///*****************************************************************************
//*
//* Copyright (C) 2012 Ittiam Systems Pvt Ltd, Bangalore
//*
//* Licensed under the Apache License, Version 2.0 (the "License");
//* you may not use this file except in compliance with the License.
//* You may obtain a copy of the License at:
//*
//* http://www.apache.org/licenses/LICENSE-2.0
//*
//* Unless required by applicable law or agreed to in writing, software
//* distributed under the License is distributed on an "AS IS" BASIS,
//* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
//* See the License for the specific language governing permissions and
//* limitations under the License.
//*
//*****************************************************************************/
///**
//*******************************************************************************
//* ,:file
//* ihevc_sao_band_offset_chroma.s
//*
//* ,:brief
//* Contains function definitions for inter prediction interpolation.
//* Functions are coded using NEON intrinsics and can be compiled using@ ARM
//* RVCT
//*
//* ,:author
//* Parthiban V
//*
//* ,:par List of Functions:
//*
//*
//* ,:remarks
//* None
//*
//*******************************************************************************
//*/
//void ihevc_sao_band_offset_chroma(UWORD8 *pu1_src,
// WORD32 src_strd,
// UWORD8 *pu1_src_left,
// UWORD8 *pu1_src_top,
// UWORD8 *pu1_src_top_left,
// WORD32 sao_band_pos_u,
// WORD32 sao_band_pos_v,
// WORD8 *pi1_sao_offset_u,
// WORD8 *pi1_sao_offset_v,
// WORD32 wd,
// WORD32 ht)
//
//**************Variables Vs Registers*****************************************
//x0 => *pu1_src
//x1 => src_strd
//x2 => *pu1_src_left
//x3 => *pu1_src_top
//x4 => *pu1_src_top_left 40
//x5 => sao_band_pos_u 44
//x6 => sao_band_pos_v 48
//x7 => *pi1_sao_offset_u 52
//x8 => *pi1_sao_offset_v 56
//x9 => wd 60
//x10=> ht 64
.text
.p2align 2
.include "ihevc_neon_macros.s"
.globl gu1_table_band_idx
.globl ihevc_sao_band_offset_chroma_av8
ihevc_sao_band_offset_chroma_av8:
mov x8,#0
mov x9,#0
mov x10,#0
ldr x8,[sp,#0]
ldr w9,[sp,#8]
ldr w10,[sp,#16]
push_v_regs
// stmfd sp!, {x4-x12, x14} //stack stores the values of the arguments
stp x19, x20,[sp,#-16]!
stp x21, x22,[sp,#-16]!
stp x23, x24,[sp,#-16]!
mov x15,x4 // pu1_src_top_left 40
mov x16,x5 // sao_band_pos_u 44
mov x17,x6 // sao_band_pos_v 48
mov x19,x7 // pi1_sao_offset_u 52
mov x20,x8 // pi1_sao_offset_v 56
mov x21,x9 // wd 60
mov x22,x10 // ht 64
MOV x4, x15 //Loads pu1_src_top_left
MOV x10, x22 //Loads ht
MOV x9, x21 //Loads wd
MOV x11,x10 //Move the ht to x9 for loop counter
ADD x12,x0,x9 //pu1_src[row * src_strd + (wd)]
ADRP x14, :got:gu1_table_band_idx
LDR x14, [x14, #:got_lo12:gu1_table_band_idx]
SUB x12,x12,#2 //wd-2
SRC_LEFT_LOOP:
LDRH w5,[x12] //Load the value
ADD x12,x12,x1
SUBS x11,x11,#1 //Decrement the loop counter
STRH w5,[x2],#2 //Store the value in pu1_src_left pointer
BNE SRC_LEFT_LOOP
MOV x5, x16 //Loads sao_band_pos_u
LD1 {v1.8b},[x14],#8 //band_table_u.val[0]
ADD x12,x3,x9 //pu1_src_top[wd]
sub x23,x12,#2
LDRH w11,[x23]
LD1 {v2.8b},[x14],#8 //band_table_u.val[1]
LSL x6,x5,#3 //sao_band_pos_u
STRH w11,[x4] //store to pu1_src_top_left[0]
LD1 {v3.8b},[x14],#8 //band_table_u.val[2]
MOV x7, x19 //Loads pi1_sao_offset_u
SUB x4,x10,#1 //ht-1
dup v31.8b,w6 //band_pos_u
mul x4, x4, x1 //ht-1 * src_strd
ADD x4,x4,x0 //pu1_src[(ht - 1) * src_strd]
LD1 {v4.8b},[x14],#8 //band_table_u.val[3]
MOV x11,x9 //Move the wd to x9 for loop counter
SRC_TOP_LOOP: //wd is always multiple of 8
LD1 {v0.8b},[x4],#8 //Load pu1_src[(ht - 1) * src_strd + col]
SUBS x11,x11,#8 //Decrement the loop counter by 8
ST1 {v0.8b},[x3],#8 //Store to pu1_src_top[col]
BNE SRC_TOP_LOOP
LD1 {v30.8b},[x7] //pi1_sao_offset_u load
ADD v5.8b, v1.8b , v31.8b //band_table_u.val[0] = vadd_u8(band_table_u.val[0], sao_band_pos_u)
dup v29.8b, v30.b[1] //vdup_n_u8(pi1_sao_offset_u[1])
ADD v6.8b, v2.8b , v31.8b //band_table_u.val[1] = vadd_u8(band_table_u.val[1], sao_band_pos_u)
dup v28.8b, v30.b[2] //vdup_n_u8(pi1_sao_offset_u[2])
ADD v7.8b, v3.8b , v31.8b //band_table_u.val[2] = vadd_u8(band_table_u.val[2], sao_band_pos_u)
dup v27.8b, v30.b[3] //vdup_n_u8(pi1_sao_offset_u[3])
ADD v8.8b, v4.8b , v31.8b //band_table_u.val[3] = vadd_u8(band_table_u.val[3], sao_band_pos_u)
CMP x5,#28
dup v26.8b, v30.b[4] //vdup_n_u8(pi1_sao_offset_u[4])
ADRP x14, :got:gu1_table_band_idx
LDR x14, [x14, #:got_lo12:gu1_table_band_idx]
movi v30.8b, #16 //vdup_n_u8(16)
ADD v1.8b, v5.8b , v29.8b //band_table_u.val[0] = vadd_u8(band_table_u.val[0], vdup_n_u8(pi1_sao_offset_u[1]))
LD1 {v9.8b},[x14],#8 //band_table_v.val[0]
ADD v2.8b, v6.8b , v28.8b //band_table_u.val[1] = vadd_u8(band_table_u.val[1], vdup_n_u8(pi1_sao_offset_u[2]))
LD1 {v10.8b},[x14],#8 //band_table_v.val[1]
ADD v3.8b, v7.8b , v27.8b //band_table_u.val[2] = vadd_u8(band_table_u.val[2], vdup_n_u8(pi1_sao_offset_u[3]))
MOV x6, x17 //Loads sao_band_pos_v
ADD v4.8b, v8.8b , v26.8b //band_table_u.val[3] = vadd_u8(band_table_u.val[3], vdup_n_u8(pi1_sao_offset_u[4]))
LSL x11,x6,#3 //sao_band_pos_v
BLT SAO_BAND_POS_U_0
SAO_BAND_POS_U_28: //case 28
cmhs v13.8b, v30.8b , v4.8b //vcle_u8(band_table.val[3], vdup_n_u8(16))
BNE SAO_BAND_POS_U_29
ORR v4.8b, v4.8b , v13.8b //band_table.val[3] = vorr_u8(band_table.val[3], au1_cmp)
B SWITCH_BREAK_U
SAO_BAND_POS_U_29: //case 29
CMP x5,#29
cmhs v14.8b, v30.8b , v3.8b //vcle_u8(band_table.val[2], vdup_n_u8(16))
BNE SAO_BAND_POS_U_30
ORR v3.8b, v3.8b , v14.8b //band_table.val[2] = vorr_u8(band_table.val[2], au1_cmp)
AND v4.8b, v4.8b , v13.8b //band_table.val[3] = vand_u8(band_table.val[3], au1_cmp)
B SWITCH_BREAK_U
SAO_BAND_POS_U_30: //case 30
CMP x5,#30
cmhs v15.8b, v30.8b , v2.8b //vcle_u8(band_table.val[1], vdup_n_u8(16))
BNE SAO_BAND_POS_U_31
ORR v2.8b, v2.8b , v15.8b //band_table.val[1] = vorr_u8(band_table.val[1], au1_cmp)
AND v3.8b, v3.8b , v14.8b //band_table.val[2] = vand_u8(band_table.val[2], au1_cmp)
SAO_BAND_POS_U_31: //case 31
CMP x5,#31
BNE SWITCH_BREAK_U
cmhs v16.8b, v30.8b , v1.8b //vcle_u8(band_table.val[0], vdup_n_u8(16))
ORR v1.8b, v1.8b , v16.8b //band_table.val[0] = vorr_u8(band_table.val[0], au1_cmp)
AND v2.8b, v2.8b , v15.8b //band_table.val[1] = vand_u8(band_table.val[1], au1_cmp)
B SWITCH_BREAK_U
SAO_BAND_POS_U_0:
CMP x5,#0 //case 0
BNE SWITCH_BREAK_U
cmhs v16.8b, v30.8b , v1.8b //vcle_u8(band_table.val[0], vdup_n_u8(16))
AND v1.8b, v1.8b , v16.8b //band_table.val[0] = vand_u8(band_table.val[0], au1_cmp)
SWITCH_BREAK_U:
dup v30.8b,w11 //band_pos_v
MOV x8, x20 //Loads pi1_sao_offset_v
LD1 {v11.8b},[x14],#8 //band_table_v.val[2]
ADD v13.8b, v9.8b , v30.8b //band_table_v.val[0] = vadd_u8(band_table_v.val[0], band_pos_v)
LD1 {v12.8b},[x14],#8 //band_table_v.val[3]
ADD v14.8b, v10.8b , v30.8b //band_table_v.val[1] = vadd_u8(band_table_v.val[1], band_pos_v)
LD1 {v25.8b},[x8] //pi1_sao_offset_v load
ADD v15.8b, v11.8b , v30.8b //band_table_v.val[2] = vadd_u8(band_table_v.val[2], band_pos_v)
dup v29.8b, v25.b[1] //vdup_n_u8(pi1_sao_offset_v[1])
ADD v16.8b, v12.8b , v30.8b //band_table_v.val[3] = vadd_u8(band_table_v.val[3], band_pos_v)
dup v28.8b, v25.b[2] //vdup_n_u8(pi1_sao_offset_v[2])
ADD v9.8b, v13.8b , v29.8b //band_table_v.val[0] = vadd_u8(band_table_v.val[0], vdup_n_u8(pi1_sao_offset_v[1]))
dup v27.8b, v25.b[3] //vdup_n_u8(pi1_sao_offset_v[3])
ADD v10.8b, v14.8b , v28.8b //band_table_v.val[1] = vadd_u8(band_table_v.val[1], vdup_n_u8(pi1_sao_offset_v[2]))
dup v26.8b, v25.b[4] //vdup_n_u8(pi1_sao_offset_v[4])
ADD v11.8b, v15.8b , v27.8b //band_table_v.val[2] = vadd_u8(band_table_v.val[2], vdup_n_u8(pi1_sao_offset_v[3]))
movi v29.8b, #16 //vdup_n_u8(16)
ADD v12.8b, v16.8b , v26.8b //band_table_v.val[3] = vadd_u8(band_table_v.val[3], vdup_n_u8(pi1_sao_offset_v[4]))
AND x12,x9,#0xf
CMP x6,#28
BLT SAO_BAND_POS_V_0
SAO_BAND_POS_V_28: //case 28
cmhs v17.8b, v29.8b , v12.8b //vcle_u8(band_table.val[3], vdup_n_u8(16))
BNE SAO_BAND_POS_V_29
ORR v12.8b, v12.8b , v17.8b //band_table.val[3] = vorr_u8(band_table.val[3], au1_cmp)
B SWITCH_BREAK_V
SAO_BAND_POS_V_29: //case 29
CMP x6,#29
cmhs v18.8b, v29.8b , v11.8b //vcle_u8(band_table.val[2], vdup_n_u8(16))
BNE SAO_BAND_POS_V_30
ORR v11.8b, v11.8b , v18.8b //band_table.val[2] = vorr_u8(band_table.val[2], au1_cmp)
AND v12.8b, v12.8b , v17.8b //band_table.val[3] = vand_u8(band_table.val[3], au1_cmp)
B SWITCH_BREAK_V
SAO_BAND_POS_V_30: //case 30
CMP x6,#30
cmhs v19.8b, v29.8b , v10.8b //vcle_u8(band_table.val[1], vdup_n_u8(16))
BNE SAO_BAND_POS_V_31
ORR v10.8b, v10.8b , v19.8b //band_table.val[1] = vorr_u8(band_table.val[1], au1_cmp)
AND v11.8b, v11.8b , v18.8b //band_table.val[2] = vand_u8(band_table.val[2], au1_cmp)
B SWITCH_BREAK_V
SAO_BAND_POS_V_31: //case 31
CMP x6,#31
BNE SWITCH_BREAK_V
cmhs v20.8b, v29.8b , v9.8b //vcle_u8(band_table.val[0], vdup_n_u8(16))
ORR v9.8b, v9.8b , v20.8b //band_table.val[0] = vorr_u8(band_table.val[0], au1_cmp)
AND v10.8b, v10.8b , v19.8b //band_table.val[1] = vand_u8(band_table.val[1], au1_cmp)
B SWITCH_BREAK_V
SAO_BAND_POS_V_0:
CMP x6,#0 //case 0
BNE SWITCH_BREAK_V
cmhs v20.8b, v29.8b , v9.8b //vcle_u8(band_table.val[0], vdup_n_u8(16))
AND v9.8b, v9.8b , v20.8b //band_table.val[0] = vand_u8(band_table.val[0], au1_cmp)
SWITCH_BREAK_V:
CMP x9,#16
MOV x4,x0 //pu1_src_cpy
mov v1.d[1],v2.d[0]
mov v2.d[0],v3.d[0]
mov v2.d[1],v4.d[0]
mov v9.d[1],v10.d[0]
mov v10.d[0],v11.d[0]
mov v10.d[1],v12.d[0]
BLT WIDTH_RESIDUE
WIDTH_LOOP: //Width is assigned to be multiple of 16
MOV x4,x0 //pu1_src_cpy
MOV x11,x10 //move ht
ADD x5,x4,x1
HEIGHT_LOOP: //unrolled for 4 rows
ADD x6,x5,x1
LD2 {v5.8b, v6.8b},[x4] //vld1q_u8(pu1_src_cpy)
ADD x7,x6,x1
LD2 {v13.8b, v14.8b},[x5] //vld1q_u8(pu1_src_cpy)
SUB v7.8b, v5.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
LD2 {v17.8b, v18.8b},[x6] //vld1q_u8(pu1_src_cpy)
SUB v8.8b, v6.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
LD2 {v21.8b, v22.8b},[x7] //vld1q_u8(pu1_src_cpy)
SUB v15.8b, v13.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
TBX v5.8b, {v1.16b- v2.16b},v7.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
SUB v16.8b, v14.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
TBX v6.8b, {v9.16b- v10.16b},v8.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
SUB v19.8b, v17.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
TBX v13.8b, {v1.16b- v2.16b},v15.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
SUB v20.8b, v18.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
TBX v14.8b, {v9.16b- v10.16b},v16.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
SUB v23.8b, v21.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
ST2 {v5.8b, v6.8b},[x4] //vst1q_u8(pu1_src_cpy, au1_cur_row)
SUB v24.8b, v22.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
SUBS x11,x11,#4 //Decrement the ht loop count by 4
TBX v17.8b, {v1.16b- v2.16b},v19.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
ST2 {v13.8b, v14.8b},[x5] //vst1q_u8(pu1_src_cpy, au1_cur_row)
TBX v18.8b, {v9.16b- v10.16b},v20.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
TBX v21.8b, {v1.16b- v2.16b},v23.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
TBX v22.8b, {v9.16b- v10.16b},v24.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
ST2 {v17.8b, v18.8b},[x6],x1 //vst1q_u8(pu1_src_cpy, au1_cur_row)
ADD x4,x6,x1
ST2 {v21.8b, v22.8b},[x7] //vst1q_u8(pu1_src_cpy, au1_cur_row)
ADD x5,x4,x1
BNE HEIGHT_LOOP
SUB x9,x9,#16 //Decrement the width loop by 16
ADD x0,x0,#16
CMP x9,#8
BGT WIDTH_LOOP
BLT END_LOOP
MOV x4,x0 //pu1_src_cpy
WIDTH_RESIDUE: //If width is not multiple of 16
ADD x5,x4,x1
LD2 {v5.8b, v6.8b},[x4] //vld1q_u8(pu1_src_cpy)
ADD x6,x5,x1
ADD x7,x6,x1
LD2 {v13.8b, v14.8b},[x5] //vld1q_u8(pu1_src_cpy)
SUB v7.8b, v5.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
LD2 {v17.8b, v18.8b},[x6] //vld1q_u8(pu1_src_cpy)
SUB v8.8b, v6.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
TBX v5.8b, {v1.16b- v2.16b},v7.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
SUB v15.8b, v13.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
TBX v6.8b, {v9.16b- v10.16b},v8.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
SUB v16.8b, v14.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
LD2 {v21.8b, v22.8b},[x7] //vld1q_u8(pu1_src_cpy)
SUB v19.8b, v17.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
TBX v13.8b, {v1.16b- v2.16b},v15.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
SUB v20.8b, v18.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
TBX v14.8b, {v9.16b- v10.16b},v16.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
ZIP1 v28.8b, v5.8b, v6.8b
ZIP2 v6.8b, v5.8b, v6.8b
mov v5.8b, v28.8b
TBX v17.8b, {v1.16b- v2.16b},v19.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
SUB v23.8b, v21.8b , v31.8b //vsub_u8(au1_cur_row_deint.val[0], band_pos_u)
ST1 {v5.8b},[x4] //vst1q_u8(pu1_src_cpy, au1_cur_row)
ZIP1 v28.8b, v13.8b, v14.8b
ZIP2 v14.8b, v13.8b, v14.8b
mov v13.8b, v28.8b
TBX v18.8b, {v9.16b- v10.16b},v20.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
SUB v24.8b, v22.8b , v30.8b //vsub_u8(au1_cur_row_deint.val[1], band_pos_v)
ST1 {v13.8b},[x5] //vst1q_u8(pu1_src_cpy, au1_cur_row)
SUBS x10,x10,#4 //Decrement the ht loop count by 4
TBX v21.8b, {v1.16b- v2.16b},v23.8b //vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u))
ZIP1 v28.8b, v17.8b, v18.8b
ZIP2 v18.8b, v17.8b, v18.8b
mov v17.8b, v28.8b
TBX v22.8b, {v9.16b- v10.16b},v24.8b //vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v))
ST1 {v17.8b},[x6],x1 //vst1q_u8(pu1_src_cpy, au1_cur_row)
ZIP1 v28.8b, v21.8b, v22.8b
ZIP2 v22.8b, v21.8b, v22.8b
mov v21.8b, v28.8b
ADD x4,x6,x1
ST1 {v21.8b},[x7] //vst1q_u8(pu1_src_cpy, au1_cur_row)
ADD x5,x4,x1
BNE WIDTH_RESIDUE
END_LOOP:
// LDMFD sp!,{x4-x12,x15} //Reload the registers from SP
ldp x23, x24,[sp],#16
ldp x21, x22,[sp],#16
ldp x19, x20,[sp],#16
pop_v_regs
ret