@/*****************************************************************************
@*
@* 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_inter_pred_chroma_vert_w16out_neon.s
@*
@* @brief
@* contains function definitions for inter prediction interpolation.
@* functions are coded using neon intrinsics and can be compiled using
@* rvct
@*
@* @author
@* yogeswaran rs/ pathiban
@*
@* @par list of functions:
@*
@*
@* @remarks
@* none
@*
@*******************************************************************************
@*/
@/**
@/**
@*******************************************************************************
@*
@* @brief
@* interprediction chroma filter to store vertical 16bit ouput
@*
@* @par description:
@* applies a vertical filter with coefficients pointed to by 'pi1_coeff' to
@* the elements pointed by 'pu1_src' and writes to the location pointed by
@* 'pu1_dst' no downshifting or clipping is done and the output is used as
@* an input for weighted prediction assumptions : the function is optimized
@* considering the fact width is multiple of 2,4 or 8. and also considering
@* height should be multiple of 2. width 4,8 is optimized further
@*
@* @param[in] pu1_src
@* uword8 pointer to the source
@*
@* @param[out] pi2_dst
@* word16 pointer to the destination
@*
@* @param[in] src_strd
@* integer source stride
@*
@* @param[in] dst_strd
@* integer destination stride
@*
@* @param[in] pi1_coeff
@* word8 pointer to the filter coefficients
@*
@* @param[in] ht
@* integer height of the array
@*
@* @param[in] wd
@* integer width of the array
@*
@* @returns
@*
@* @remarks
@* none
@*
@*****************************************************************************
@*/
@void ihevc_inter_pred_chroma_vert_w16out(uword8 *pu1_src,
@ word16 *pi2_dst,
@ word32 src_strd,
@ word32 dst_strd,
@ word8 *pi1_coeff,
@ word32 ht,
@ word32 wd)
@**************variables vs registers*****************************************
@r0 => *pu1_src
@r1 => *pi2_dst
@r2 => src_strd
@r3 => dst_strd
.text
.align 4
.globl ihevc_inter_pred_chroma_vert_w16out_a9q
.type ihevc_inter_pred_chroma_vert_w16out_a9q, %function
ihevc_inter_pred_chroma_vert_w16out_a9q:
stmfd sp!,{r4-r12,r14} @stack stores the values of the arguments
ldr r4,[sp,#44] @loads ht
ldr r12,[sp,#40] @loads pi1_coeff
cmp r4,#0 @checks ht == 0
ldr r6,[sp,#48] @loads wd
sub r0,r0,r2 @pu1_src - src_strd
vld1.8 {d0},[r12] @loads pi1_coeff
ble end_loops @jumps to end
tst r6,#3 @checks (wd & 3)
vabs.s8 d3,d0 @vabs_s8(coeff)
lsl r10,r6,#1 @2*wd
vdup.8 d0,d3[0] @coeffabs_0
vdup.8 d1,d3[1] @coeffabs_1
vdup.8 d2,d3[2] @coeffabs_2
vdup.8 d3,d3[3] @coeffabs_3
bgt outer_loop_wd_2 @jumps to loop handling wd ==2
tst r4,#7 @checks ht for mul of 8
beq core_loop_ht_8 @when height is multiple of 8
lsl r7,r3,#2 @2*dst_strd
sub r9,r7,r10,lsl #1 @4*dst_strd - 4wd
lsl r12,r2,#1 @2*src_strd
sub r8,r12,r10 @2*src_strd - 2wd
mov r3,r3,lsl #1
mov r5,r10 @2wd
inner_loop_ht_2: @called when wd is multiple of 4 and ht is 4,2
add r6,r0,r2 @pu1_src +src_strd
vld1.8 {d9},[r6],r2 @loads pu1_src
subs r5,r5,#8 @2wd - 8
vld1.8 {d5},[r0]! @loads src
vmull.u8 q3,d9,d1 @vmull_u8(vreinterpret_u8_u32(src_tmp2), coeffabs_1)
vld1.8 {d4},[r6],r2 @loads incremented src
vmlsl.u8 q3,d5,d0 @vmlsl_u8(mul_res1, vreinterpret_u8_u32(src_tmp1), coeffabs_0)
vld1.8 {d8},[r6],r2 @loads incremented src
vmlal.u8 q3,d4,d2 @vmlal_u8(mul_res1, vreinterpret_u8_u32(src_tmp3), coeffabs_2)
vmull.u8 q2,d4,d1
vld1.8 {d10},[r6] @loads the incremented src
vmlsl.u8 q3,d8,d3
vmlsl.u8 q2,d9,d0
vmlal.u8 q2,d8,d2
vmlsl.u8 q2,d10,d3
add r6,r1,r3 @pu1_dst + dst_strd
vst1.8 {q3},[r1]! @stores the loaded value
vst1.8 {q2},[r6] @stores the loaded value
bgt inner_loop_ht_2 @inner loop again
subs r4,r4,#2 @ht - 2
add r1,r1,r9 @pu1_dst += (2*dst_strd - 2wd)
mov r5,r10 @2wd
add r0,r0,r8 @pu1_src += (2*src_strd - 2wd)
bgt inner_loop_ht_2 @loop again
b end_loops @jumps to end
outer_loop_wd_2: @called when width is multiple of 2
lsl r5,r3,#2 @2*dst_strd
mov r12,r10 @2wd
sub r9,r5,r10,lsl #1 @4*dst_strd - 4wd
lsl r7,r2,#1 @2*src_strd
sub r8,r7,r10 @2*src_strd - 2wd
inner_loop_wd_2:
add r6,r0,r2 @pu1_src + src_strd
vld1.32 {d6[0]},[r0] @vld1_lane_u32((uint32_t *)pu1_src_tmp, src_tmp1, 0
subs r12,r12,#4 @2wd - 4
add r0,r0,#4 @pu1_src + 4
vld1.32 {d6[1]},[r6],r2 @loads pu1_src_tmp
vdup.32 d7,d6[1]
vld1.32 {d7[1]},[r6],r2 @loads pu1_src_tmp
vmull.u8 q2,d7,d1 @vmull_u8(vreinterpret_u8_u32(src_tmp2), coeffabs_1)
vdup.32 d7,d7[1]
vld1.32 {d7[1]},[r6],r2
vmlsl.u8 q2,d6,d0
vmlal.u8 q2,d7,d2
vdup.32 d7,d7[1]
vld1.32 {d7[1]},[r6]
add r6,r1,r3,lsl #1 @pu1_dst + dst_strd
vmlsl.u8 q2,d7,d3
vst1.32 {d4},[r1] @stores the loaded value
add r1,r1,#8 @pu1_dst += 4
vst1.32 {d5},[r6] @stores the loaded value
bgt inner_loop_wd_2 @inner loop again
@inner loop ends
subs r4,r4,#2 @ht - 2
add r1,r1,r9 @pu1_dst += 2*dst_strd - 2*wd
mov r12,r10 @2wd
add r0,r0,r8 @pu1_src += 2*src_strd - 2*wd
bgt inner_loop_wd_2 @loop again
b end_loops @jumps to end
core_loop_ht_8: @when wd & ht is multiple of 8
lsl r12,r3,#3 @4*dst_strd
sub r8,r12,r10,lsl #1 @4*dst_strd - 2wd
lsl r12,r2,#2 @4*src_strd
sub r9,r12,r10 @4*src_strd - 2wd
bic r5,r10,#7 @r5 ->wd
mov r14,r10,lsr #3 @divide by 8
mul r12,r4,r14 @multiply height by width
sub r12,#4 @subtract by one for epilog
mov r3,r3,lsl #1
prolog:
add r6,r0,r2 @pu1_src + src_strd
vld1.8 {d5},[r6],r2 @loads pu1_src
subs r5,r5,#8 @2wd - 8
vld1.8 {d4},[r0]! @loads the source
vld1.8 {d6},[r6],r2 @load and increment
vmull.u8 q15,d5,d1 @mul with coeff 1
vld1.8 {d7},[r6],r2 @load and increment
vmlsl.u8 q15,d4,d0
add r7,r1,r3 @pu1_dst
vmlal.u8 q15,d6,d2
vmlsl.u8 q15,d7,d3
vld1.8 {d8},[r6],r2 @load and increment
vmull.u8 q14,d6,d1 @mul_res 2
addle r0,r0,r9 @pu1_dst += 4*dst_strd - 2*wd
vmlsl.u8 q14,d5,d0
bicle r5,r10,#7 @r5 ->wd
vmlal.u8 q14,d7,d2
vld1.8 {d9},[r6],r2
vmlsl.u8 q14,d8,d3
vld1.8 {d10},[r6],r2
vmull.u8 q13,d7,d1
add r6,r0,r2 @pu1_src + src_strd
vmlsl.u8 q13,d6,d0
vst1.8 {q15},[r1]! @stores the loaded value
vmlal.u8 q13,d8,d2
vld1.8 {d4},[r0]! @loads the source
vmlsl.u8 q13,d9,d3
addle r1,r1,r8 @pu1_src += 4*src_strd - 2*wd
vmull.u8 q12,d8,d1
vld1.8 {d5},[r6],r2 @loads pu1_src
vmlsl.u8 q12,d7,d0
subs r12,r12,#4
vld1.8 {d6},[r6],r2 @load and increment
vmlal.u8 q12,d9,d2
vld1.8 {d7},[r6],r2 @load and increment
vmlsl.u8 q12,d10,d3
rsb r11,r2,r2,lsl #3
add r14,r2,r2,lsl #1
add r14,r14,r11
vst1.8 {q14},[r7],r3 @stores the loaded value
ble epilog @jumps to epilog
kernel_8:
vmull.u8 q15,d5,d1 @mul with coeff 1
subs r5,r5,#8 @2wd - 8
vmlsl.u8 q15,d4,d0
addle r0,r0,r9 @pu1_dst += 4*dst_strd - 2*wd
vmlal.u8 q15,d6,d2
rsble r11,r2,r2,lsl #3
vmlsl.u8 q15,d7,d3
vst1.8 {q13},[r7],r3 @stores the loaded value
vld1.8 {d8},[r6],r2 @load and increment
vmull.u8 q14,d6,d1 @mul_res 2
bicle r5,r10,#7 @r5 ->wd
vmlsl.u8 q14,d5,d0
vst1.8 {q12},[r7],r3 @stores the loaded value
vmlal.u8 q14,d7,d2
vld1.8 {d9},[r6],r2
vmlsl.u8 q14,d8,d3
vld1.8 {d10},[r6],r2
add r7,r1,r3 @pu1_dst
vmull.u8 q13,d7,d1
add r6,r0,r2 @pu1_src + src_strd
pld [r0,r11]
vmlsl.u8 q13,d6,d0
vld1.8 {d4},[r0]! @loads the source
add r11,r11,r2
vmlal.u8 q13,d8,d2
vst1.8 {q15},[r1]! @stores the loaded value
vmlsl.u8 q13,d9,d3
vld1.8 {d5},[r6],r2 @loads pu1_src
vmull.u8 q12,d8,d1
vld1.8 {d6},[r6],r2 @load and increment
addle r1,r1,r8 @pu1_src += 4*src_strd - 2*wd
cmp r11,r14
rsbgt r11,r2,r2,lsl #3
vmlsl.u8 q12,d7,d0
subs r12,r12,#4
vmlal.u8 q12,d9,d2
vld1.8 {d7},[r6],r2 @load and increment
vmlsl.u8 q12,d10,d3
vst1.8 {q14},[r7],r3 @stores the loaded value
bgt kernel_8 @jumps to kernel_8
epilog:
vmull.u8 q15,d5,d1 @mul with coeff 1
vmlsl.u8 q15,d4,d0
vmlal.u8 q15,d6,d2
vmlsl.u8 q15,d7,d3
vst1.8 {q13},[r7],r3 @stores the loaded value
vld1.8 {d8},[r6],r2 @load and increment
vmull.u8 q14,d6,d1 @mul_res 2
vmlsl.u8 q14,d5,d0
vmlal.u8 q14,d7,d2
vmlsl.u8 q14,d8,d3
vst1.8 {q12},[r7],r3 @stores the loaded value
vld1.8 {d9},[r6],r2
vmull.u8 q13,d7,d1
add r7,r1,r3 @pu1_dst
vmlsl.u8 q13,d6,d0
vst1.8 {q15},[r1]! @stores the loaded value
vmlal.u8 q13,d8,d2
vld1.8 {d10},[r6],r2
vmlsl.u8 q13,d9,d3
vmull.u8 q12,d8,d1
vst1.8 {q14},[r7],r3 @stores the loaded value
vmlsl.u8 q12,d7,d0
vmlal.u8 q12,d9,d2
vst1.8 {q13},[r7],r3 @stores the loaded value
vmlsl.u8 q12,d10,d3
vst1.8 {q12},[r7],r3 @stores the loaded value
end_loops:
ldmfd sp!,{r4-r12,r15} @reload the registers from sp