@/*****************************************************************************
@*
@* 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_weighted_pred_bi_default.s
@*
@* @brief
@* contains function definitions for weighted prediction used in inter
@* prediction
@*
@* @author
@* parthiban v
@*
@* @par list of functions:
@* - ihevc_weighted_pred_bi_default()
@*
@* @remarks
@* none
@*
@*******************************************************************************
@*/
@/**
@*******************************************************************************
@*
@* @brief
@* does default bi-weighted prediction on the arrays pointed by pi2_src1 and
@* pi2_src2 and stores it at location pointed by pi2_dst assumptions : the
@* function is optimized considering the fact width and height are multiple
@* of 2.
@*
@* @par description:
@* dst = ( (src1 + lvl_shift1) + (src2 + lvl_shift2) + 1 << (shift - 1) )
@* >> shift where shift = 15 - bitdepth
@*
@* @param[in] pi2_src1
@* pointer to source 1
@*
@* @param[in] pi2_src2
@* pointer to source 2
@*
@* @param[out] pu1_dst
@* pointer to destination
@*
@* @param[in] src_strd1
@* source stride 1
@*
@* @param[in] src_strd2
@* source stride 2
@*
@* @param[in] dst_strd
@* destination stride
@*
@* @param[in] lvl_shift1
@* added before shift and offset
@*
@* @param[in] lvl_shift2
@* added before shift and offset
@*
@* @param[in] ht
@* height of the source
@*
@* @param[in] wd
@* width of the source
@*
@* @returns
@*
@* @remarks
@* none
@*
@*******************************************************************************
@*/
@void ihevc_weighted_pred_bi_default(word16 *pi2_src1,
@ word16 *pi2_src2,
@ uword8 *pu1_dst,
@ word32 src_strd1,
@ word32 src_strd2,
@ word32 dst_strd,
@ word32 lvl_shift1,
@ word32 lvl_shift2,
@ word32 ht,
@ word32 wd)
@**************variables vs registers*****************************************
@ r0 => *pi2_src1
@ r1 => *pi2_src2
@ r2 => *pu1_dst
@ r3 => src_strd1
@ r4 => src_strd2
@ r5 => dst_strd
@ r6 => lvl_shift1
@ r7 => lvl_shift2
@ r8 => ht
@ r9 => wd
.equ src_strd2_offset, 104
.equ dst_strd_offset, 108
.equ lvl_shift1_offset, 112
.equ lvl_shift2_offset, 116
.equ ht_offset, 120
.equ wd_offset, 124
.text
.syntax unified
.align 4
.globl ihevc_weighted_pred_bi_default_a9q
.type ihevc_weighted_pred_bi_default_a9q, %function
ihevc_weighted_pred_bi_default_a9q:
stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments
vpush {d8 - d15}
ldr r4,[sp,#src_strd2_offset] @load src_strd2
lsl r3,r3,#1
ldr r5,[sp,#dst_strd_offset] @load dst_strd
ldr r6,[sp,#lvl_shift1_offset] @load lvl_shift1
lsl r4,r4,#1
ldr r7,[sp,#lvl_shift2_offset] @load lvl_shift2
ldr r8,[sp,#ht_offset] @load ht
ldr r9,[sp,#wd_offset] @load wd
vdup.16 q2,r6 @lvl_shift1_t = vmov_n_s16((int16_t)lvl_shift1)
vdup.16 q3,r7 @lvl_shift2_t = vmov_n_s16((int16_t)lvl_shift2)
vmov.i16 q0,#0x40 @tmp_lvl_shift = 1 << (shift - 1)
vadd.i16 q2,q3
vadd.s16 q0,q0,q2
@ vmvn.i32 q1,#0x6 @vmovq_n_s32(tmp_shift)
lsl r6,r9,#1
rsb r7,r6,r3,lsl #2 @4*src_strd1 - wd
rsb r10,r6,r4,lsl #2 @4*src_strd2 - wd
@asr r6,#1
@rsb r6,r6,r5,lsl #2 @4*dst_strd - wd
cmp r8,#0 @check ht == 0
beq end_loops @if equal, then end the function
chroma_decision:
orr r14,r8,r9
cmp r14,#10
beq outer_loop_chroma_8x2
cmp r14,#6
beq outer_loop_chroma_4x2
luma_decision:
cmp r9,#24
beq outer_loop_8
cmp r9,#16
bge outer_loop_16
cmp r9,#12
beq outer_loop_4
cmp r9,#8
bge outer_loop_8
outer_loop_4:
cmp r9,#0 @check wd == 0
beq end_loops @if equal, then end the function
core_loop_4:
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
vld1.s16 {d6},[r0]! @load and increment the pi2_src1
add r14,r2,r5 @pu1_dst_tmp = pu1_dst + dst_strd
vld1.s16 {d7},[r1]! @load and increment the pi2_src2
vld1.s16 {d8},[r11],r3 @load and increment the pi2_src1 ii iteration
vqadd.s16 d18,d6,d7
vqadd.s16 d18,d18,d0 @vaddq_s32(i4_tmp1_t1, tmp_lvl_shift_t)
vld1.s16 {d9},[r12],r4 @load and increment the pi2_src2 ii iteration
vqadd.s16 d20,d8,d9 @vaddq_s32(i4_tmp2_t1, i4_tmp2_t2)
vqadd.s16 d19,d20,d0 @vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t)
vqshrun.s16 d20,q9,#7
vld1.s16 {d22},[r11],r3 @load and increment the pi2_src1 iii iteration
vld1.s16 {d23},[r12],r4 @load and increment the pi2_src2 iii iteration
vqadd.s16 d30,d22,d23
vqadd.s16 d30,d30,d0 @vaddq_s32(i4_tmp1_t1, tmp_lvl_shift_t) iii iteration
vld1.s16 {d24},[r11],r3 @load and increment the pi2_src1 iv iteration
vld1.s16 {d25},[r12],r4 @load and increment the pi2_src2 iv iteration
vqadd.s16 d18,d24,d25 @vaddq_s32(i4_tmp2_t1, i4_tmp2_t2) iv iteration
vqadd.s16 d31,d18,d0
vst1.32 {d20[0]},[r2]! @store pu1_dst i iteration
vst1.32 {d20[1]},[r14],r5 @store pu1_dst ii iteration
vqshrun.s16 d30,q15,#7
vst1.32 {d30[0]},[r14],r5 @store pu1_dst iii iteration @vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t) iv iteratio
subs r9,r9,#4 @decrement wd by 4 and check for 0
vst1.32 {d30[1]},[r14],r5 @store pu1_dst iv iteration
bgt core_loop_4 @if greater than 0 repeat the core loop again
end_core_loop_4:
subs r8,r8,#4 @decrement the ht by 4
add r0,r0,r7 @pi2_src1 + 4*src_strd1 - 2*wd(since pi2_src1 is 16 bit pointer double the increment with double the wd decrement)
asr r9,r6,#1
add r1,r1,r10 @pi2_src2 + 4*src_strd2 - 2*wd
rsb r14,r9,r5,lsl #2 @4*dst_strd - wd
add r2,r2,r14
@pu1_dst + dst_std - wd
bgt core_loop_4 @if ht is greater than 0 goto outer_loop
b end_loops
@ this is only for chroma module with input 2x2
outer_loop_chroma_4x2:
cmp r9,#0 @check wd == 0
beq end_loops @if equal, then end the function
rsb r7,r6,r3,lsl #1 @2*src_strd1 - wd
rsb r10,r6,r4,lsl #1 @2*src_strd2 - wd
core_loop_chroma_4x2:
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
vld1.s16 {d6},[r0]! @load and increment the pi2_src1
add r14,r2,r5 @pu1_dst_tmp = pu1_dst + dst_strd
vld1.s16 {d7},[r1]! @load and increment the pi2_src2
vld1.s16 {d8},[r11],r3 @load and increment the pi2_src1 ii iteration
vqadd.s16 d18,d6,d7
vqadd.s16 d18,d18,d0 @vaddq_s32(i4_tmp1_t1, tmp_lvl_shift_t)
vld1.s16 {d9},[r12],r4 @load and increment the pi2_src2 ii iteration
vqadd.s16 d20,d8,d9 @vaddq_s32(i4_tmp2_t1, i4_tmp2_t2)
vqadd.s16 d19,d20,d0 @vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t)
vqshrun.s16 d20,q9,#7
vst1.32 {d20[0]},[r2]! @store pu1_dst i iteration
vst1.32 {d20[1]},[r14],r5 @store pu1_dst ii iteration
subs r9,r9,#4 @decrement wd by 4 and check for 0
bgt core_loop_chroma_4x2 @if greater than 0 repeat the core loop again
end_core_loop_chorma_4x2:
subs r8,r8,#2 @decrement the ht by 4
add r0,r0,r7 @pi2_src1 + 2*src_strd1 - 2*wd(since pi2_src1 is 16 bit pointer double the increment with double the wd decrement)
asr r9,r6,#1
add r1,r1,r10 @pi2_src2 + 2*src_strd2 - 2*wd
rsb r14,r9,r5,lsl #1 @2*dst_strd - wd
add r2,r2,r14
@pu1_dst + dst_std - wd
bgt core_loop_chroma_4x2 @if ht is greater than 0 goto outer_loop
b end_loops
outer_loop_8:
cmp r9,#0 @check wd == 0
beq end_loops @if equal, then end the function
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
core_loop_8:
vld1.s16 {q12},[r0]! @load and increment the pi2_src1
add r14,r2,r5 @pu1_dst_tmp = pu1_dst + dst_strd
vld1.s16 {q13},[r1]! @load and increment the pi2_src2
vqadd.s16 q12,q12,q13
vld1.s16 {q14},[r11],r3 @load and increment the pi2_src1 ii iteration
vqadd.s16 q12,q12,q0 @vaddq_s32(i4_tmp1_t1, tmp_lvl_shift_t)
vld1.s16 {q15},[r12],r4 @load and increment the pi2_src2 ii iteration
vld1.s16 {q8},[r11],r3 @load and increment the pi2_src1 iii iteration
vqadd.s16 q11,q14,q15 @vaddq_s32(i4_tmp2_t1, i4_tmp2_t2)
vld1.s16 {q9},[r12],r4 @load and increment the pi2_src2 iii iteration
vqadd.s16 q11,q11,q0 @vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t)
vqshrun.s16 d20,q12,#7
vld1.s16 {q6},[r11],r3 @load and increment the pi2_src1 iv iteration
vqadd.s16 q15,q8,q9
vqshrun.s16 d21,q11,#7
vld1.s16 {q7},[r12],r4 @load and increment the pi2_src2 iv iteration
vqadd.s16 q15,q15,q0 @vaddq_s32(i4_tmp1_t1, tmp_lvl_shift_t) iii iteration
vst1.32 {d20},[r2]! @store pu1_dst i iteration
vqadd.s16 q4,q6,q7 @vaddq_s32(i4_tmp2_t1, i4_tmp2_t2) iv iteration
vst1.32 {d21},[r14],r5 @store pu1_dst ii iteration
vqadd.s16 q4,q4,q0
vqshrun.s16 d30,q15,#7
vqshrun.s16 d31,q4,#7
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
vst1.32 {d30},[r14],r5 @store pu1_dst iii iteration @vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t) iv iteratio
subs r9,r9,#8 @decrement wd by 4 and check for 0
vst1.32 {d31},[r14],r5 @store pu1_dst iv iteration
bgt core_loop_8 @if greater than 0 repeat the core loop again
end_core_loop_8:
subs r8,r8,#4 @decrement the ht by 4
add r0,r0,r7 @pi2_src1 + 4*src_strd1 - 2*wd(since pi2_src1 is 16 bit pointer double the increment with double the wd decrement)
asr r9,r6,#1
add r1,r1,r10 @pi2_src2 + 4*src_strd2 - 2*wd
rsb r14,r9,r5,lsl #2 @4*dst_strd - wd
add r2,r2,r14
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer) @pu1_dst + dst_std - wd
bgt core_loop_8
b end_loops
@ this is only for chroma module with inpput 4x2
outer_loop_chroma_8x2:
cmp r9,#0 @check wd == 0
beq end_loops @if equal, then end the function
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
rsb r7,r6,r3,lsl #1 @2*src_strd1 - wd
rsb r10,r6,r4,lsl #1 @2*src_strd2 - wd
core_loop_chroma_8x2:
vld1.s16 {q12},[r0]! @load and increment the pi2_src1
add r14,r2,r5 @pu1_dst_tmp = pu1_dst + dst_strd
vld1.s16 {q13},[r1]! @load and increment the pi2_src2
vqadd.s16 q12,q12,q13
vld1.s16 {q14},[r11],r3 @load and increment the pi2_src1 ii iteration
vqadd.s16 q12,q12,q0 @vaddq_s32(i4_tmp1_t1, tmp_lvl_shift_t)
vld1.s16 {q15},[r12],r4 @load and increment the pi2_src2 ii iteration
vld1.s16 {q8},[r11],r3 @load and increment the pi2_src1 iii iteration
vqadd.s16 q11,q14,q15 @vaddq_s32(i4_tmp2_t1, i4_tmp2_t2)
vqadd.s16 q11,q11,q0 @vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t)
vqshrun.s16 d20,q12,#7
vqshrun.s16 d21,q11,#7
vst1.32 {d20},[r2]! @store pu1_dst i iteration
vst1.32 {d21},[r14],r5 @store pu1_dst ii iteration
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
@vaddq_s32(i4_tmp2_t1, tmp_lvl_shift_t) iv iteratio
subs r9,r9,#8 @decrement wd by 4 and check for 0
bgt core_loop_chroma_8x2 @if greater than 0 repeat the core loop again
end_core_loop_chroma_8x2:
subs r8,r8,#2 @decrement the ht by 4
add r0,r0,r7 @pi2_src1 + 4*src_strd1 - 2*wd(since pi2_src1 is 16 bit pointer double the increment with double the wd decrement)
asr r9,r6,#1
add r1,r1,r10 @pi2_src2 + 4*src_strd2 - 2*wd
rsb r14,r9,r5,lsl #1 @4*dst_strd - wd
add r2,r2,r14
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer) @pu1_dst + dst_std - wd
bgt core_loop_chroma_8x2
b end_loops
outer_loop_16:
cmp r9,#0 @check wd == 0
beq end_loops @if equal, then end the function
add r11,r0,r3 @pi2_src_tmp1 = pi2_src1 + 2*src_strd1(2* because pi1_src is a 16 bit pointer)
add r12,r1,r4 @pi2_src_tmp2 = pi2_src2 + 2*src_strd2(2* because pi2_src is a 16 bit pointer)
rsb r7,r6,r3,lsl #1 @2*src_strd1 - wd
mov r14,#16
sub r10,r14,r5
sub r11,r3,r14
sub r12,r14,r3
rsb r14,r9,r5,lsl #1 @2*dst_strd - wd
prolog_16:
vld1.s16 {q1},[r0]! @load and increment the pi2_src1
vld1.s16 {q2},[r1]! @load and increment the pi2_src2
vld1.s16 {q5},[r0],r11 @load and increment the pi2_src1
vld1.s16 {q6},[r1],r11 @load and increment the pi2_src2
vld1.s16 {q3},[r0]! @load and increment the pi2_src1 ii iteration
subs r9,r9,#16
vld1.s16 {q4},[r1]! @load and increment the pi2_src2 ii iteration
subeq r8,r8,#2
vqadd.s16 q11,q1,q2
vld1.s16 {q7},[r0],r12 @load and increment the pi2_src1 ii iteration
vqadd.s16 q14,q5,q6
vld1.s16 {q8},[r1],r12 @load and increment the pi2_src2 ii iteration
addeq r0,r0,r7
addeq r1,r1,r7
vqadd.s16 q12,q3,q4
vld1.s16 {q1},[r0]!
vqadd.s16 q13,q7,q8
@ if the input is chroma with 8x2 block size
cmp r8,#0
beq epilog_16
vld1.s16 {q2},[r1]! @load and increment the pi2_src2
vqadd.s16 q11,q11,q0
vld1.s16 {q5},[r0],r11 @load and increment the pi2_src1
vqadd.s16 q14,q14,q0
vld1.s16 {q6},[r1],r11 @load and increment the pi2_src2
vqadd.s16 q12,q12,q0
vld1.s16 {q3},[r0]! @load and increment the pi2_src1 ii iteration
vqadd.s16 q15,q13,q0
vqshrun.s16 d20,q11,#7
vld1.s16 {q4},[r1]! @load and increment the pi2_src2 ii iteration
vqshrun.s16 d21,q14,#7
vld1.s16 {q7},[r0],r12 @load and increment the pi2_src1 ii iteration
vqshrun.s16 d26,q12,#7
vld1.s16 {q8},[r1],r12 @load and increment the pi2_src2 ii iteration
vqshrun.s16 d27,q15,#7
core_loop_16:
cmp r9,#0
vqadd.s16 q11,q1,q2
asreq r9,r6,#1
vst1.32 {q10},[r2],r5
vqadd.s16 q14,q5,q6
vst1.32 {q13},[r2],r10
addeq r2,r2,r14
vqadd.s16 q12,q3,q4
subs r9,r9,#16
addeq r0,r0,r7
vqadd.s16 q13,q7,q8
addeq r1,r1,r7
subseq r8,r8,#2 @decrement the ht by 2
beq epilog_16
vqadd.s16 q11,q11,q0
vld1.s16 {q1},[r0]! @load and increment the pi2_src1
vqadd.s16 q14,q14,q0
vld1.s16 {q2},[r1]! @load and increment the pi2_src2
vqadd.s16 q12,q12,q0
vld1.s16 {q5},[r0],r11 @load and increment the pi2_src1
vqadd.s16 q15,q13,q0
vld1.s16 {q6},[r1],r11 @load and increment the pi2_src2
vqshrun.s16 d20,q11,#7
vld1.s16 {q3},[r0]! @load and increment the pi2_src1 ii iteration
vqshrun.s16 d21,q14,#7
vld1.s16 {q4},[r1]! @load and increment the pi2_src2 ii iteration
vqshrun.s16 d26,q12,#7
vld1.s16 {q7},[r0],r12 @load and increment the pi2_src1 ii iteration
vqshrun.s16 d27,q15,#7
vld1.s16 {q8},[r1],r12 @load and increment the pi2_src2 ii iteration
b core_loop_16
epilog_16:
vqadd.s16 q11,q11,q0
vqadd.s16 q14,q14,q0
vqadd.s16 q12,q12,q0
vqadd.s16 q15,q13,q0
vqshrun.s16 d20,q11,#7
vqshrun.s16 d21,q14,#7
vqshrun.s16 d26,q12,#7
vqshrun.s16 d27,q15,#7
vst1.32 {q10},[r2],r5
vst1.32 {q13},[r2]
end_core_loop_16:
end_loops:
vpop {d8 - d15}
ldmfd sp!,{r4-r12,r15} @reload the registers from sp