ffmpeg/libavcodec/aarch64/h264qpel_neon.S
Martin Storsjö fd3bd5c492 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>
2021-10-18 14:27:58 +03:00

936 lines
33 KiB
ArmAsm

/*
* 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
//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]
.if \narrow
sqrshrun \d0\().8B, \d0\().8H, #5
sqrshrun \d1\().8B, \d1\().8H, #5
.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
.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]
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]
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
lowpass_8.16 v16, v17, v18, v19, v20, v21
lowpass_8.16 v17, v18, v19, v20, v21, v22
lowpass_8.16 v18, v19, v20, v21, v22, v23
lowpass_8.16 v19, v20, v21, v22, v23, v24
lowpass_8.16 v20, v21, v22, v23, v24, v25
lowpass_8.16 v21, v22, v23, v24, v25, v26
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