ffmpeg/libavcodec/aarch64/vvc/inter.S

621 lines
21 KiB
ArmAsm

/*
* Copyright (c) 2024 Zhao Zhili <quinkblack@foxmail.com>
*
* 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"
#define VVC_MAX_PB_SIZE 128
.macro vvc_avg type, bit_depth
.macro vvc_\type\()_\bit_depth\()_2_4 tap
.if \tap == 2
ldr s0, [src0]
ldr s2, [src1]
.else
ldr d0, [src0]
ldr d2, [src1]
.endif
.ifc \type, avg
saddl v4.4s, v0.4h, v2.4h
add v4.4s, v4.4s, v16.4s
sqshrn v4.4h, v4.4s, #(15 - \bit_depth)
.else
mov v4.16b, v16.16b
smlal v4.4s, v0.4h, v19.4h
smlal v4.4s, v2.4h, v20.4h
sqshl v4.4s, v4.4s, v22.4s
sqxtn v4.4h, v4.4s
.endif
.if \bit_depth == 8
sqxtun v4.8b, v4.8h
.if \tap == 2
str h4, [dst]
.else // tap == 4
str s4, [dst]
.endif
.else // bit_depth > 8
smin v4.4h, v4.4h, v17.4h
smax v4.4h, v4.4h, v18.4h
.if \tap == 2
str s4, [dst]
.else
str d4, [dst]
.endif
.endif
add src0, src0, x10
add src1, src1, x10
add dst, dst, dst_stride
.endm
function ff_vvc_\type\()_\bit_depth\()_neon, export=1
dst .req x0
dst_stride .req x1
src0 .req x2
src1 .req x3
width .req w4
height .req w5
mov x10, #(VVC_MAX_PB_SIZE * 2)
cmp width, #8
.ifc \type, avg
movi v16.4s, #(1 << (14 - \bit_depth))
.else
lsr x11, x6, #32 // weight0
mov w12, w6 // weight1
lsr x13, x7, #32 // offset
mov w14, w7 // shift
dup v19.8h, w11
neg w14, w14 // so we can use sqshl
dup v20.8h, w12
dup v16.4s, w13
dup v22.4s, w14
.endif // avg
.if \bit_depth >= 10
// clip pixel
mov w6, #((1 << \bit_depth) - 1)
movi v18.8h, #0
dup v17.8h, w6
.endif
b.eq 8f
b.hi 16f
cmp width, #4
b.eq 4f
2: // width == 2
subs height, height, #1
vvc_\type\()_\bit_depth\()_2_4 2
b.ne 2b
b 32f
4: // width == 4
subs height, height, #1
vvc_\type\()_\bit_depth\()_2_4 4
b.ne 4b
b 32f
8: // width == 8
ld1 {v0.8h}, [src0], x10
ld1 {v2.8h}, [src1], x10
.ifc \type, avg
saddl v4.4s, v0.4h, v2.4h
saddl2 v5.4s, v0.8h, v2.8h
add v4.4s, v4.4s, v16.4s
add v5.4s, v5.4s, v16.4s
sqshrn v4.4h, v4.4s, #(15 - \bit_depth)
sqshrn2 v4.8h, v5.4s, #(15 - \bit_depth)
.else
mov v4.16b, v16.16b
mov v5.16b, v16.16b
smlal v4.4s, v0.4h, v19.4h
smlal v4.4s, v2.4h, v20.4h
smlal2 v5.4s, v0.8h, v19.8h
smlal2 v5.4s, v2.8h, v20.8h
sqshl v4.4s, v4.4s, v22.4s
sqshl v5.4s, v5.4s, v22.4s
sqxtn v4.4h, v4.4s
sqxtn2 v4.8h, v5.4s
.endif
subs height, height, #1
.if \bit_depth == 8
sqxtun v4.8b, v4.8h
st1 {v4.8b}, [dst], dst_stride
.else
smin v4.8h, v4.8h, v17.8h
smax v4.8h, v4.8h, v18.8h
st1 {v4.8h}, [dst], dst_stride
.endif
b.ne 8b
b 32f
16: // width >= 16
mov w6, width
mov x7, src0
mov x8, src1
mov x9, dst
17:
ldp q0, q1, [x7], #32
ldp q2, q3, [x8], #32
.ifc \type, avg
saddl v4.4s, v0.4h, v2.4h
saddl2 v5.4s, v0.8h, v2.8h
saddl v6.4s, v1.4h, v3.4h
saddl2 v7.4s, v1.8h, v3.8h
add v4.4s, v4.4s, v16.4s
add v5.4s, v5.4s, v16.4s
add v6.4s, v6.4s, v16.4s
add v7.4s, v7.4s, v16.4s
sqshrn v4.4h, v4.4s, #(15 - \bit_depth)
sqshrn2 v4.8h, v5.4s, #(15 - \bit_depth)
sqshrn v6.4h, v6.4s, #(15 - \bit_depth)
sqshrn2 v6.8h, v7.4s, #(15 - \bit_depth)
.else // avg
mov v4.16b, v16.16b
mov v5.16b, v16.16b
mov v6.16b, v16.16b
mov v7.16b, v16.16b
smlal v4.4s, v0.4h, v19.4h
smlal v4.4s, v2.4h, v20.4h
smlal2 v5.4s, v0.8h, v19.8h
smlal2 v5.4s, v2.8h, v20.8h
smlal v6.4s, v1.4h, v19.4h
smlal v6.4s, v3.4h, v20.4h
smlal2 v7.4s, v1.8h, v19.8h
smlal2 v7.4s, v3.8h, v20.8h
sqshl v4.4s, v4.4s, v22.4s
sqshl v5.4s, v5.4s, v22.4s
sqshl v6.4s, v6.4s, v22.4s
sqshl v7.4s, v7.4s, v22.4s
sqxtn v4.4h, v4.4s
sqxtn v6.4h, v6.4s
sqxtn2 v4.8h, v5.4s
sqxtn2 v6.8h, v7.4s
.endif // w_avg
subs w6, w6, #16
.if \bit_depth == 8
sqxtun v4.8b, v4.8h
sqxtun2 v4.16b, v6.8h
str q4, [x9], #16
.else
smin v4.8h, v4.8h, v17.8h
smin v6.8h, v6.8h, v17.8h
smax v4.8h, v4.8h, v18.8h
smax v6.8h, v6.8h, v18.8h
stp q4, q6, [x9], #32
.endif
b.ne 17b
subs height, height, #1
add src0, src0, x10
add src1, src1, x10
add dst, dst, dst_stride
b.ne 16b
32:
ret
.unreq dst
.unreq dst_stride
.unreq src0
.unreq src1
.unreq width
.unreq height
endfunc
.endm
vvc_avg avg, 8
vvc_avg avg, 10
vvc_avg avg, 12
vvc_avg w_avg, 8
vvc_avg w_avg, 10
vvc_avg w_avg, 12
/* x0: int16_t *dst
* x1: const uint8_t *_src
* x2: ptrdiff_t _src_stride
* w3: int height
* x4: intptr_t mx
* x5: intptr_t my
* w6: int width
*/
function ff_vvc_dmvr_8_neon, export=1
dst .req x0
src .req x1
src_stride .req x2
height .req w3
mx .req x4
my .req x5
width .req w6
sxtw x6, w6
mov x7, #(VVC_MAX_PB_SIZE * 2 + 8)
cmp width, #16
sub src_stride, src_stride, x6
cset w15, gt // width > 16
movi v16.8h, #2 // DMVR_SHIFT
sub x7, x7, x6, lsl #1
1:
cbz w15, 2f
ldr q0, [src], #16
uxtl v1.8h, v0.8b
uxtl2 v2.8h, v0.16b
ushl v1.8h, v1.8h, v16.8h
ushl v2.8h, v2.8h, v16.8h
stp q1, q2, [dst], #32
b 3f
2:
ldr d0, [src], #8
uxtl v1.8h, v0.8b
ushl v1.8h, v1.8h, v16.8h
str q1, [dst], #16
3:
subs height, height, #1
ldr s3, [src], #4
uxtl v4.8h, v3.8b
ushl v4.4h, v4.4h, v16.4h
st1 {v4.4h}, [dst], x7
add src, src, src_stride
b.ne 1b
ret
endfunc
function ff_vvc_dmvr_12_neon, export=1
sxtw x6, w6
mov x7, #(VVC_MAX_PB_SIZE * 2 + 8)
cmp width, #16
sub src_stride, src_stride, x6, lsl #1
cset w15, gt // width > 16
movi v16.8h, #2 // offset4
sub x7, x7, x6, lsl #1
1:
cbz w15, 2f
ldp q0, q1, [src], #32
uaddl v2.4s, v0.4h, v16.4h
uaddl2 v3.4s, v0.8h, v16.8h
uaddl v4.4s, v1.4h, v16.4h
uaddl2 v5.4s, v1.8h, v16.8h
ushr v2.4s, v2.4s, #2
ushr v3.4s, v3.4s, #2
ushr v4.4s, v4.4s, #2
ushr v5.4s, v5.4s, #2
uqxtn v2.4h, v2.4s
uqxtn2 v2.8h, v3.4s
uqxtn v4.4h, v4.4s
uqxtn2 v4.8h, v5.4s
stp q2, q4, [dst], #32
b 3f
2:
ldr q0, [src], #16
uaddl v2.4s, v0.4h, v16.4h
uaddl2 v3.4s, v0.8h, v16.8h
ushr v2.4s, v2.4s, #2
ushr v3.4s, v3.4s, #2
uqxtn v2.4h, v2.4s
uqxtn2 v2.8h, v3.4s
str q2, [dst], #16
3:
subs height, height, #1
ldr d0, [src], #8
uaddl v3.4s, v0.4h, v16.4h
ushr v3.4s, v3.4s, #2
uqxtn v3.4h, v3.4s
st1 {v3.4h}, [dst], x7
add src, src, src_stride
b.ne 1b
ret
endfunc
function ff_vvc_dmvr_hv_8_neon, export=1
tmp0 .req x7
tmp1 .req x8
sub sp, sp, #(VVC_MAX_PB_SIZE * 4)
movrel x9, X(ff_vvc_inter_luma_dmvr_filters)
add x12, x9, mx, lsl #1
ldrb w10, [x12]
ldrb w11, [x12, #1]
mov tmp0, sp
add tmp1, tmp0, #(VVC_MAX_PB_SIZE * 2)
// We know the value are positive
dup v0.8h, w10 // filter_x[0]
dup v1.8h, w11 // filter_x[1]
add x12, x9, my, lsl #1
ldrb w10, [x12]
ldrb w11, [x12, #1]
sxtw x6, w6
movi v30.8h, #(1 << (8 - 7)) // offset1
movi v31.8h, #8 // offset2
dup v2.8h, w10 // filter_y[0]
dup v3.8h, w11 // filter_y[1]
// Valid value for width can only be 8 + 4, 16 + 4
cmp width, #16
mov w10, #0 // start filter_y or not
add height, height, #1
sub dst, dst, #(VVC_MAX_PB_SIZE * 2)
sub src_stride, src_stride, x6
cset w15, gt // width > 16
1:
mov x12, tmp0
mov x13, tmp1
mov x14, dst
cbz w15, 2f
// width > 16
ldur q5, [src, #1]
ldr q4, [src], #16
uxtl v7.8h, v5.8b
uxtl2 v17.8h, v5.16b
uxtl v6.8h, v4.8b
uxtl2 v16.8h, v4.16b
mul v6.8h, v6.8h, v0.8h
mul v16.8h, v16.8h, v0.8h
mla v6.8h, v7.8h, v1.8h
mla v16.8h, v17.8h, v1.8h
add v6.8h, v6.8h, v30.8h
add v16.8h, v16.8h, v30.8h
ushr v6.8h, v6.8h, #(8 - 6)
ushr v7.8h, v16.8h, #(8 - 6)
stp q6, q7, [x13], #32
cbz w10, 3f
ldp q16, q17, [x12], #32
mul v16.8h, v16.8h, v2.8h
mul v17.8h, v17.8h, v2.8h
mla v16.8h, v6.8h, v3.8h
mla v17.8h, v7.8h, v3.8h
add v16.8h, v16.8h, v31.8h
add v17.8h, v17.8h, v31.8h
ushr v16.8h, v16.8h, #4
ushr v17.8h, v17.8h, #4
stp q16, q17, [x14], #32
b 3f
2:
// width > 8
ldur d5, [src, #1]
ldr d4, [src], #8
uxtl v7.8h, v5.8b
uxtl v6.8h, v4.8b
mul v6.8h, v6.8h, v0.8h
mla v6.8h, v7.8h, v1.8h
add v6.8h, v6.8h, v30.8h
ushr v6.8h, v6.8h, #(8 - 6)
str q6, [x13], #16
cbz w10, 3f
ldr q16, [x12], #16
mul v16.8h, v16.8h, v2.8h
mla v16.8h, v6.8h, v3.8h
add v16.8h, v16.8h, v31.8h
ushr v16.8h, v16.8h, #4
str q16, [x14], #16
3:
ldr s5, [src, #1]
ldr s4, [src], #4
uxtl v7.8h, v5.8b
uxtl v6.8h, v4.8b
mul v6.4h, v6.4h, v0.4h
mla v6.4h, v7.4h, v1.4h
add v6.4h, v6.4h, v30.4h
ushr v6.4h, v6.4h, #(8 - 6)
str d6, [x13], #8
cbz w10, 4f
ldr d16, [x12], #8
mul v16.4h, v16.4h, v2.4h
mla v16.4h, v6.4h, v3.4h
add v16.4h, v16.4h, v31.4h
ushr v16.4h, v16.4h, #4
str d16, [x14], #8
4:
subs height, height, #1
mov w10, #1
add src, src, src_stride
add dst, dst, #(VVC_MAX_PB_SIZE * 2)
eor tmp0, tmp0, tmp1
eor tmp1, tmp0, tmp1
eor tmp0, tmp0, tmp1
b.ne 1b
add sp, sp, #(VVC_MAX_PB_SIZE * 4)
ret
endfunc
function ff_vvc_dmvr_hv_12_neon, export=1
movi v29.4s, #(12 - 6)
movi v30.4s, #(1 << (12 - 7)) // offset1
b 0f
endfunc
function ff_vvc_dmvr_hv_10_neon, export=1
movi v29.4s, #(10 - 6)
movi v30.4s, #(1 << (10 - 7)) // offset1
0:
movi v31.4s, #8 // offset2
neg v29.4s, v29.4s
sub sp, sp, #(VVC_MAX_PB_SIZE * 4)
movrel x9, X(ff_vvc_inter_luma_dmvr_filters)
add x12, x9, mx, lsl #1
ldrb w10, [x12]
ldrb w11, [x12, #1]
mov tmp0, sp
add tmp1, tmp0, #(VVC_MAX_PB_SIZE * 2)
// We know the value are positive
dup v0.8h, w10 // filter_x[0]
dup v1.8h, w11 // filter_x[1]
add x12, x9, my, lsl #1
ldrb w10, [x12]
ldrb w11, [x12, #1]
sxtw x6, w6
dup v2.8h, w10 // filter_y[0]
dup v3.8h, w11 // filter_y[1]
// Valid value for width can only be 8 + 4, 16 + 4
cmp width, #16
mov w10, #0 // start filter_y or not
add height, height, #1
sub dst, dst, #(VVC_MAX_PB_SIZE * 2)
sub src_stride, src_stride, x6, lsl #1
cset w15, gt // width > 16
1:
mov x12, tmp0
mov x13, tmp1
mov x14, dst
cbz w15, 2f
// width > 16
add x16, src, #2
ldp q6, q16, [src], #32
ldp q7, q17, [x16]
umull v4.4s, v6.4h, v0.4h
umull2 v5.4s, v6.8h, v0.8h
umull v18.4s, v16.4h, v0.4h
umull2 v19.4s, v16.8h, v0.8h
umlal v4.4s, v7.4h, v1.4h
umlal2 v5.4s, v7.8h, v1.8h
umlal v18.4s, v17.4h, v1.4h
umlal2 v19.4s, v17.8h, v1.8h
add v4.4s, v4.4s, v30.4s
add v5.4s, v5.4s, v30.4s
add v18.4s, v18.4s, v30.4s
add v19.4s, v19.4s, v30.4s
ushl v4.4s, v4.4s, v29.4s
ushl v5.4s, v5.4s, v29.4s
ushl v18.4s, v18.4s, v29.4s
ushl v19.4s, v19.4s, v29.4s
uqxtn v6.4h, v4.4s
uqxtn2 v6.8h, v5.4s
uqxtn v7.4h, v18.4s
uqxtn2 v7.8h, v19.4s
stp q6, q7, [x13], #32
cbz w10, 3f
ldp q4, q5, [x12], #32
umull v17.4s, v4.4h, v2.4h
umull2 v18.4s, v4.8h, v2.8h
umull v19.4s, v5.4h, v2.4h
umull2 v20.4s, v5.8h, v2.8h
umlal v17.4s, v6.4h, v3.4h
umlal2 v18.4s, v6.8h, v3.8h
umlal v19.4s, v7.4h, v3.4h
umlal2 v20.4s, v7.8h, v3.8h
add v17.4s, v17.4s, v31.4s
add v18.4s, v18.4s, v31.4s
add v19.4s, v19.4s, v31.4s
add v20.4s, v20.4s, v31.4s
ushr v17.4s, v17.4s, #4
ushr v18.4s, v18.4s, #4
ushr v19.4s, v19.4s, #4
ushr v20.4s, v20.4s, #4
uqxtn v6.4h, v17.4s
uqxtn2 v6.8h, v18.4s
uqxtn v7.4h, v19.4s
uqxtn2 v7.8h, v20.4s
stp q6, q7, [x14], #32
b 3f
2:
// width > 8
ldur q7, [src, #2]
ldr q6, [src], #16
umull v4.4s, v6.4h, v0.4h
umull2 v5.4s, v6.8h, v0.8h
umlal v4.4s, v7.4h, v1.4h
umlal2 v5.4s, v7.8h, v1.8h
add v4.4s, v4.4s, v30.4s
add v5.4s, v5.4s, v30.4s
ushl v4.4s, v4.4s, v29.4s
ushl v5.4s, v5.4s, v29.4s
uqxtn v6.4h, v4.4s
uqxtn2 v6.8h, v5.4s
str q6, [x13], #16
cbz w10, 3f
ldr q16, [x12], #16
umull v17.4s, v16.4h, v2.4h
umull2 v18.4s, v16.8h, v2.8h
umlal v17.4s, v6.4h, v3.4h
umlal2 v18.4s, v6.8h, v3.8h
add v17.4s, v17.4s, v31.4s
add v18.4s, v18.4s, v31.4s
ushr v17.4s, v17.4s, #4
ushr v18.4s, v18.4s, #4
uqxtn v16.4h, v17.4s
uqxtn2 v16.8h, v18.4s
str q16, [x14], #16
3:
ldr d7, [src, #2]
ldr d6, [src], #8
umull v4.4s, v7.4h, v1.4h
umlal v4.4s, v6.4h, v0.4h
add v4.4s, v4.4s, v30.4s
ushl v4.4s, v4.4s, v29.4s
uqxtn v6.4h, v4.4s
str d6, [x13], #8
cbz w10, 4f
ldr d16, [x12], #8
umull v17.4s, v16.4h, v2.4h
umlal v17.4s, v6.4h, v3.4h
add v17.4s, v17.4s, v31.4s
ushr v17.4s, v17.4s, #4
uqxtn v16.4h, v17.4s
str d16, [x14], #8
4:
subs height, height, #1
mov w10, #1
add src, src, src_stride
add dst, dst, #(VVC_MAX_PB_SIZE * 2)
eor tmp0, tmp0, tmp1
eor tmp1, tmp0, tmp1
eor tmp0, tmp0, tmp1
b.ne 1b
add sp, sp, #(VVC_MAX_PB_SIZE * 4)
ret
.unreq dst
.unreq src
.unreq src_stride
.unreq height
.unreq mx
.unreq my
.unreq width
.unreq tmp0
.unreq tmp1
endfunc