/*
* Copyright © 2018, VideoLAN and dav1d authors
* Copyright © 2018, Martin Storsjo
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice, this
* list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "src/arm/asm.S"
#include "util.S"
// void dav1d_sgr_box3_vert_neon(int32_t **sumsq, int16_t **sum,
// int32_t *AA, int16_t *BB,
// const int w, const int s,
// const int bitdepth_max);
function sgr_box3_vert_neon, export=1
stp d8, d9, [sp, #-0x30]!
stp d10, d11, [sp, #0x10]
stp d12, d13, [sp, #0x20]
add w4, w4, #2
clz w9, w6 // bitdepth_max
dup v28.4s, w5 // strength
ldp x5, x6, [x0]
ldr x0, [x0, #16]
ldp x7, x8, [x1]
ldr x1, [x1, #16]
movi v31.4s, #9 // n
sub w9, w9, #24 // -bitdepth_min_8
movrel x12, X(sgr_x_by_x)
mov w13, #455 // one_by_x
ld1 {v16.16b, v17.16b, v18.16b}, [x12]
dup v6.8h, w9 // -bitdepth_min_8
movi v19.16b, #5
movi v20.8b, #55 // idx of last 5
movi v21.8b, #72 // idx of last 4
movi v22.8b, #101 // idx of last 3
movi v23.8b, #169 // idx of last 2
movi v24.8b, #254 // idx of last 1
saddl v7.4s, v6.4h, v6.4h // -2*bitdepth_min_8
movi v29.8h, #1, lsl #8
dup v30.4s, w13 // one_by_x
sub v16.16b, v16.16b, v19.16b
sub v17.16b, v17.16b, v19.16b
sub v18.16b, v18.16b, v19.16b
ld1 {v8.4s, v9.4s}, [x5], #32
ld1 {v10.4s, v11.4s}, [x6], #32
ld1 {v12.8h}, [x7], #16
ld1 {v13.8h}, [x8], #16
ld1 {v0.4s, v1.4s}, [x0], #32
ld1 {v2.8h}, [x1], #16
1:
add v8.4s, v8.4s, v10.4s
add v9.4s, v9.4s, v11.4s
add v12.8h, v12.8h, v13.8h
subs w4, w4, #8
add v0.4s, v0.4s, v8.4s
add v1.4s, v1.4s, v9.4s
add v2.8h, v2.8h, v12.8h
srshl v0.4s, v0.4s, v7.4s
srshl v1.4s, v1.4s, v7.4s
srshl v4.8h, v2.8h, v6.8h
mul v0.4s, v0.4s, v31.4s // a * n
mul v1.4s, v1.4s, v31.4s // a * n
umull v3.4s, v4.4h, v4.4h // b * b
umull2 v4.4s, v4.8h, v4.8h // b * b
uqsub v0.4s, v0.4s, v3.4s // imax(a * n - b * b, 0)
uqsub v1.4s, v1.4s, v4.4s // imax(a * n - b * b, 0)
mul v0.4s, v0.4s, v28.4s // p * s
mul v1.4s, v1.4s, v28.4s // p * s
ld1 {v8.4s, v9.4s}, [x5], #32
uqshrn v0.4h, v0.4s, #16
uqshrn2 v0.8h, v1.4s, #16
ld1 {v10.4s, v11.4s}, [x6], #32
uqrshrn v0.8b, v0.8h, #4 // imin(z, 255)
ld1 {v12.8h}, [x7], #16
cmhi v25.8b, v0.8b, v20.8b // = -1 if sgr_x_by_x[v0] < 5
cmhi v26.8b, v0.8b, v21.8b // = -1 if sgr_x_by_x[v0] < 4
tbl v1.8b, {v16.16b, v17.16b, v18.16b}, v0.8b
cmhi v27.8b, v0.8b, v22.8b // = -1 if sgr_x_by_x[v0] < 3
cmhi v4.8b, v0.8b, v23.8b // = -1 if sgr_x_by_x[v0] < 2
add v25.8b, v25.8b, v26.8b
cmhi v5.8b, v0.8b, v24.8b // = -1 if sgr_x_by_x[v0] < 1
add v27.8b, v27.8b, v4.8b
add v5.8b, v5.8b, v19.8b
add v25.8b, v25.8b, v27.8b
add v5.8b, v1.8b, v5.8b
ld1 {v13.8h}, [x8], #16
add v5.8b, v5.8b, v25.8b
ld1 {v0.4s, v1.4s}, [x0], #32
uxtl v5.8h, v5.8b // x
umull v3.4s, v5.4h, v2.4h // x * BB[i]
umull2 v4.4s, v5.8h, v2.8h // x * BB[i]
mul v3.4s, v3.4s, v30.4s // x * BB[i] * sgr_one_by_x
mul v4.4s, v4.4s, v30.4s // x * BB[i] * sgr_one_by_x
srshr v3.4s, v3.4s, #12 // AA[i]
srshr v4.4s, v4.4s, #12 // AA[i]
sub v5.8h, v29.8h, v5.8h // 256 - x
ld1 {v2.8h}, [x1], #16
st1 {v3.4s, v4.4s}, [x2], #32
st1 {v5.8h}, [x3], #16
b.gt 1b
ldp d12, d13, [sp, #0x20]
ldp d10, d11, [sp, #0x10]
ldp d8, d9, [sp], 0x30
ret
endfunc
// void dav1d_sgr_box5_vert_neon(int32_t **sumsq, int16_t **sum,
// int32_t *AA, int16_t *BB,
// const int w, const int s,
// const int bitdepth_max);
function sgr_box5_vert_neon, export=1
stp d8, d9, [sp, #-0x40]!
stp d10, d11, [sp, #0x10]
stp d12, d13, [sp, #0x20]
stp d14, d15, [sp, #0x30]
add w4, w4, #2
clz w15, w6 // bitdepth_max
dup v28.4s, w5 // strength
ldp x5, x6, [x0]
ldp x7, x8, [x0, #16]
ldr x0, [x0, #32]
ldp x9, x10, [x1]
ldp x11, x12, [x1, #16]
ldr x1, [x1, #32]
movi v31.4s, #25 // n
sub w15, w15, #24 // -bitdepth_min_8
movrel x13, X(sgr_x_by_x)
mov w14, #164 // one_by_x
ld1 {v16.16b, v17.16b, v18.16b}, [x13]
dup v6.8h, w15 // -bitdepth_min_8
movi v19.16b, #5
movi v24.8b, #254 // idx of last 1
saddl v7.4s, v6.4h, v6.4h // -2*bitdepth_min_8
movi v29.8h, #1, lsl #8
dup v30.4s, w14 // one_by_x
sub v16.16b, v16.16b, v19.16b
sub v17.16b, v17.16b, v19.16b
sub v18.16b, v18.16b, v19.16b
ld1 {v8.4s, v9.4s}, [x5], #32
ld1 {v10.4s, v11.4s}, [x6], #32
ld1 {v12.4s, v13.4s}, [x7], #32
ld1 {v14.4s, v15.4s}, [x8], #32
ld1 {v20.8h}, [x9], #16
ld1 {v21.8h}, [x10], #16
ld1 {v22.8h}, [x11], #16
ld1 {v23.8h}, [x12], #16
ld1 {v0.4s, v1.4s}, [x0], #32
ld1 {v2.8h}, [x1], #16
1:
add v8.4s, v8.4s, v10.4s
add v9.4s, v9.4s, v11.4s
add v12.4s, v12.4s, v14.4s
add v13.4s, v13.4s, v15.4s
add v20.8h, v20.8h, v21.8h
add v22.8h, v22.8h, v23.8h
add v0.4s, v0.4s, v8.4s
add v1.4s, v1.4s, v9.4s
add v2.8h, v2.8h, v20.8h
add v0.4s, v0.4s, v12.4s
add v1.4s, v1.4s, v13.4s
add v2.8h, v2.8h, v22.8h
subs w4, w4, #8
movi v20.8b, #55 // idx of last 5
movi v21.8b, #72 // idx of last 4
movi v22.8b, #101 // idx of last 3
movi v23.8b, #169 // idx of last 2
srshl v0.4s, v0.4s, v7.4s
srshl v1.4s, v1.4s, v7.4s
srshl v4.8h, v2.8h, v6.8h
mul v0.4s, v0.4s, v31.4s // a * n
mul v1.4s, v1.4s, v31.4s // a * n
umull v3.4s, v4.4h, v4.4h // b * b
umull2 v4.4s, v4.8h, v4.8h // b * b
uqsub v0.4s, v0.4s, v3.4s // imax(a * n - b * b, 0)
uqsub v1.4s, v1.4s, v4.4s // imax(a * n - b * b, 0)
mul v0.4s, v0.4s, v28.4s // p * s
mul v1.4s, v1.4s, v28.4s // p * s
ld1 {v8.4s, v9.4s}, [x5], #32
uqshrn v0.4h, v0.4s, #16
uqshrn2 v0.8h, v1.4s, #16
ld1 {v10.4s, v11.4s}, [x6], #32
uqrshrn v0.8b, v0.8h, #4 // imin(z, 255)
ld1 {v12.4s, v13.4s}, [x7], #32
cmhi v25.8b, v0.8b, v20.8b // = -1 if sgr_x_by_x[v0] < 5
cmhi v26.8b, v0.8b, v21.8b // = -1 if sgr_x_by_x[v0] < 4
tbl v1.8b, {v16.16b, v17.16b, v18.16b}, v0.8b
cmhi v27.8b, v0.8b, v22.8b // = -1 if sgr_x_by_x[v0] < 3
cmhi v4.8b, v0.8b, v23.8b // = -1 if sgr_x_by_x[v0] < 2
ld1 {v14.4s, v15.4s}, [x8], #32
add v25.8b, v25.8b, v26.8b
cmhi v5.8b, v0.8b, v24.8b // = -1 if sgr_x_by_x[v0] < 1
add v27.8b, v27.8b, v4.8b
ld1 {v20.8h}, [x9], #16
add v5.8b, v5.8b, v19.8b
add v25.8b, v25.8b, v27.8b
ld1 {v21.8h}, [x10], #16
add v5.8b, v1.8b, v5.8b
ld1 {v22.8h}, [x11], #16
add v5.8b, v5.8b, v25.8b
ld1 {v23.8h}, [x12], #16
uxtl v5.8h, v5.8b // x
ld1 {v0.4s, v1.4s}, [x0], #32
umull v3.4s, v5.4h, v2.4h // x * BB[i]
umull2 v4.4s, v5.8h, v2.8h // x * BB[i]
mul v3.4s, v3.4s, v30.4s // x * BB[i] * sgr_one_by_x
mul v4.4s, v4.4s, v30.4s // x * BB[i] * sgr_one_by_x
srshr v3.4s, v3.4s, #12 // AA[i]
srshr v4.4s, v4.4s, #12 // AA[i]
sub v5.8h, v29.8h, v5.8h // 256 - x
ld1 {v2.8h}, [x1], #16
st1 {v3.4s, v4.4s}, [x2], #32
st1 {v5.8h}, [x3], #16
b.gt 1b
ldp d14, d15, [sp, #0x30]
ldp d12, d13, [sp, #0x20]
ldp d10, d11, [sp, #0x10]
ldp d8, d9, [sp], 0x40
ret
endfunc