v_lowpass_neon
mov x4, x0
mov x0, x8
sub x1, x9, x3, lsl #1
sub x1, x1, #2
sub x2, x4, #64
bl \type\()_h264_qpel8_hv_lowpass_l2_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc22_neon, export=1
mov x14, x30
mov x11, sp
sub x1, x1, x2, lsl #1
sub x1, x1, #2
mov x3, x2
bl \type\()_h264_qpel8_hv_lowpass_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc32_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, #1
b \type\()_h264_qpel8_mc12
endfunc
function ff_\type\()_h264_qpel8_mc03_neon, export=1
mov x14, x30
add x12, x1, x2
b \type\()_h264_qpel8_mc01
endfunc
function ff_\type\()_h264_qpel8_mc13_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
b \type\()_h264_qpel8_mc11
endfunc
function ff_\type\()_h264_qpel8_mc23_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
b \type\()_h264_qpel8_mc21
endfunc
function ff_\type\()_h264_qpel8_mc33_neon, export=1
add x1, x1, #1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
sub x1, x1, #1
b \type\()_h264_qpel8_mc11
endfunc
.endm
h264_qpel8 put
h264_qpel8 avg
.macro h264_qpel16 type
function ff_\type\()_h264_qpel16_mc10_neon, export=1
lowpass_const w3
mov x3, x1
sub x1, x1, #2
b \type\()_h264_qpel16_h_lowpass_l2_neon
endfunc
function ff_\type\()_h264_qpel16_mc20_neon, export=1
lowpass_const w3
sub x1, x1, #2
mov x3, x2
b \type\()_h264_qpel16_h_lowpass_neon
endfunc
function ff_\type\()_h264_qpel16_mc30_neon, export=1
lowpass_const w3
add x3, x1, #1
sub x1, x1, #2
b \type\()_h264_qpel16_h_lowpass_l2_neon
endfunc
function ff_\type\()_h264_qpel16_mc01_neon, export=1
mov x14, x30
mov x12, x1
\type\()_h264_qpel16_mc01:
lowpass_const w3
mov x3, x2
sub x1, x1, x2, lsl #1
bl \type\()_h264_qpel16_v_lowpass_l2_neon
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc11_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel16_mc11:
lowpass_const w3
mov x11, sp
sub sp, sp, #256
mov x0, sp
sub x1, x1, #2
mov x3, #16
bl put_h264_qpel16_h_lowpass_neon
mov x0, x8
mov x3, x2
mov x12, sp
sub x1, x9, x2, lsl #1
mov x2, #16
bl \type\()_h264_qpel16_v_lowpass_l2_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc21_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel16_mc21:
lowpass_const w3
mov x11, sp
sub sp, sp, #(16*16+16*12)
sub x1, x1, #2
mov x0, sp
bl put_h264_qpel16_h_lowpass_neon_packed
mov x4, x0
mov x0, x8
sub x1, x9, x2, lsl #1
sub x1, x1, #2
mov x3, x2
bl \type\()_h264_qpel16_hv_lowpass_l2_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc31_neon, export=1
add x1, x1, #1
mov x14, x30
mov x8, x0
mov x9, x1
sub x1, x1, #1
b \type\()_h264_qpel16_mc11
endfunc
function ff_\type\()_h264_qpel16_mc02_neon, export=1
mov x14, x30
lowpass_const w3
sub x1, x1, x2, lsl #1
mov x3, x2
bl \type\()_h264_qpel16_v_lowpass_neon
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc12_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel16_mc12:
lowpass_const w3
mov x11, sp
sub sp, sp, #(16*16+16*12)
sub x1, x1, x2, lsl #1
mov x0, sp
mov x3, x2
bl put_h264_qpel16_v_lowpass_neon_packed
mov x4, x0
mov x0, x8
sub x1, x9, x3, lsl #1
sub x1, x1, #2
mov x2, x3
bl \type\()_h264_qpel16_hv_lowpass_l2_neon
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc22_neon, export=1
mov x14, x30
lowpass_const w3
mov x11, sp
sub x1, x1, x2, lsl #1
sub x1, x1, #2
mov x3, x2
bl \type\()_h264_qpel16_hv_lowpass_neon
mov sp, x11 // restore stack
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc32_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, #1
b \type\()_h264_qpel16_mc12
endfunc
function ff_\type\()_h264_qpel16_mc03_neon, export=1
mov x14, x30
add x12, x1, x2
b \type\()_h264_qpel16_mc01
endfunc
function ff_\type\()_h264_qpel16_mc13_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
b \type\()_h264_qpel16_mc11
endfunc
function ff_\type\()_h264_qpel16_mc23_neon, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
b \type\()_h264_qpel16_mc21
endfunc
function ff_\type\()_h264_qpel16_mc33_neon, export=1
add x1, x1, #1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
sub x1, x1, #1
b \type\()_h264_qpel16_mc11
endfunc
.endm
h264_qpel16 put
h264_qpel16 avg
lavc/aarch64: h264qpel, add 10-bit lowpass_8_10 based functions Benchmarks A53 A55 A72 A76 avg_h264_qpel_8_mc01_10_c: 936.5 924.0 656.0 504.7 avg_h264_qpel_8_mc01_10_neon: 234.7 202.0 120.7 63.2 avg_h264_qpel_8_mc02_10_c: 921.0 920.0 669.2 493.7 avg_h264_qpel_8_mc02_10_neon: 202.0 173.2 102.7 58.5 avg_h264_qpel_8_mc03_10_c: 936.5 924.0 656.0 509.5 avg_h264_qpel_8_mc03_10_neon: 236.2 203.7 120.0 63.2 avg_h264_qpel_8_mc10_10_c: 1441.0 1437.7 806.7 478.5 avg_h264_qpel_8_mc10_10_neon: 325.7 324.0 153.7 94.2 avg_h264_qpel_8_mc11_10_c: 2160.7 2148.2 1366.7 906.7 avg_h264_qpel_8_mc11_10_neon: 492.0 464.0 242.5 134.5 avg_h264_qpel_8_mc13_10_c: 2157.0 2138.2 1357.0 908.2 avg_h264_qpel_8_mc13_10_neon: 494.0 467.2 242.0 140.0 avg_h264_qpel_8_mc20_10_c: 1433.5 1410.0 785.2 486.0 avg_h264_qpel_8_mc20_10_neon: 293.7 289.7 138.0 91.5 avg_h264_qpel_8_mc30_10_c: 1458.5 1461.7 813.7 483.2 avg_h264_qpel_8_mc30_10_neon: 341.7 339.2 154.0 95.2 avg_h264_qpel_8_mc31_10_c: 2194.7 2197.2 1358.7 928.0 avg_h264_qpel_8_mc31_10_neon: 520.0 495.0 245.5 142.5 avg_h264_qpel_8_mc33_10_c: 2188.0 2205.5 1356.7 910.7 avg_h264_qpel_8_mc33_10_neon: 521.0 494.5 245.7 145.7 avg_h264_qpel_16_mc01_10_c: 3717.2 3595.0 2610.0 2012.0 avg_h264_qpel_16_mc01_10_neon: 920.5 791.5 483.2 240.5 avg_h264_qpel_16_mc02_10_c: 3684.0 3633.0 2659.0 1919.7 avg_h264_qpel_16_mc02_10_neon: 790.7 678.2 409.2 217.0 avg_h264_qpel_16_mc03_10_c: 3726.5 3596.0 2606.7 2010.0 avg_h264_qpel_16_mc03_10_neon: 922.0 792.5 483.2 239.7 avg_h264_qpel_16_mc10_10_c: 5912.0 5803.2 3241.5 1916.7 avg_h264_qpel_16_mc10_10_neon: 1267.5 1277.2 616.5 365.0 avg_h264_qpel_16_mc11_10_c: 8599.2 8482.5 5338.0 3616.2 avg_h264_qpel_16_mc11_10_neon: 1913.0 1827.0 956.2 542.2 avg_h264_qpel_16_mc13_10_c: 8643.7 8488.5 5388.0 3628.5 avg_h264_qpel_16_mc13_10_neon: 1914.7 1828.7 969.2 530.5 avg_h264_qpel_16_mc20_10_c: 5719.5 5641.0 3147.0 1946.2 avg_h264_qpel_16_mc20_10_neon: 1139.5 1150.0 539.5 344.0 avg_h264_qpel_16_mc30_10_c: 5930.0 5872.5 3267.5 1918.0 avg_h264_qpel_16_mc30_10_neon: 1331.5 1341.2 616.5 369.5 avg_h264_qpel_16_mc31_10_c: 8758.7 8697.7 5353.0 3630.7 avg_h264_qpel_16_mc31_10_neon: 2018.7 1941.7 982.2 574.7 avg_h264_qpel_16_mc33_10_c: 8683.2 8675.2 5339.2 3634.7 avg_h264_qpel_16_mc33_10_neon: 2019.7 1940.2 994.5 566.0 put_h264_qpel_8_mc01_10_c: 854.2 843.0 599.2 478.0 put_h264_qpel_8_mc01_10_neon: 192.7 168.0 101.7 56.7 put_h264_qpel_8_mc02_10_c: 766.5 760.0 550.2 441.0 put_h264_qpel_8_mc02_10_neon: 160.0 139.2 88.7 53.0 put_h264_qpel_8_mc03_10_c: 854.2 843.0 599.2 479.0 put_h264_qpel_8_mc03_10_neon: 194.2 169.7 102.0 56.2 put_h264_qpel_8_mc10_10_c: 1352.7 1353.7 749.7 446.7 put_h264_qpel_8_mc10_10_neon: 289.7 294.2 135.5 88.5 put_h264_qpel_8_mc11_10_c: 2080.0 2066.2 1309.5 876.7 put_h264_qpel_8_mc11_10_neon: 450.0 429.7 229.7 131.2 put_h264_qpel_8_mc13_10_c: 2074.7 2060.2 1294.5 870.5 put_h264_qpel_8_mc13_10_neon: 452.5 434.5 226.5 130.0 put_h264_qpel_8_mc20_10_c: 1221.5 1216.0 684.5 399.7 put_h264_qpel_8_mc20_10_neon: 257.7 262.5 121.2 78.7 put_h264_qpel_8_mc30_10_c: 1379.0 1374.7 757.2 449.5 put_h264_qpel_8_mc30_10_neon: 305.7 310.2 135.5 86.5 put_h264_qpel_8_mc31_10_c: 2109.2 2119.7 1299.5 878.0 put_h264_qpel_8_mc31_10_neon: 478.0 458.5 226.0 137.2 put_h264_qpel_8_mc33_10_c: 2101.5 2115.2 1306.5 887.0 put_h264_qpel_8_mc33_10_neon: 479.0 458.7 229.7 141.7 put_h264_qpel_16_mc01_10_c: 3485.7 3396.7 2460.5 1914.5 put_h264_qpel_16_mc01_10_neon: 752.5 665.5 397.0 213.2 put_h264_qpel_16_mc02_10_c: 3103.5 3023.2 2154.7 1720.7 put_h264_qpel_16_mc02_10_neon: 622.7 551.2 347.7 196.2 put_h264_qpel_16_mc03_10_c: 3486.2 3394.0 2436.5 1917.7 put_h264_qpel_16_mc03_10_neon: 754.0 666.5 397.0 215.7 put_h264_qpel_16_mc10_10_c: 5533.0 5488.5 2989.0 1783.0 put_h264_qpel_16_mc10_10_neon: 1123.5 1165.2 535.2 334.7 put_h264_qpel_16_mc11_10_c: 8437.7 8281.2 5209.0 3510.7 put_h264_qpel_16_mc11_10_neon: 1745.0 1697.0 878.5 513.5 put_h264_qpel_16_mc13_10_c: 8567.7 8468.0 5221.5 3528.0 put_h264_qpel_16_mc13_10_neon: 1751.7 1698.2 889.2 507.0 put_h264_qpel_16_mc20_10_c: 4907.5 4885.0 2786.2 1607.5 put_h264_qpel_16_mc20_10_neon: 995.5 1034.5 475.5 307.0 put_h264_qpel_16_mc30_10_c: 5579.7 5537.7 3045.2 1789.5 put_h264_qpel_16_mc30_10_neon: 1187.5 1231.2 532.5 334.5 put_h264_qpel_16_mc31_10_c: 8677.2 8672.5 5204.2 3516.0 put_h264_qpel_16_mc31_10_neon: 1850.7 1813.2 893.0 545.2 put_h264_qpel_16_mc33_10_c: 8688.7 8671.2 5223.2 3512.0 put_h264_qpel_16_mc33_10_neon: 1851.7 1814.2 908.5 535.2 Signed-off-by: Mikhail Nitenko <mnitenko@gmail.com> Signed-off-by: Martin Storsjö <martin@martin.st>
2023-12-04 12:00:35 +02:00
//trashes v0-v5
.macro lowpass_8_10 r0, r1, r2, r3, d0, d1
ext v2.16b, \r0\().16b, \r1\().16b, #4
ext v3.16b, \r0\().16b, \r1\().16b, #6
add v2.8h, v2.8h, v3.8h
ext v4.16b, \r0\().16b, \r1\().16b, #2
ext v5.16b, \r0\().16b, \r1\().16b, #8
add v4.8h, v4.8h, v5.8h
ext v1.16b, \r0\().16b, \r1\().16b, #10
add \d0\().8h, \r0\().8h, v1.8h
ext v0.16b, \r2\().16b, \r3\().16b, #4
mla \d0\().8h, v2.8h, v6.h[1]
ext v1.16b, \r2\().16b, \r3\().16b, #6
add v0.8h, v0.8h, v1.8h
ext v1.16b, \r2\().16b, \r3\().16b, #2
mul v5.8h, v4.8h, v6.h[0]
uqsub \d0\().8h, \d0\().8h, v5.8h
urshr \d0\().8h, \d0\().8h, #5
ext v3.16b, \r2\().16b, \r3\().16b, #8
add v1.8h, v1.8h, v3.8h
ext v2.16b, \r2\().16b, \r3\().16b, #10
add \d1\().8h, \r2\().8h, v2.8h
mla \d1\().8h, v0.8h, v6.h[1]
mul v5.8h, v1.8h, v6.h[0]
uqsub \d1\().8h, \d1\().8h, v5.8h
mvni v5.8h, #0xFC, lsl #8 // 1023 for clipping
urshr \d1\().8h, \d1\().8h, #5
umin \d0\().8h, \d0\().8h, v5.8h
umin \d1\().8h, \d1\().8h, v5.8h
.endm
//trashes v0-v4
.macro lowpass_8_10_v r0, r1, r2, r3, r4, r5, r6, d0, d1
add v2.8h, \r2\().8h, \r3\().8h
add v0.8h, \r3\().8h, \r4\().8h
add v4.8h, \r1\().8h, \r4\().8h
add v1.8h, \r2\().8h, \r5\().8h
add \d0\().8h, \r0\().8h, \r5\().8h
add \d1\().8h, \r1\().8h, \r6\().8h
mla \d0\().8h, v2.8h, v6.h[1]
mla \d1\().8h, v0.8h, v6.h[1]
mul v2.8h, v4.8h, v6.h[0]
mul v0.8h, v1.8h, v6.h[0]
uqsub \d0\().8h, \d0\().8h, v2.8h
uqsub \d1\().8h, \d1\().8h, v0.8h
mvni v0.8h, #0xFC, lsl #8 // 1023 for clipping
urshr \d0\().8h, \d0\().8h, #5
urshr \d1\().8h, \d1\().8h, #5
umin \d0\().8h, \d0\().8h, v0.8h
umin \d1\().8h, \d1\().8h, v0.8h
.endm
function put_h264_qpel16_h_lowpass_neon_packed_10
mov x4, x30
mov x12, #32
mov x3, #16
bl put_h264_qpel8_h_lowpass_neon_10
sub x1, x1, x2, lsl #4
add x1, x1, #16
mov x12, #32
mov x30, x4
b put_h264_qpel8_h_lowpass_neon_10
endfunc
.macro h264_qpel_h_lowpass_10 type
function \type\()_h264_qpel16_h_lowpass_neon_10
mov x13, x30
mov x12, #32
bl \type\()_h264_qpel8_h_lowpass_neon_10
sub x0, x0, x3, lsl #4
sub x1, x1, x2, lsl #4
add x0, x0, #16
add x1, x1, #16
mov x12, #32
mov x30, x13
endfunc
function \type\()_h264_qpel8_h_lowpass_neon_10
1: ld1 {v28.8h, v29.8h}, [x1], x2
ld1 {v16.8h, v17.8h}, [x1], x2
subs x12, x12, #4
lowpass_8_10 v28, v29, v16, v17, v28, v20
.ifc \type,avg
ld1 {v2.8h}, [x0], x3
ld1 {v3.8h}, [x0]
urhadd v28.8h, v28.8h, v2.8h
urhadd v20.8h, v20.8h, v3.8h
sub x0, x0, x3
.endif
st1 {v28.8h}, [x0], x3
st1 {v20.8h}, [x0], x3
b.ne 1b
ret
endfunc
.endm
h264_qpel_h_lowpass_10 put
h264_qpel_h_lowpass_10 avg
.macro h264_qpel_h_lowpass_l2_10 type
function \type\()_h264_qpel16_h_lowpass_l2_neon_10
mov x13, x30
mov x12, #32
bl \type\()_h264_qpel8_h_lowpass_l2_neon_10
sub x0, x0, x2, lsl #4
sub x1, x1, x2, lsl #4
sub x3, x3, x2, lsl #4
add x0, x0, #16
add x1, x1, #16
add x3, x3, #16
mov x12, #32
mov x30, x13
endfunc
function \type\()_h264_qpel8_h_lowpass_l2_neon_10
1: ld1 {v26.8h, v27.8h}, [x1], x2
ld1 {v16.8h, v17.8h}, [x1], x2
ld1 {v28.8h}, [x3], x2
ld1 {v29.8h}, [x3], x2
subs x12, x12, #4
lowpass_8_10 v26, v27, v16, v17, v26, v27
urhadd v26.8h, v26.8h, v28.8h
urhadd v27.8h, v27.8h, v29.8h
.ifc \type,avg
ld1 {v2.8h}, [x0], x2
ld1 {v3.8h}, [x0]
urhadd v26.8h, v26.8h, v2.8h
urhadd v27.8h, v27.8h, v3.8h
sub x0, x0, x2
.endif
st1 {v26.8h}, [x0], x2
st1 {v27.8h}, [x0], x2
b.ne 1b
ret
endfunc
.endm
h264_qpel_h_lowpass_l2_10 put
h264_qpel_h_lowpass_l2_10 avg
function put_h264_qpel16_v_lowpass_neon_packed_10
mov x4, x30
mov x2, #8
bl put_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #2
bl put_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #8
bl put_h264_qpel8_v_lowpass_neon
sub x1, x1, x3, lsl #2
mov x30, x4
b put_h264_qpel8_v_lowpass_neon
endfunc
.macro h264_qpel_v_lowpass_10 type
function \type\()_h264_qpel16_v_lowpass_neon_10
mov x4, x30
bl \type\()_h264_qpel8_v_lowpass_neon_10
sub x1, x1, x3, lsl #2
bl \type\()_h264_qpel8_v_lowpass_neon_10
sub x0, x0, x2, lsl #4
add x0, x0, #16
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #16
bl \type\()_h264_qpel8_v_lowpass_neon_10
sub x1, x1, x3, lsl #2
mov x30, x4
endfunc
function \type\()_h264_qpel8_v_lowpass_neon_10
ld1 {v16.8h}, [x1], x3
ld1 {v17.8h}, [x1], x3
ld1 {v18.8h}, [x1], x3
ld1 {v19.8h}, [x1], x3
ld1 {v20.8h}, [x1], x3
ld1 {v21.8h}, [x1], x3
ld1 {v22.8h}, [x1], x3
ld1 {v23.8h}, [x1], x3
ld1 {v24.8h}, [x1], x3
ld1 {v25.8h}, [x1], x3
ld1 {v26.8h}, [x1], x3
ld1 {v27.8h}, [x1], x3
ld1 {v28.8h}, [x1]
lowpass_8_10_v v16, v17, v18, v19, v20, v21, v22, v16, v17
lowpass_8_10_v v18, v19, v20, v21, v22, v23, v24, v18, v19
lowpass_8_10_v v20, v21, v22, v23, v24, v25, v26, v20, v21
lowpass_8_10_v v22, v23, v24, v25, v26, v27, v28, v22, v23
.ifc \type,avg
ld1 {v24.8h}, [x0], x2
ld1 {v25.8h}, [x0], x2
ld1 {v26.8h}, [x0], x2
urhadd v16.8h, v16.8h, v24.8h
ld1 {v27.8h}, [x0], x2
urhadd v17.8h, v17.8h, v25.8h
ld1 {v28.8h}, [x0], x2
urhadd v18.8h, v18.8h, v26.8h
ld1 {v29.8h}, [x0], x2
urhadd v19.8h, v19.8h, v27.8h
ld1 {v30.8h}, [x0], x2
urhadd v20.8h, v20.8h, v28.8h
ld1 {v31.8h}, [x0], x2
urhadd v21.8h, v21.8h, v29.8h
urhadd v22.8h, v22.8h, v30.8h
urhadd v23.8h, v23.8h, v31.8h
sub x0, x0, x2, lsl #3
.endif
st1 {v16.8h}, [x0], x2
st1 {v17.8h}, [x0], x2
st1 {v18.8h}, [x0], x2
st1 {v19.8h}, [x0], x2
st1 {v20.8h}, [x0], x2
st1 {v21.8h}, [x0], x2
st1 {v22.8h}, [x0], x2
st1 {v23.8h}, [x0], x2
ret
endfunc
.endm
h264_qpel_v_lowpass_10 put
h264_qpel_v_lowpass_10 avg
.macro h264_qpel_v_lowpass_l2_10 type
function \type\()_h264_qpel16_v_lowpass_l2_neon_10
mov x4, x30
bl \type\()_h264_qpel8_v_lowpass_l2_neon_10
sub x1, x1, x3, lsl #2
bl \type\()_h264_qpel8_v_lowpass_l2_neon_10
sub x0, x0, x3, lsl #4
sub x12, x12, x2, lsl #4
add x0, x0, #16
add x12, x12, #16
sub x1, x1, x3, lsl #4
sub x1, x1, x3, lsl #2
add x1, x1, #16
bl \type\()_h264_qpel8_v_lowpass_l2_neon_10
sub x1, x1, x3, lsl #2
mov x30, x4
endfunc
function \type\()_h264_qpel8_v_lowpass_l2_neon_10
ld1 {v16.8h}, [x1], x3
ld1 {v17.8h}, [x1], x3
ld1 {v18.8h}, [x1], x3
ld1 {v19.8h}, [x1], x3
ld1 {v20.8h}, [x1], x3
ld1 {v21.8h}, [x1], x3
ld1 {v22.8h}, [x1], x3
ld1 {v23.8h}, [x1], x3
ld1 {v24.8h}, [x1], x3
ld1 {v25.8h}, [x1], x3
ld1 {v26.8h}, [x1], x3
ld1 {v27.8h}, [x1], x3
ld1 {v28.8h}, [x1]
lowpass_8_10_v v16, v17, v18, v19, v20, v21, v22, v16, v17
lowpass_8_10_v v18, v19, v20, v21, v22, v23, v24, v18, v19
lowpass_8_10_v v20, v21, v22, v23, v24, v25, v26, v20, v21
lowpass_8_10_v v22, v23, v24, v25, v26, v27, v28, v22, v23
ld1 {v24.8h}, [x12], x2
ld1 {v25.8h}, [x12], x2
ld1 {v26.8h}, [x12], x2
ld1 {v27.8h}, [x12], x2
ld1 {v28.8h}, [x12], x2
urhadd v16.8h, v24.8h, v16.8h
urhadd v17.8h, v25.8h, v17.8h
ld1 {v29.8h}, [x12], x2
urhadd v18.8h, v26.8h, v18.8h
urhadd v19.8h, v27.8h, v19.8h
ld1 {v30.8h}, [x12], x2
urhadd v20.8h, v28.8h, v20.8h
urhadd v21.8h, v29.8h, v21.8h
ld1 {v31.8h}, [x12], x2
urhadd v22.8h, v30.8h, v22.8h
urhadd v23.8h, v31.8h, v23.8h
.ifc \type,avg
ld1 {v24.8h}, [x0], x3
ld1 {v25.8h}, [x0], x3
ld1 {v26.8h}, [x0], x3
urhadd v16.8h, v16.8h, v24.8h
ld1 {v27.8h}, [x0], x3
urhadd v17.8h, v17.8h, v25.8h
ld1 {v28.8h}, [x0], x3
urhadd v18.8h, v18.8h, v26.8h
ld1 {v29.8h}, [x0], x3
urhadd v19.8h, v19.8h, v27.8h
ld1 {v30.8h}, [x0], x3
urhadd v20.8h, v20.8h, v28.8h
ld1 {v31.8h}, [x0], x3
urhadd v21.8h, v21.8h, v29.8h
urhadd v22.8h, v22.8h, v30.8h
urhadd v23.8h, v23.8h, v31.8h
sub x0, x0, x3, lsl #3
.endif
st1 {v16.8h}, [x0], x3
st1 {v17.8h}, [x0], x3
st1 {v18.8h}, [x0], x3
st1 {v19.8h}, [x0], x3
st1 {v20.8h}, [x0], x3
st1 {v21.8h}, [x0], x3
st1 {v22.8h}, [x0], x3
st1 {v23.8h}, [x0], x3
ret
endfunc
.endm
h264_qpel_v_lowpass_l2_10 put
h264_qpel_v_lowpass_l2_10 avg
.macro h264_qpel8_10 type
function ff_\type\()_h264_qpel8_mc10_neon_10, export=1
lowpass_const w3
mov x3, x1
sub x1, x1, #4
mov x12, #16
b \type\()_h264_qpel8_h_lowpass_l2_neon_10
endfunc
function ff_\type\()_h264_qpel8_mc20_neon_10, export=1
lowpass_const w3
sub x1, x1, #4
mov x3, x2
mov x12, #16
b \type\()_h264_qpel8_h_lowpass_neon_10
endfunc
function ff_\type\()_h264_qpel8_mc30_neon_10, export=1
lowpass_const w3
add x3, x1, #2
sub x1, x1, #4
mov x12, #16
b \type\()_h264_qpel8_h_lowpass_l2_neon_10
endfunc
function ff_\type\()_h264_qpel8_mc01_neon_10, export=1
mov x14, x30
mov x12, x1
\type\()_h264_qpel8_mc01_10:
lowpass_const w3
mov x3, x2
sub x1, x1, x2, lsl #1
bl \type\()_h264_qpel8_v_lowpass_l2_neon_10
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc11_neon_10, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel8_mc11_10:
lowpass_const w3
mov x11, sp
sub sp, sp, #128
mov x0, sp
sub x1, x1, #4
mov x3, #16
mov x12, #16
bl put_h264_qpel8_h_lowpass_neon_10
mov x0, x8
mov x3, x2
mov x12, sp
sub x1, x9, x2, lsl #1
mov x2, #16
bl \type\()_h264_qpel8_v_lowpass_l2_neon_10
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc31_neon_10, export=1
add x1, x1, #2
mov x14, x30
mov x8, x0
mov x9, x1
sub x1, x1, #2
b \type\()_h264_qpel8_mc11_10
endfunc
function ff_\type\()_h264_qpel8_mc02_neon_10, export=1
mov x14, x30
lowpass_const w3
sub x1, x1, x2, lsl #1
mov x3, x2
bl \type\()_h264_qpel8_v_lowpass_neon_10
ret x14
endfunc
function ff_\type\()_h264_qpel8_mc03_neon_10, export=1
mov x14, x30
add x12, x1, x2
b \type\()_h264_qpel8_mc01_10
endfunc
function ff_\type\()_h264_qpel8_mc13_neon_10, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
b \type\()_h264_qpel8_mc11_10
endfunc
function ff_\type\()_h264_qpel8_mc33_neon_10, export=1
add x1, x1, #2
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
sub x1, x1, #2
b \type\()_h264_qpel8_mc11_10
endfunc
.endm
h264_qpel8_10 put
h264_qpel8_10 avg
.macro h264_qpel16_10 type
function ff_\type\()_h264_qpel16_mc10_neon_10, export=1
lowpass_const w3
mov x3, x1
sub x1, x1, #4
b \type\()_h264_qpel16_h_lowpass_l2_neon_10
endfunc
function ff_\type\()_h264_qpel16_mc20_neon_10, export=1
lowpass_const w3
sub x1, x1, #4
mov x3, x2
b \type\()_h264_qpel16_h_lowpass_neon_10
endfunc
function ff_\type\()_h264_qpel16_mc30_neon_10, export=1
lowpass_const w3
add x3, x1, #2
sub x1, x1, #4
b \type\()_h264_qpel16_h_lowpass_l2_neon_10
endfunc
function ff_\type\()_h264_qpel16_mc01_neon_10, export=1
mov x14, x30
mov x12, x1
\type\()_h264_qpel16_mc01_10:
lowpass_const w3
mov x3, x2
sub x1, x1, x2, lsl #1
bl \type\()_h264_qpel16_v_lowpass_l2_neon_10
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc11_neon_10, export=1
mov x14, x30
mov x8, x0
mov x9, x1
\type\()_h264_qpel16_mc11_10:
lowpass_const w3
mov x11, sp
sub sp, sp, #512
mov x0, sp
sub x1, x1, #4
mov x3, #32
bl put_h264_qpel16_h_lowpass_neon_10
mov x0, x8
mov x3, x2
mov x12, sp
sub x1, x9, x2, lsl #1
mov x2, #32
bl \type\()_h264_qpel16_v_lowpass_l2_neon_10
mov sp, x11
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc31_neon_10, export=1
add x1, x1, #2
mov x14, x30
mov x8, x0
mov x9, x1
sub x1, x1, #2
b \type\()_h264_qpel16_mc11_10
endfunc
function ff_\type\()_h264_qpel16_mc02_neon_10, export=1
mov x14, x30
lowpass_const w3
sub x1, x1, x2, lsl #1
mov x3, x2
bl \type\()_h264_qpel16_v_lowpass_neon_10
ret x14
endfunc
function ff_\type\()_h264_qpel16_mc03_neon_10, export=1
mov x14, x30
add x12, x1, x2
b \type\()_h264_qpel16_mc01_10
endfunc
function ff_\type\()_h264_qpel16_mc13_neon_10, export=1
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
b \type\()_h264_qpel16_mc11_10
endfunc
function ff_\type\()_h264_qpel16_mc33_neon_10, export=1
add x1, x1, #2
mov x14, x30
mov x8, x0
mov x9, x1
add x1, x1, x2
sub x1, x1, #2
b \type\()_h264_qpel16_mc11_10
endfunc
.endm
h264_qpel16_10 put
h264_qpel16_10 avg