mirror of
https://github.com/cemu-project/Cemu.git
synced 2024-12-27 10:11:52 +01:00
1093 lines
51 KiB
ArmAsm
1093 lines
51 KiB
ArmAsm
|
@/******************************************************************************
|
||
|
@ *
|
||
|
@ * Copyright (C) 2015 The Android Open Source Project
|
||
|
@ *
|
||
|
@ * 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.
|
||
|
@ *
|
||
|
@ *****************************************************************************
|
||
|
@ * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
|
||
|
@*/
|
||
|
@/*****************************************************************************/
|
||
|
@/* */
|
||
|
@/* File Name : ih264_deblk_luma_a9.s */
|
||
|
@/* */
|
||
|
@/* Description : Contains function definitions for deblocking luma */
|
||
|
@/* edge. Functions are coded in NEON assembly and can */
|
||
|
@/* be compiled using ARM RVDS. */
|
||
|
@/* */
|
||
|
@/* List of Functions : ih264_deblk_luma_vert_bs4_a9() */
|
||
|
@/* ih264_deblk_luma_vert_bslt4_a9() */
|
||
|
@/* ih264_deblk_luma_horz_bs4_a9() */
|
||
|
@/* ih264_deblk_luma_horz_bslt4_a9() */
|
||
|
@/* ih264_deblk_luma_vert_bs4_mbaff_a9() */
|
||
|
@/* ih264_deblk_luma_vert_bslt4_mbaff_a9() */
|
||
|
@/* */
|
||
|
@/* Issues / Problems : None */
|
||
|
@/* */
|
||
|
@/* Revision History : */
|
||
|
@/* */
|
||
|
@/* DD MM YYYY Author(s) Changes (Describe the changes made) */
|
||
|
@/* 28 11 2013 Ittiam Draft */
|
||
|
@/* 05 01 2015 Kaushik Added double-call functions for */
|
||
|
@/* Senthoor vertical deblocking. */
|
||
|
@/* */
|
||
|
@/*****************************************************************************/
|
||
|
|
||
|
|
||
|
.text
|
||
|
.p2align 2
|
||
|
|
||
|
@**
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
@* @brief
|
||
|
@* Performs filtering of a luma block horizontal edge for cases where the
|
||
|
@* boundary strength is less than 4
|
||
|
@*
|
||
|
@* @par Description:
|
||
|
@* This operation is described in Sec. 8.7.2.4 under the title
|
||
|
@* "Filtering process for edges for bS equal to 4" in ITU T Rec H.264.
|
||
|
@*
|
||
|
@* @param[in] r0 - pu1_src
|
||
|
@* Pointer to the src sample q0
|
||
|
@*
|
||
|
@* @param[in] r1 - src_strd
|
||
|
@* Source stride
|
||
|
@*
|
||
|
@* @param[in] r2 - alpha
|
||
|
@* Alpha Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] r3 - beta
|
||
|
@* Beta Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] sp(0) - u4_bs
|
||
|
@* Packed Boundary strength array
|
||
|
@*
|
||
|
@* @param[in] sp(4) - pu1_cliptab
|
||
|
@* tc0_table
|
||
|
@*
|
||
|
@* @returns
|
||
|
@* None
|
||
|
@*
|
||
|
@* @remarks
|
||
|
@* None
|
||
|
@*
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
|
||
|
.global ih264_deblk_luma_horz_bslt4_a9
|
||
|
|
||
|
ih264_deblk_luma_horz_bslt4_a9:
|
||
|
|
||
|
stmfd sp!, {r4-r7, lr}
|
||
|
|
||
|
ldrd r4, r5, [sp, #0x14] @r4 = ui_Bs , r5 = *puc_ClpTab
|
||
|
vpush {d8 - d15}
|
||
|
sub r0, r0, r1, lsl #1 @R1 = uc_Horizonpad
|
||
|
sub r0, r0, r1 @r0 pointer to p2
|
||
|
rev r4, r4 @
|
||
|
vld1.8 {q5}, [r0], r1 @p2 values are loaded into q5
|
||
|
vmov.32 d12[0], r4 @d12[0] = ui_Bs
|
||
|
mov r6, r0 @keeping backup of pointer to p1
|
||
|
vld1.8 {q4}, [r0], r1 @p1 values are loaded into q4
|
||
|
mov r7, r0 @keeping backup of pointer to p0
|
||
|
vld1.8 {q3}, [r0], r1 @p0 values are loaded into q3
|
||
|
vmovl.u8 q6, d12 @q6 = uc_Bs in each 16 bt scalar
|
||
|
vld1.8 {q0}, [r0], r1 @q0 values are loaded into q0
|
||
|
vabd.u8 q13, q4, q3 @Q13 = ABS(p1 - p0)
|
||
|
vld1.8 {q1}, [r0], r1 @q1 values are loaded into q1
|
||
|
vabd.u8 q11, q3, q0 @Q11 = ABS(p0 - q0)
|
||
|
vld1.32 d16[0], [r5] @D16[0] contains cliptab
|
||
|
vabd.u8 q12, q1, q0 @Q12 = ABS(q1 - q0)
|
||
|
vld1.8 {q2}, [r0], r1 @q2 values are loaded into q2
|
||
|
vtbl.8 d14, {d16}, d12 @
|
||
|
vdup.8 q10, r2 @Q10 contains alpha
|
||
|
vdup.8 q8, r3 @Q8 contains beta
|
||
|
vmovl.u16 q6, d12 @
|
||
|
vmovl.u16 q7, d14 @
|
||
|
vabd.u8 q14, q5, q3 @Q14 = Ap = ABS(p2 - p0)
|
||
|
vabd.u8 q15, q2, q0 @Q15 = Aq = ABS(q2 - q0)
|
||
|
vcgt.s32 q6, q6, #0 @Q6 = (us_Bs > 0)
|
||
|
vsli.32 q7, q7, #8 @
|
||
|
vcge.u8 q9, q11, q10 @Q9 = ( ABS(p0 - q0) >= Alpha )
|
||
|
vcge.u8 q12, q12, q8 @Q12=( ABS(q1 - q0) >= Beta )
|
||
|
vcge.u8 q13, q13, q8 @Q13=( ABS(p1 - p0) >= Beta )
|
||
|
vcgt.u8 q10, q8, q14 @Q10=(Ap<Beta)
|
||
|
vcgt.u8 q11, q8, q15 @Q11=(Aq<Beta)
|
||
|
vsli.32 q7, q7, #16 @Q7 = C0
|
||
|
vorr q9, q9, q12 @Q9 = ( ABS(p0 - q0) >= Alpha ) | ( ABS(q1 - q0) >= Beta )
|
||
|
vsubl.u8 q15, d1, d7 @
|
||
|
vsubl.u8 q12, d0, d6 @Q15,Q12 = (q0 - p0)
|
||
|
vorr q9, q9, q13 @Q9 = ( ABS(p0 - q0) >= Alpha ) | ( ABS(q1 - q0) >= Beta ) | ( ABS(p1 - p0) >= Beta )
|
||
|
vsubl.u8 q14, d8, d2 @Q14 = (p1 - q1)L
|
||
|
vshl.i16 q13, q15, #2 @Q13 = (q0 - p0)<<2
|
||
|
vshl.i16 q12, q12, #2 @Q12 = (q0 - p0)<<2
|
||
|
vsubl.u8 q15, d9, d3 @Q15 = (p1 - q1)H
|
||
|
vbic q6, q6, q9 @final condition
|
||
|
vadd.i16 q12, q12, q14 @
|
||
|
vadd.i16 q13, q13, q15 @Q13,Q12 = [ (q0 - p0)<<2 ] + (p1 - q1)
|
||
|
vsub.i8 q9, q7, q10 @Q9 = C0 + (Ap < Beta)
|
||
|
vrhadd.u8 q8, q3, q0 @Q8 = ((p0+q0+1) >> 1)
|
||
|
vqrshrn.s16 d24, q12, #3 @
|
||
|
vqrshrn.s16 d25, q13, #3 @Q12 = i_macro = (((q0 - p0)<<2) + (p1 - q1) + 4)>>3
|
||
|
vsub.i8 q9, q9, q11 @Q9 = C0 + (Ap < Beta) + (Aq < Beta)
|
||
|
vand.i8 q10, q10, q6 @
|
||
|
vand.i8 q11, q11, q6 @
|
||
|
vabs.s8 q13, q12 @Q13 = ABS (i_macro)
|
||
|
vaddl.u8 q14, d17, d11 @
|
||
|
vaddl.u8 q5, d16, d10 @Q14,Q5 = p2 + (p0+q0+1)>>1
|
||
|
vaddl.u8 q15, d17, d5 @
|
||
|
vmin.u8 q9, q13, q9 @Q9 = delta = (ABS(i_macro) > C) ? C : ABS(i_macro)
|
||
|
vshll.u8 q13, d9, #1 @
|
||
|
vaddl.u8 q2, d16, d4 @Q15,Q2 = q2 + (p0+q0+1)>>1
|
||
|
vshll.u8 q8, d8, #1 @Q13,Q8 = (p1<<1)
|
||
|
vand q9, q9, q6 @Making delta zero in places where values shouldn be filterd
|
||
|
vsub.i16 q14, q14, q13 @Q14,Q5 = [p2 + (p0+q0+1)>>1] - (p1<<1)
|
||
|
vsub.i16 q5, q5, q8 @
|
||
|
vshll.u8 q8, d2, #1 @
|
||
|
vshll.u8 q13, d3, #1 @Q13,Q8 = (q1<<1)
|
||
|
vqshrn.s16 d29, q14, #1 @
|
||
|
vqshrn.s16 d28, q5, #1 @Q14 = i_macro_p1
|
||
|
vsub.i16 q2, q2, q8 @
|
||
|
vsub.i16 q15, q15, q13 @Q15,Q2 = [q2 + (p0+q0+1)>>1] - (q1<<1)
|
||
|
vneg.s8 q13, q7 @Q13 = -C0
|
||
|
vmin.s8 q14, q14, q7 @Q14 = min(C0,i_macro_p1)
|
||
|
vcge.s8 q12, q12, #0 @Q12 = (i_macro >= 0)
|
||
|
vqshrn.s16 d31, q15, #1 @
|
||
|
vqshrn.s16 d30, q2, #1 @Q15 = i_macro_q1
|
||
|
vmax.s8 q14, q14, q13 @Q14 = max( - C0 , min(C0, i_macro_p1) )
|
||
|
vqadd.u8 q8, q3, q9 @Q8 = p0 + delta
|
||
|
vqsub.u8 q3, q3, q9 @Q3 = p0 - delta
|
||
|
vmin.s8 q15, q15, q7 @Q15 = min(C0,i_macro_q1)
|
||
|
vand.i8 q14, q10, q14 @condition check Ap<beta
|
||
|
vqadd.u8 q7, q0, q9 @Q7 = q0 + delta
|
||
|
vqsub.u8 q0, q0, q9 @Q0 = q0 - delta
|
||
|
vmax.s8 q15, q15, q13 @Q15 = max( - C0 , min(C0, i_macro_q1) )
|
||
|
vbif q8, q3, q12 @Q8 = (i_macro >= 0 ) ? (p0+delta) : (p0-delta)
|
||
|
vbif q0, q7, q12 @Q0 = (i_macro >= 0 ) ? (q0-delta) : (q0+delta)
|
||
|
vadd.i8 q14, q14, q4 @
|
||
|
vand.i8 q15, q11, q15 @condition check Aq<beta
|
||
|
vst1.8 {q8}, [r7], r1 @writting back filtered value of p0
|
||
|
vadd.i8 q15, q15, q1 @
|
||
|
vst1.8 {q0}, [r7], r1 @writting back filtered value of q0
|
||
|
vst1.8 {q14}, [r6] @writting back filtered value of p1
|
||
|
vst1.8 {q15}, [r7], r1 @writting back filtered value of q1
|
||
|
vpop {d8 - d15}
|
||
|
ldmfd sp!, {r4-r7, pc}
|
||
|
|
||
|
|
||
|
|
||
|
@**
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
@* @brief
|
||
|
@* Performs filtering of a luma block horizontal edge when the
|
||
|
@* boundary strength is set to 4
|
||
|
@*
|
||
|
@* @par Description:
|
||
|
@* This operation is described in Sec. 8.7.2.4 under the title
|
||
|
@* "Filtering process for edges for bS equal to 4" in ITU T Rec H.264.
|
||
|
@*
|
||
|
@* @param[in] r0 - pu1_src
|
||
|
@* Pointer to the src sample q0
|
||
|
@*
|
||
|
@* @param[in] r1 - src_strd
|
||
|
@* Source stride
|
||
|
@*
|
||
|
@* @param[in] r2 - alpha
|
||
|
@* Alpha Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] r3 - beta
|
||
|
@* Beta Value for the boundary
|
||
|
@*
|
||
|
@* @returns
|
||
|
@* None
|
||
|
@*
|
||
|
@* @remarks
|
||
|
@* None
|
||
|
@*
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
|
||
|
.global ih264_deblk_luma_horz_bs4_a9
|
||
|
|
||
|
ih264_deblk_luma_horz_bs4_a9:
|
||
|
|
||
|
@ Back up necessary registers on stack
|
||
|
stmfd sp!, {r12, r14}
|
||
|
vpush {d8 - d15}
|
||
|
@ Init
|
||
|
vdup.8 q0, r2 @duplicate alpha
|
||
|
sub r12, r0, r1 @pointer to p0 = q0 - src_strd
|
||
|
vdup.8 q1, r3 @duplicate beta
|
||
|
sub r14, r0, r1, lsl#1 @pointer to p1 = q0 - src_strd*2
|
||
|
sub r2, r0, r1, lsl#2 @pointer to p3 = q0 - src_strd*4
|
||
|
sub r3, r14, r1 @pointer to p2 = p1 - src_strd
|
||
|
|
||
|
@ Load Data
|
||
|
vld1.8 {d4, d5}, [r0], r1 @load q0 to Q2, q0 = q0 + src_strd
|
||
|
vld1.8 {d6, d7}, [r12] @load p0 to Q3
|
||
|
vld1.8 {d8, d9}, [r0], r1 @load q1 to Q4, q0 = q0 + src_strd
|
||
|
vld1.8 {d10, d11}, [r14] @load p1 to Q5
|
||
|
|
||
|
@ Filter Decision
|
||
|
vabd.u8 q6, q2, q3 @ABS(p0 - q0)
|
||
|
vabd.u8 q7, q4, q2 @ABS(q1 - q0)
|
||
|
vabd.u8 q8, q5, q3 @ABS(p1 - p0)
|
||
|
vcge.u8 q9, q6, q0 @ABS(p0 - q0) >= Alpha
|
||
|
vcge.u8 q7, q7, q1 @ABS(q1 - q0) >= Beta
|
||
|
vcge.u8 q8, q8, q1 @ABS(p1 - p0) >= Beta
|
||
|
vmov.i8 q10, #2
|
||
|
vorr q9, q9, q7 @ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta
|
||
|
vld1.8 {d14, d15}, [r0], r1 @load q2 to Q7, q0 = q0 + src_strd
|
||
|
vorr q9, q9, q8 @ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta
|
||
|
vsra.u8 q10, q0, #2 @((Alpha >> 2) + 2)
|
||
|
vabd.u8 q11, q7, q2 @Aq = ABS(q2 - q0)
|
||
|
vaddl.u8 q12, d4, d6 @p0+q0 L
|
||
|
vaddl.u8 q13, d5, d7 @p0+q0 H
|
||
|
vclt.u8 q11, q11, q1 @Aq < Beta
|
||
|
vclt.u8 q10, q6, q10 @(ABS(p0 - q0) <((Alpha >>2) + 2))
|
||
|
|
||
|
@ Deblock Filtering q0', q1', q2'
|
||
|
vaddw.u8 q14, q12, d8 @p0+q0+q1 L
|
||
|
vaddw.u8 q15, q13, d9 @p0+q0+q1 H
|
||
|
vand q11, q11, q10 @(Aq < Beta && ABS(p0 - q0) <((Alpha >>2) + 2))
|
||
|
@ q0' if (Aq < Beta && ABS(p0 - q0) <((Alpha >>2) + 2)) TRUE
|
||
|
vadd.i16 q8, q14, q14 @2*(p0+q0+q1)L
|
||
|
vadd.i16 q0, q15, q15 @2*(p0+q0+q1)H
|
||
|
vaddw.u8 q8, q8, d14 @2*(p0+q0+q1)+q2 L
|
||
|
vaddw.u8 q0, q0, d15 @2*(p0+q0+q1)+q2 H
|
||
|
vaddw.u8 q8, q8, d10 @2*(p0+q0+q1)+q2 +p1 L
|
||
|
vaddw.u8 q0, q0, d11 @2*(p0+q0+q1)+q2 +p1 H
|
||
|
vrshrn.u16 d12, q8, #3 @(2*(p0+q0+q1)+q2 +p1 +4)>> 3 L [q0']
|
||
|
vrshrn.u16 d13, q0, #3 @(2*(p0+q0+q1)+q2 +p1 +4)>> 3 H [q0']
|
||
|
@ q0" if (Aq < Beta && ABS(p0 - q0) <((Alpha >>2) + 2)) FALSE
|
||
|
vaddl.u8 q8, d8, d8 @2*q1 L
|
||
|
vaddl.u8 q0, d9, d9 @2*q1 H
|
||
|
vaddw.u8 q8, q8, d4 @2*q1+q0 L
|
||
|
vaddw.u8 q0, q0, d5 @2*q1+q0 H
|
||
|
vaddw.u8 q8, q8, d10 @2*q1+q0+p1 L
|
||
|
vaddw.u8 q0, q0, d11 @2*q1+q0+p1 H
|
||
|
vrshrn.u16 d16, q8, #2 @(2*q1+q0+p1+2)>>2 L [q0"]
|
||
|
vrshrn.u16 d17, q0, #2 @(2*q1+q0+p1+2)>>2 H [q0"]
|
||
|
@ q1'
|
||
|
vaddw.u8 q14, q14, d14 @p0+q0+q1+q2 L
|
||
|
vaddw.u8 q15, q15, d15 @p0+q0+q1+q2 H
|
||
|
vld1.8 {q0}, [r0], r1 @load q3 to Q0, q0 = q0 + src_strd
|
||
|
vbit q8, q6, q11 @choosing between q0' and q0" depending on condn
|
||
|
sub r0, r0, r1, lsl #2 @pointer to q0
|
||
|
vbic q11, q11, q9 @((ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta))
|
||
|
@ && (Aq < Beta && ABS(p0 - q0) <((Alpha >>2) + 2))
|
||
|
vrshrn.u16 d12, q14, #2 @(p0+q0+q1+q2+2)>>2 L [q1']
|
||
|
vrshrn.u16 d13, q15, #2 @(p0+q0+q1+q2+2)>>2 H [q1']
|
||
|
vbif q2, q8, q9 @choose q0 or filtered q0
|
||
|
@ q2'
|
||
|
vaddl.u8 q8, d14, d0 @q2+q3,L
|
||
|
vaddl.u8 q0, d15, d1 @q2+q3,H
|
||
|
vadd.i16 q14, q14, q8 @p0+q0+q1+2*q2+q3 L
|
||
|
vst1.8 {d4, d5}, [r0], r1 @store q0
|
||
|
vadd.i16 q15, q15, q0 @p0+q0+q1+2*q2+q3 H
|
||
|
vadd.i16 q14, q14, q8 @p0+q0+q1+3*q2+2*q3 L
|
||
|
vadd.i16 q15, q15, q0 @p0+q0+q1+3*q2+2*q3 H
|
||
|
vrshrn.u16 d0, q14, #3 @(p0+q0+q1+3*q2+2*q3+4)>>3 L [q2']
|
||
|
vrshrn.u16 d1, q15, #3 @(p0+q0+q1+3*q2+2*q3+4)>>3 H [q2']
|
||
|
vld1.8 {d30, d31}, [r3] @load p2 to Q15
|
||
|
vbif q6, q4, q11 @choose q1 or filtered value of q1
|
||
|
|
||
|
vabd.u8 q8, q15, q3 @Ap,ABS(p2 - p0)
|
||
|
vaddw.u8 q12, q12, d10 @p0+q0+p1 L
|
||
|
vbif q0, q7, q11 @choose q2 or filtered q2
|
||
|
vaddw.u8 q13, q13, d11 @p0+q0+p1 H
|
||
|
vst1.8 {d12, d13}, [r0], r1 @store q1
|
||
|
vclt.u8 q8, q8, q1 @Ap < Beta
|
||
|
vadd.i16 q14, q12, q12 @2*(p0+q0+p1) L
|
||
|
vadd.i16 q2, q13, q13 @2*(p0+q0+p1) H
|
||
|
vst1.8 {d0, d1}, [r0], r1 @store q2
|
||
|
vand q10, q10, q8 @((Ap < Beta) && (ABS(p0 - q0) <((Alpha >>2) + 2)))
|
||
|
vaddw.u8 q14, q14, d30 @2*(p0+q0+p1)+p2 l
|
||
|
vaddw.u8 q2, q2, d31 @2*(p0+q0+p1)+p2 H
|
||
|
vaddw.u8 q14, q14, d8 @2*(p0+q0+p1)+p2+q1 L
|
||
|
vaddw.u8 q2, q2, d9 @2*(p0+q0+p1)+p2+q1 H
|
||
|
vrshrn.u16 d28, q14, #3 @(2*(p0+q0+p1)+p2+q1+4)>>3 L,p0'
|
||
|
vrshrn.u16 d29, q2, #3 @(2*(p0+q0+p1)+p2+q1+4)>>3 H,p0'
|
||
|
vmov.i8 d0, #2
|
||
|
vmov.i16 d1, #2
|
||
|
vaddl.u8 q1, d6, d8 @p0+q1 L
|
||
|
vmlal.u8 q1, d10, d0 @2*p1+p0+q1 L
|
||
|
vaddl.u8 q8, d7, d9 @p0+q1 H
|
||
|
vmlal.u8 q8, d11, d0 @2*p1+p0+q1 H
|
||
|
vaddw.u8 q6, q12, d30 @(p0+q0+p1) +p2 L
|
||
|
vld1.8 {d24, d25}, [r2] @load p3,Q12
|
||
|
vaddw.u8 q2, q13, d31 @(p0+q0+p1) +p2 H
|
||
|
vaddl.u8 q4, d30, d24 @p2+p3 L
|
||
|
vrshrn.u16 d26, q6, #2 @((p0+q0+p1)+p2 +2)>>2,p1' L
|
||
|
vrshrn.u16 d2, q1, #2 @(2*p1+p0+q1+2)>>2,p0"L
|
||
|
vrshrn.u16 d27, q2, #2 @((p0+q0+p1)+p2 +2)>>2,p1' H
|
||
|
vrshrn.u16 d3, q8, #2 @(2*p1+p0+q1+2)>>2,p0" H
|
||
|
vaddl.u8 q8, d31, d25 @p2+p3 H
|
||
|
vmla.u16 q6, q4, d1[0] @(p0+q0+p1)+3*p2+2*p3 L
|
||
|
vmla.u16 q2, q8, d1[0] @(p0+q0+p1)+3*p2+2*p3 H
|
||
|
vbic q8, q10, q9 @((ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta))
|
||
|
@&& (Ap < Beta && ABS(p0 - q0) <((Alpha >>2) + 2))
|
||
|
vbit q1, q14, q10 @choosing between po' and p0"
|
||
|
vrshrn.u16 d12, q6, #3 @((p0+q0+p1)+3*p2+2*p3+4)>>3 L p2'
|
||
|
vrshrn.u16 d13, q2, #3 @((p0+q0+p1)+3*p2+2*p3+4)>>3 H p2'
|
||
|
vbif q3, q1, q9 @choosing between p0 and filtered value of p0
|
||
|
vbit q5, q13, q8 @choosing between p1 and p1'
|
||
|
vbit q15, q6, q8 @choosing between p2 and p2'
|
||
|
vst1.8 {d6, d7}, [r12] @store p0
|
||
|
vst1.8 {d10, d11}, [r14] @store p1
|
||
|
vst1.8 {d30, d31}, [r3] @store p2
|
||
|
vpop {d8 - d15}
|
||
|
ldmfd sp!, {r12, pc}
|
||
|
|
||
|
|
||
|
|
||
|
@**
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
@* @brief
|
||
|
@* Performs filtering of a luma block vertical edge for cases where the
|
||
|
@* boundary strength is less than 4
|
||
|
@*
|
||
|
@* @par Description:
|
||
|
@* This operation is described in Sec. 8.7.2.4 under the title
|
||
|
@* "Filtering process for edges for bS equal to 4" in ITU T Rec H.264.
|
||
|
@*
|
||
|
@* @param[in] r0 - pu1_src
|
||
|
@* Pointer to the src sample q0
|
||
|
@*
|
||
|
@* @param[in] r1 - src_strd
|
||
|
@* Source stride
|
||
|
@*
|
||
|
@* @param[in] r2 - alpha
|
||
|
@* Alpha Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] r3 - beta
|
||
|
@* Beta Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] sp(0) - u4_bs
|
||
|
@* Packed Boundary strength array
|
||
|
@*
|
||
|
@* @param[in] sp(4) - pu1_cliptab
|
||
|
@* tc0_table
|
||
|
@*
|
||
|
@* @returns
|
||
|
@* None
|
||
|
@*
|
||
|
@* @remarks
|
||
|
@* None
|
||
|
@*
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
|
||
|
.global ih264_deblk_luma_vert_bslt4_a9
|
||
|
|
||
|
ih264_deblk_luma_vert_bslt4_a9:
|
||
|
|
||
|
stmfd sp!, {r12, lr}
|
||
|
|
||
|
sub r0, r0, #4 @pointer uc_edgePixel-4
|
||
|
ldr r12, [sp, #8] @r12 = ui_Bs
|
||
|
ldr r14, [sp, #12] @r14 = *puc_ClpTab
|
||
|
vpush {d8 - d15}
|
||
|
@loading p3:p2:p1:p0:q0:q1:q2:q3 for every row
|
||
|
vld1.8 {d0}, [r0], r1 @row1
|
||
|
vld1.8 d2, [r0], r1 @row2
|
||
|
vld1.8 d4, [r0], r1 @row3
|
||
|
rev r12, r12 @reversing ui_bs
|
||
|
vld1.8 d6, [r0], r1 @row4
|
||
|
vmov.32 d18[0], r12 @d12[0] = ui_Bs
|
||
|
vld1.32 d16[0], [r14] @D16[0] contains cliptab
|
||
|
vld1.8 d8, [r0], r1 @row5
|
||
|
vmovl.u8 q9, d18 @q6 = uc_Bs in each 16 bt scalar
|
||
|
vld1.8 d10, [r0], r1 @row6
|
||
|
vld1.8 d12, [r0], r1 @row7
|
||
|
vtbl.8 d16, {d16}, d18 @puc_ClipTab[uc_Bs]
|
||
|
vld1.8 d14, [r0], r1 @row8
|
||
|
vld1.8 d1, [r0], r1 @row9
|
||
|
vmovl.u16 q8, d16 @
|
||
|
vld1.8 d3, [r0], r1 @row10
|
||
|
vld1.8 d5, [r0], r1 @row11
|
||
|
vld1.8 d7, [r0], r1 @row12
|
||
|
vsli.32 q8, q8, #8 @
|
||
|
vld1.8 d9, [r0], r1 @row13
|
||
|
vld1.8 d11, [r0], r1 @row14
|
||
|
vld1.8 d13, [r0], r1 @row15
|
||
|
vsli.32 q8, q8, #16 @Q8 = C0
|
||
|
vld1.8 d15, [r0], r1 @row16
|
||
|
|
||
|
@taking two 8x8 transposes
|
||
|
@2X2 transposes
|
||
|
vtrn.8 d0, d2 @row1 &2
|
||
|
vtrn.8 d4, d6 @row3&row4
|
||
|
vtrn.8 d8, d10 @row5&6
|
||
|
vtrn.8 d12, d14 @row7 & 8
|
||
|
vtrn.8 d1, d3 @row9 &10
|
||
|
vtrn.8 d5, d7 @row11 & 12
|
||
|
vtrn.8 d9, d11 @row13 &14
|
||
|
vtrn.8 d13, d15 @row15 & 16
|
||
|
@4x4 transposes
|
||
|
vtrn.16 d2, d6 @row2 & row4
|
||
|
vtrn.16 d10, d14 @row6 & row8
|
||
|
vtrn.16 d3, d7 @row10 & 12
|
||
|
vtrn.16 d11, d15 @row14 & row16
|
||
|
vtrn.32 d6, d14 @row4 & 8
|
||
|
vtrn.32 d7, d15 @row 12 & 16
|
||
|
|
||
|
@now Q3 ->p0 and Q7->q3
|
||
|
vtrn.16 d0, d4 @row1 & 3
|
||
|
vtrn.16 d8, d12 @row 5 & 7
|
||
|
vtrn.16 d1, d5 @row9 & row11
|
||
|
vtrn.16 d9, d13 @row13 & row15
|
||
|
vtrn.32 d0, d8 @row1 & row5
|
||
|
vtrn.32 d1, d9 @row9 & 13
|
||
|
|
||
|
@now Q0->p3 & Q4->q0
|
||
|
@starting processing as p0 and q0 are now ready
|
||
|
vtrn.32 d2, d10 @row2 &6
|
||
|
vrhadd.u8 q10, q3, q4 @((p0 + q0 + 1) >> 1)
|
||
|
vtrn.32 d3, d11 @row10&row14
|
||
|
vmov.i8 d19, #2
|
||
|
@now Q1->p2 & Q5->q1
|
||
|
vtrn.32 d4, d12 @row3 & 7
|
||
|
vabd.u8 q11, q3, q4 @ABS(p0 - q0)
|
||
|
vtrn.32 d5, d13 @row11 & row15
|
||
|
vaddl.u8 q12, d20, d2 @(p2 + ((p0 + q0 + 1) >> 1) L
|
||
|
@now Q2->p1,Q6->q2
|
||
|
vaddl.u8 q13, d21, d3 @(p2 + ((p0 + q0 + 1) >> 1) H
|
||
|
vmlsl.u8 q12, d4, d19 @(p2 + ((p0 + q0 + 1) >> 1) - (p1 << 1)) L
|
||
|
vmlsl.u8 q13, d5, d19 @(p2 + ((p0 + q0 + 1) >> 1) - (p1 << 1)) H
|
||
|
vdup.8 q14, r2 @alpha
|
||
|
vcle.u8 q11, q14, q11 @ABS(p0 - q0) >= Alpha(Alpha <=ABS(p0 - q0))
|
||
|
vdup.i8 q14, r3 @beta
|
||
|
vabd.u8 q15, q5, q4 @ABS(q1 - q0)
|
||
|
vqshrn.s16 d24, q12, #1 @((p2 + ((p0 + q0 + 1) >> 1) - (p1 << 1)) >> 1) L
|
||
|
vqshrn.s16 d25 , q13, #1 @((p2 + ((p0 + q0 + 1) >> 1) - (p1 << 1)) >> 1) H
|
||
|
vcge.u8 q15, q15, q14 @ABS(q1 - q0) >= Beta
|
||
|
vabd.u8 q13, q2, q3 @ABS(p1 - p0)
|
||
|
vmin.s8 q12, q12, q8 @min(deltap1 ,C0)
|
||
|
vorr q11, q11, q15 @ABS(q1 - q0) >= Beta ||ABS(p0 - q0) >= Alpha
|
||
|
vneg.s8 q15, q8 @-C0
|
||
|
vcge.u8 q13, q13, q14 @ABS(p1 - p0) >= Beta
|
||
|
vmax.s8 q12, q12, q15 @max(deltap1,-C0)
|
||
|
vorr q11, q11, q13 @ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta)
|
||
|
vmovl.u16 q13, d18 @ui_bs
|
||
|
vaddl.u8 q9, d20, d12 @q2 + ((p0 + q0 + 1) >> 1) L
|
||
|
vceq.u32 q13, q13, #0 @ui_bs == 0
|
||
|
vsubw.u8 q9, q9, d10 @(q2 + ((p0 + q0 + 1) >> 1) - q1) L
|
||
|
vaddl.u8 q10, d21, d13 @q2 + ((p0 + q0 + 1) >> 1) H
|
||
|
vsubw.u8 q9, q9, d10 @(q2 + ((p0 + q0 + 1) >> 1) - 2*q1)L
|
||
|
vsubw.u8 q10, q10, d11 @(q2 + ((p0 + q0 + 1) >> 1) - q1) H
|
||
|
vorr q13, q13, q11 @(ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta)) &&(ui_bs)
|
||
|
vsubw.u8 q10, q10, d11 @(q2 + ((p0 + q0 + 1) >> 1) - 2*q1) H
|
||
|
vqshrn.s16 d18, q9, #1 @((q2 + ((p0 + q0 + 1) >> 1) - (q1 << 1)) >> 1) L
|
||
|
vabd.u8 q11, q1, q3 @Ap = ABS(p2 - p0)
|
||
|
vqshrn.s16 d19, q10, #1 @((q2 + ((p0 + q0 + 1) >> 1) - (q1 << 1)) >> 1) H
|
||
|
vabd.u8 q10, q6, q4 @Aq= ABS(q2 - q0)
|
||
|
vclt.u8 q11, q11, q14 @Ap < Beta
|
||
|
vmin.s8 q9, q9, q8 @min(delatq1,C0)
|
||
|
vclt.u8 q10, q10, q14 @Aq <Beta
|
||
|
vsubl.u8 q14, d8, d6 @(q0 - p0) L
|
||
|
vmax.s8 q9, q9, q15 @max(deltaq1,-C0)
|
||
|
vsubl.u8 q15, d9, d7 @(q0 - p0) H
|
||
|
vshl.s16 q14, q14, #2 @(q0 - p0)<<2 L
|
||
|
vsub.u8 q8, q8, q11 @C0 + (Ap < Beta)
|
||
|
vshl.s16 q15, q15, #2 @(q0 - p0) << 2) H
|
||
|
vaddw.u8 q14, q14, d4 @((q0 - p0) << 2) + (p1 L
|
||
|
vaddw.u8 q15, q15, d5 @((q0 - p0) << 2) + (p1 H
|
||
|
vsubw.u8 q14, q14, d10 @((q0 - p0) << 2) + (p1 - q1) L
|
||
|
vsubw.u8 q15, q15, d11 @((q0 - p0) << 2) + (p1 - q1) H
|
||
|
vbic q11, q11, q13 @final condition for p1
|
||
|
vrshrn.s16 d28, q14, #3 @delta = ((((q0 - p0) << 2) + (p1 - q1) + 4) >> 3); L
|
||
|
vrshrn.s16 d29, q15, #3 @delta = ((((q0 - p0) << 2) + (p1 - q1) + 4) >> 3) H
|
||
|
vsub.u8 q8, q8, q10 @C0 + (Ap < Beta) + (Aq < Beta)
|
||
|
vbic q10, q10, q13 @final condition for q1
|
||
|
vabs.s8 q15, q14 @abs(delta)
|
||
|
vand q12, q12, q11 @delatp1
|
||
|
vand q9, q9, q10 @delta q1
|
||
|
vmin.u8 q15, q15, q8 @min((abs(delta),C)
|
||
|
vadd.i8 q2, q2, q12 @p1+deltap1
|
||
|
vadd.i8 q5, q5, q9 @q1+deltaq1
|
||
|
vbic q15, q15, q13 @abs(delta) of pixels to be changed only
|
||
|
vcge.s8 q14, q14, #0 @sign(delta)
|
||
|
vqsub.u8 q11, q3, q15 @clip(p0-delta)
|
||
|
vtrn.8 d0, d2 @row1 &2
|
||
|
vqadd.u8 q3, q3, q15 @clip(p0+delta)
|
||
|
vtrn.8 d1, d3 @row9 &10
|
||
|
vqadd.u8 q12, q4, q15 @clip(q0+delta)
|
||
|
vtrn.8 d12, d14 @row7 & 8
|
||
|
vqsub.u8 q4, q4, q15 @clip(q0-delta)
|
||
|
vtrn.8 d13, d15 @row15 & 16
|
||
|
vbif q3, q11, q14 @p0
|
||
|
vbif q4, q12, q14 @q0
|
||
|
vtrn.8 d4, d6 @row3&row4
|
||
|
vtrn.8 d8, d10 @row5&6
|
||
|
vtrn.8 d5, d7 @row11 & 12
|
||
|
vtrn.8 d9, d11 @row13 &14
|
||
|
vtrn.16 d2, d6 @row2 & row4
|
||
|
vtrn.16 d10, d14 @row6 & row8
|
||
|
vtrn.16 d3, d7 @row10 & 12
|
||
|
vtrn.16 d11, d15 @row14 & row16
|
||
|
vtrn.32 d6, d14 @row4 & 8
|
||
|
vtrn.32 d7, d15 @row 12 & 16
|
||
|
@now Q3 ->p0 and Q7->q3
|
||
|
vtrn.16 d0, d4 @row1 & 3
|
||
|
vtrn.16 d8, d12 @row 5 & 7
|
||
|
vtrn.16 d1, d5 @row9 & row11
|
||
|
vtrn.16 d9, d13 @row13 & row15
|
||
|
sub r0, r0, r1, lsl#4 @restore pointer
|
||
|
vtrn.32 d0, d8 @row1 & row5
|
||
|
vtrn.32 d1, d9 @row9 & 13
|
||
|
vtrn.32 d2, d10 @row2 &6
|
||
|
vtrn.32 d3, d11 @row10&row14
|
||
|
vtrn.32 d4, d12 @row3 & 7
|
||
|
vtrn.32 d5, d13 @row11 & row15
|
||
|
vst1.8 {d0}, [r0], r1 @row1
|
||
|
vst1.8 d2, [r0], r1 @row2
|
||
|
vst1.8 d4, [r0], r1 @row3
|
||
|
vst1.8 d6, [r0], r1 @row4
|
||
|
vst1.8 d8, [r0], r1 @row5
|
||
|
vst1.8 d10, [r0], r1 @row6
|
||
|
vst1.8 d12, [r0], r1 @row7
|
||
|
vst1.8 d14, [r0], r1 @row8
|
||
|
vst1.8 d1, [r0], r1 @row9
|
||
|
vst1.8 d3, [r0], r1 @row10
|
||
|
vst1.8 d5, [r0], r1 @row11
|
||
|
vst1.8 d7, [r0], r1 @row12
|
||
|
vst1.8 d9, [r0], r1 @row13
|
||
|
vst1.8 d11, [r0], r1 @row14
|
||
|
vst1.8 d13, [r0], r1 @row15
|
||
|
vst1.8 d15, [r0], r1 @row16
|
||
|
vpop {d8 - d15}
|
||
|
ldmfd sp!, {r12, pc}
|
||
|
|
||
|
|
||
|
|
||
|
@**
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
@* @brief
|
||
|
@* Performs filtering of a luma block vertical edge when the
|
||
|
@* boundary strength is set to 4
|
||
|
@*
|
||
|
@* @par Description:
|
||
|
@* This operation is described in Sec. 8.7.2.4 under the title
|
||
|
@* "Filtering process for edges for bS equal to 4" in ITU T Rec H.264.
|
||
|
@*
|
||
|
@* @param[in] r0 - pu1_src
|
||
|
@* Pointer to the src sample q0
|
||
|
@*
|
||
|
@* @param[in] r1 - src_strd
|
||
|
@* Source stride
|
||
|
@*
|
||
|
@* @param[in] r2 - alpha
|
||
|
@* Alpha Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] r3 - beta
|
||
|
@* Beta Value for the boundary
|
||
|
@*
|
||
|
@* @returns
|
||
|
@* None
|
||
|
@*
|
||
|
@* @remarks
|
||
|
@* None
|
||
|
@*
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
|
||
|
.global ih264_deblk_luma_vert_bs4_a9
|
||
|
|
||
|
ih264_deblk_luma_vert_bs4_a9:
|
||
|
|
||
|
stmfd sp!, {r12, lr}
|
||
|
vpush {d8 - d15}
|
||
|
sub r0, r0, #4 @pointer uc_edgePixel-4
|
||
|
@loading p3:p2:p1:p0:q0:q1:q2:q3 for every row
|
||
|
vld1.8 d0, [r0], r1 @row1
|
||
|
vld1.8 d2, [r0], r1 @row2
|
||
|
vld1.8 d4, [r0], r1 @row3
|
||
|
vld1.8 d6, [r0], r1 @row4
|
||
|
vld1.8 d8, [r0], r1 @row5
|
||
|
vld1.8 d10, [r0], r1 @row6
|
||
|
vld1.8 d12, [r0], r1 @row7
|
||
|
vld1.8 d14, [r0], r1 @row8
|
||
|
vld1.8 d1, [r0], r1 @row9
|
||
|
vld1.8 d3, [r0], r1 @row10
|
||
|
vld1.8 d5, [r0], r1 @row11
|
||
|
vld1.8 d7, [r0], r1 @row12
|
||
|
vld1.8 d9, [r0], r1 @row13
|
||
|
vld1.8 d11, [r0], r1 @row14
|
||
|
vld1.8 d13, [r0], r1 @row15
|
||
|
vld1.8 d15, [r0], r1 @row16
|
||
|
@taking two 8x8 transposes
|
||
|
@2X2 transposes
|
||
|
vtrn.8 d0, d2 @row1 &2
|
||
|
vtrn.8 d4, d6 @row3&row4
|
||
|
vtrn.8 d8, d10 @row5&6
|
||
|
vtrn.8 d12, d14 @row7 & 8
|
||
|
vtrn.8 d1, d3 @row9 &10
|
||
|
vtrn.8 d5, d7 @row11 & 12
|
||
|
vtrn.8 d9, d11 @row13 &14
|
||
|
vtrn.8 d13, d15 @row15 & 16
|
||
|
@4x4 transposes
|
||
|
vtrn.16 d2, d6 @row2 & row4
|
||
|
vtrn.16 d10, d14 @row6 & row8
|
||
|
vtrn.16 d3, d7 @row10 & 12
|
||
|
vtrn.16 d11, d15 @row14 & row16
|
||
|
vtrn.32 d6, d14 @row4 & 8
|
||
|
vtrn.32 d7, d15 @row 12 & 16
|
||
|
@now Q3 ->p0 and Q7->q3
|
||
|
vtrn.16 d0, d4 @row1 & 3
|
||
|
vtrn.16 d8, d12 @row 5 & 7
|
||
|
vtrn.16 d1, d5 @row9 & row11
|
||
|
vtrn.16 d9, d13 @row13 & row15
|
||
|
vtrn.32 d0, d8 @row1 & row5
|
||
|
vtrn.32 d1, d9 @row9 & 13
|
||
|
@now Q0->p3 & Q4->q0
|
||
|
@starting processing as p0 and q0 are now ready
|
||
|
@now Q1->p2 & Q5->q1
|
||
|
vpush {q7} @saving in stack
|
||
|
vtrn.32 d4, d12 @row3 & 7
|
||
|
vmov.i16 q14, #2
|
||
|
vtrn.32 d5, d13 @row11 & row15
|
||
|
vaddl.u8 q8, d6, d8 @p0+q0 L
|
||
|
vtrn.32 d2, d10 @row2 &6
|
||
|
vaddl.u8 q9, d7, d9 @p0+q0 H
|
||
|
vtrn.32 d3, d11 @row10&row14
|
||
|
vaddw.u8 q10, q8, d4 @p0+q0+p1 L
|
||
|
vaddw.u8 q11, q9, d5 @p0+q0+p1 H
|
||
|
vaddl.u8 q12, d2, d10 @p2+q1 L
|
||
|
vaddl.u8 q13, d3, d11 @p2+q1 H
|
||
|
vmla.u16 q12, q10, q14 @p2 + X2(p1) + X2(p0) + X2(q0) + q1 L
|
||
|
vmla.u16 q13, q11, q14 @p2 + X2(p1) + X2(p0) + X2(q0) + q1 H
|
||
|
vmov.i8 q14, #2
|
||
|
vaddw.u8 q8, q10, d2 @p0+q0+p1+p2 L
|
||
|
vaddw.u8 q9, q11, d3 @p0+q0+p1+p2 H
|
||
|
vdup.i8 q15, r2 @duplicate alpha
|
||
|
vrshrn.u16 d20, q8, #2 @(p2 + p1 + p0 + q0 + 2) >> 2)L p1'
|
||
|
vrshrn.u16 d21, q9, #2 @(p2 + p1 + p0 + q0 + 2) >> 2)H p1'
|
||
|
vabd.u8 q11, q3, q4 @ABD(p0-q0)
|
||
|
vsra.u8 q14, q15, #2 @alpha >>2 +2
|
||
|
vabd.u8 q15, q1, q3 @Ap = ABD(p2-p0)
|
||
|
vrshrn.u16 d24, q12, #3 @((p2 + X2(p1) + X2(p0) + X2(q0) + q1 + 4) >> 3) L p0'
|
||
|
vrshrn.u16 d25, q13, #3 @((p2 + X2(p1) + X2(p0) + X2(q0) + q1 + 4) >> 3) H p0'
|
||
|
vdup.i8 q13, r3 @beta
|
||
|
vcgt.u8 q14, q14, q11 @ABS(p0 - q0) <((Alpha >>2) + 2)
|
||
|
vaddl.u8 q11, d6, d10 @p0+q1 L
|
||
|
vcgt.u8 q7, q13, q15 @beta>Ap
|
||
|
vaddl.u8 q15, d7, d11 @p0+q1 H
|
||
|
vaddw.u8 q11, q11, d4 @p0+q1+p1 L
|
||
|
vaddw.u8 q15, q15, d5 @p0+q1+p1 H
|
||
|
vaddw.u8 q11, q11, d4 @p0+q1+2*p1 L
|
||
|
vaddw.u8 q15, q15, d5 @p0+q1+2*p1 H
|
||
|
vand q7, q7, q14 @(Ap < Beta && ABS(p0 - q0) <((Alpha >>2) + 2)
|
||
|
vrshrn.u16 d22, q11, #2 @((X2(p1) + p0 + q1 + 2) >> 2) L p0"
|
||
|
vrshrn.u16 d23, q15, #2 @((X2(p1) + p0 + q1 + 2) >> 2) H p0"
|
||
|
vaddl.u8 q15, d2, d0 @p2+p3 L
|
||
|
vbif q12, q11, q7 @p0' or p0 "
|
||
|
vaddl.u8 q11, d3, d1 @p2+p3 H
|
||
|
vadd.u16 q15, q15, q15 @2*(p2+p3) L
|
||
|
vadd.u16 q11, q11, q11 @2*(p2+p3)H
|
||
|
vadd.u16 q8, q8, q15 @(X2(p3) + X3(p2) + p1 + p0 + q0) L
|
||
|
vadd.u16 q9, q9, q11 @(X2(p3) + X3(p2) + p1 + p0 + q0) H
|
||
|
vabd.u8 q15, q6, q4 @Aq = abs(q2-q0)
|
||
|
vabd.u8 q11, q5, q4 @ABS(Q1-Q0)
|
||
|
vrshrn.u16 d16, q8, #3 @((X2(p3) + X3(p2) + p1 + p0 + q0 + 4) >> 3); L p2'
|
||
|
vrshrn.u16 d17, q9, #3 @((X2(p3) + X3(p2) + p1 + p0 + q0 + 4) >> 3); H p2'
|
||
|
vabd.u8 q9, q2, q3 @ABS(p1-p0)
|
||
|
vcgt.u8 q15, q13, q15 @Aq < Beta
|
||
|
vcge.u8 q11, q11, q13 @ABS(q1 - q0) >= Beta
|
||
|
vcge.u8 q9, q9, q13 @ABS(p1 - p0) >= beta
|
||
|
vdup.i8 q13, r2 @duplicate alpha
|
||
|
vand q15, q15, q14 @(Aq < Beta && ABS(p0 - q0) <((Alpha >>2) + 2))
|
||
|
vabd.u8 q14, q3, q4 @abs(p0-q0)
|
||
|
vorr q11, q11, q9 @ABS(p1 - p0) >= Beta || ABS(q1 - q0) >= Beta
|
||
|
vaddl.u8 q9, d6, d8 @p0+q0 L
|
||
|
vcge.u8 q14, q14, q13 @ABS(p0 - q0) >= Alpha
|
||
|
vaddl.u8 q13, d7, d9 @p0+q0 H
|
||
|
vaddw.u8 q9, q9, d10 @p0+q0+q1 L
|
||
|
vorr q11, q11, q14 @ABS(p1 - p0) >= Beta || ABS(q1 - q0) >= Beta||ABS(p0 - q0) >= Alpha
|
||
|
vaddw.u8 q13, q13, d11 @p0+q0+q1 H
|
||
|
vbic q7, q7, q11 @final condn for p's
|
||
|
vmov.i8 q14, #2
|
||
|
vbif q3, q12, q11 @final p0
|
||
|
vbit q1, q8, q7 @final p2
|
||
|
vbif q10, q2, q7 @final p1
|
||
|
vaddl.u8 q12, d8, d4 @q0+p1 L
|
||
|
vmlal.u8 q12, d10, d28 @X2(q1) + q0 + p1 L
|
||
|
vaddl.u8 q8, d9, d5 @q0+p1 H
|
||
|
vmlal.u8 q8, d11, d28 @X2(q1) + q0 + p1 H
|
||
|
vmov.i16 q14, #2
|
||
|
vaddl.u8 q7, d4, d12 @p1+q2 L
|
||
|
vmla.u16 q7, q9, q14 @p1 + X2(p0) + X2(q0) + X2(q1) + q2L
|
||
|
vaddl.u8 q2, d5, d13 @p1+q2H
|
||
|
vmla.u16 q2, q13, q14 @p1 + X2(p0) + X2(q0) + X2(q1) + q2H
|
||
|
vrshrn.u16 d24, q12, #2 @(X2(q1) + q0 + p1 + 2) >> 2; L q0'
|
||
|
vrshrn.u16 d25, q8, #2 @(X2(q1) + q0 + p1 + 2) >> 2; H q0'
|
||
|
vaddw.u8 q9, q9, d12 @p0 + q0 + q1 + q2 L
|
||
|
vaddw.u8 q13, q13, d13 @p0 + q0 + q1 + q2 H
|
||
|
vrshrn.u16 d16, q7, #3 @(p1 + X2(p0) + X2(q0) + X2(q1) + q2 + 4) >> 3 L qo"
|
||
|
vpop {q7}
|
||
|
vrshrn.u16 d17, q2, #3 @(p1 + X2(p0) + X2(q0) + X2(q1) + q2 + 4) >> 3 H qo"
|
||
|
vrshrn.u16 d4, q9, #2 @p0 + q0 + q1 + q2 + 2)>>2 L q1'
|
||
|
vrshrn.u16 d5, q13, #2 @p0 + q0 + q1 + q2 + 2)>>2 H q1'
|
||
|
vbit q12, q8, q15 @q0' or q0"
|
||
|
vbic q15, q15, q11 @final condn for q's
|
||
|
vtrn.8 d0, d2 @row1 &2
|
||
|
vbit q5, q2, q15 @final q1
|
||
|
vtrn.8 d1, d3 @row9 &10
|
||
|
vaddl.u8 q8, d12, d14 @q2+q3 L
|
||
|
vtrn.8 d20, d6 @row3&row4
|
||
|
vaddl.u8 q2, d13, d15 @q2+q3 H
|
||
|
vtrn.8 d21, d7 @row11 & 12
|
||
|
vmla.u16 q9, q8, q14 @X2(q3) + X3(q2) + q1 + q0 + p0 L
|
||
|
vtrn.16 d2, d6 @row2 & row4
|
||
|
vmla.u16 q13, q2, q14 @X2(q3) + X3(q2) + q1 + q0 + p0 H
|
||
|
vtrn.16 d3, d7 @row10 & 12
|
||
|
vbif q4, q12, q11 @final q0
|
||
|
vtrn.16 d0, d20 @row1 & 3
|
||
|
vrshrn.u16 d18, q9, #3 @(X2(q3) + X3(q2) + q1 + q0 + p0 + 4) >> 3; L
|
||
|
vtrn.16 d1, d21 @row9 & row11
|
||
|
vrshrn.u16 d19, q13, #3 @(X2(q3) + X3(q2) + q1 + q0 + p0 + 4) >> 3; H
|
||
|
vtrn.8 d8, d10 @row5&6
|
||
|
vbit q6, q9, q15 @final q2
|
||
|
vtrn.8 d9, d11 @row13 &14
|
||
|
vtrn.8 d12, d14 @row7 & 8
|
||
|
vtrn.8 d13, d15 @row15 & 16
|
||
|
vtrn.16 d10, d14 @row6 & row8
|
||
|
vtrn.16 d11, d15 @row14 & row16
|
||
|
@now Q3 ->p0 and Q7->q3
|
||
|
vtrn.16 d8, d12 @row 5 & 7
|
||
|
vtrn.16 d9, d13 @row13 & row15
|
||
|
sub r0, r0, r1, lsl#4 @restore pointer
|
||
|
vtrn.32 d6, d14 @row4 & 8
|
||
|
vtrn.32 d7, d15 @row 12 & 16
|
||
|
vtrn.32 d0, d8 @row1 & row5
|
||
|
vtrn.32 d1, d9 @row9 & 13
|
||
|
vtrn.32 d2, d10 @row2 &6
|
||
|
vtrn.32 d3, d11 @row10&row14
|
||
|
vtrn.32 d20, d12 @row3 & 7
|
||
|
vtrn.32 d21, d13 @row11 & row15
|
||
|
vst1.8 d0, [r0], r1 @row1
|
||
|
vst1.8 d2, [r0], r1 @row2
|
||
|
vst1.8 d20, [r0], r1 @row3
|
||
|
vst1.8 d6, [r0], r1 @row4
|
||
|
vst1.8 d8, [r0], r1 @row5
|
||
|
vst1.8 d10, [r0], r1 @row6
|
||
|
vst1.8 d12, [r0], r1 @row7
|
||
|
vst1.8 d14, [r0], r1 @row8
|
||
|
vst1.8 d1, [r0], r1 @row9
|
||
|
vst1.8 d3, [r0], r1 @row10
|
||
|
vst1.8 d21, [r0], r1 @row11
|
||
|
vst1.8 d7, [r0], r1 @row12
|
||
|
vst1.8 d9, [r0], r1 @row13
|
||
|
vst1.8 d11, [r0], r1 @row14
|
||
|
vst1.8 d13, [r0], r1 @row15
|
||
|
vst1.8 d15, [r0], r1 @row16
|
||
|
vpop {d8 - d15}
|
||
|
ldmfd sp!, {r12, pc}
|
||
|
|
||
|
|
||
|
|
||
|
@**
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
@* @brief
|
||
|
@* Performs filtering of a luma block vertical edge when the
|
||
|
@* boundary strength is set to 4 on calling twice
|
||
|
@*
|
||
|
@* @par Description:
|
||
|
@* This operation is described in Sec. 8.7.2.4 under the title
|
||
|
@* "Filtering process for edges for bS equal to 4" in ITU T Rec H.264.
|
||
|
@*
|
||
|
@* @param[in] r0 - pu1_src
|
||
|
@* Pointer to the src sample q0
|
||
|
@*
|
||
|
@* @param[in] r1 - src_strd
|
||
|
@* Source stride
|
||
|
@*
|
||
|
@* @param[in] r2 - alpha
|
||
|
@* Alpha Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] r3 - beta
|
||
|
@* Beta Value for the boundary
|
||
|
@*
|
||
|
@* @returns
|
||
|
@* None
|
||
|
@*
|
||
|
@* @remarks
|
||
|
@* None
|
||
|
@*
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
|
||
|
.global ih264_deblk_luma_vert_bs4_mbaff_a9
|
||
|
|
||
|
ih264_deblk_luma_vert_bs4_mbaff_a9:
|
||
|
|
||
|
stmfd sp!, {lr}
|
||
|
|
||
|
sub r0, r0, #4 @pointer uc_edgePixel-4
|
||
|
vpush {d8 - d15}
|
||
|
@loading [p3:p2],[p1:p0]:[q0:q1]:[q2:q3] for every row
|
||
|
vld4.16 {d0[0], d2[0], d4[0], d6[0]}, [r0], r1
|
||
|
vld4.16 {d0[1], d2[1], d4[1], d6[1]}, [r0], r1
|
||
|
vld4.16 {d0[2], d2[2], d4[2], d6[2]}, [r0], r1
|
||
|
vld4.16 {d0[3], d2[3], d4[3], d6[3]}, [r0], r1
|
||
|
vld4.16 {d1[0], d3[0], d5[0], d7[0]}, [r0], r1
|
||
|
vld4.16 {d1[1], d3[1], d5[1], d7[1]}, [r0], r1
|
||
|
vld4.16 {d1[2], d3[2], d5[2], d7[2]}, [r0], r1
|
||
|
vld4.16 {d1[3], d3[3], d5[3], d7[3]}, [r0], r1
|
||
|
|
||
|
vuzp.8 d0, d1 @D0->p3, D1->p2
|
||
|
vuzp.8 d2, d3 @D2->p1, D3->p0
|
||
|
vuzp.8 d4, d5 @D4->q0, D5->q1
|
||
|
vuzp.8 d6, d7 @D6->q2, D7->q3
|
||
|
|
||
|
vmov.i16 q14, #2
|
||
|
vaddl.u8 q4, d3, d4 @p0+q0
|
||
|
vaddw.u8 q5, q4, d2 @p0+q0+p1
|
||
|
vaddl.u8 q6, d1, d5 @p2+q1
|
||
|
vmla.u16 q6, q5, q14 @p2 + X2(p1) + X2(p0) + X2(q0) + q1
|
||
|
|
||
|
vmov.i8 d14, #2
|
||
|
vaddw.u8 q4, q5, d1 @p0+q0+p1+p2
|
||
|
vdup.i8 d15, r2 @duplicate alpha
|
||
|
vrshrn.u16 d10, q4, #2 @(p2 + p1 + p0 + q0 + 2) >> 2) p1'
|
||
|
vabd.u8 d11, d3, d4 @ABD(p0-q0)
|
||
|
vsra.u8 d14, d15, #2 @alpha >>2 +2
|
||
|
vabd.u8 d15, d1, d3 @Ap = ABD(p2-p0)
|
||
|
vrshrn.u16 d12, q6, #3 @((p2 + X2(p1) + X2(p0) + X2(q0) + q1 + 4) >> 3) p0'
|
||
|
vdup.i8 d13, r3 @beta
|
||
|
vcgt.u8 d14, d14, d11 @ABS(p0 - q0) <((Alpha >>2) + 2)
|
||
|
vaddl.u8 q8, d3, d5 @p0+q1
|
||
|
vcgt.u8 d26, d13, d15 @beta>Ap
|
||
|
vaddw.u8 q8, q8, d2 @p0+q1+p1
|
||
|
vaddw.u8 q8, q8, d2 @p0+q1+2*p1
|
||
|
vand d26, d26, d14 @(Ap < Beta && ABS(p0 - q0) <((Alpha >>2) + 2)
|
||
|
vrshrn.u16 d11, q8, #2 @((X2(p1) + p0 + q1 + 2) >> 2) p0"
|
||
|
vbif d12, d11, d26 @p0' or p0 "
|
||
|
vaddl.u8 q9, d1, d0 @p2+p3
|
||
|
vadd.u16 q9, q9, q9 @2*(p2+p3)
|
||
|
vadd.u16 q4, q4, q9 @(X2(p3) + X3(p2) + p1 + p0 + q0)
|
||
|
vabd.u8 d15, d6, d4 @Aq = abs(q2-q0)
|
||
|
vabd.u8 d11, d5, d4 @ABS(q1-q0)
|
||
|
vrshrn.u16 d8, q4, #3 @((X2(p3) + X3(p2) + p1 + p0 + q0 + 4) >> 3); p2'
|
||
|
vabd.u8 d9, d2, d3 @ABS(p1-p0)
|
||
|
vcgt.u8 d15, d13, d15 @Aq < Beta
|
||
|
vcge.u8 d11, d11, d13 @ABS(q1 - q0) >= Beta
|
||
|
vcge.u8 d9, d9, d13 @ABS(p1 - p0) >= beta
|
||
|
vdup.i8 d13, r2 @duplicate alpha
|
||
|
vand d15, d15, d14 @(Aq < Beta && ABS(p0 - q0) <((Alpha >>2) + 2))
|
||
|
vabd.u8 d14, d3, d4 @abs(p0-q0)
|
||
|
vorr d11, d11, d9 @ABS(p1 - p0) >= Beta || ABS(q1 - q0) >= Beta
|
||
|
vcge.u8 d14, d14, d13 @ABS(p0 - q0) >= Alpha
|
||
|
vaddl.u8 q10, d3, d4 @p0+q0
|
||
|
vorr d11, d11, d14 @ABS(p1 - p0) >= Beta || ABS(q1 - q0) >= Beta||ABS(p0 - q0) >= Alpha
|
||
|
vaddw.u8 q10, q10, d5 @p0+q0+q1
|
||
|
vbic d26, d26, d11 @final condn for p's
|
||
|
vmov.i8 d14, #2
|
||
|
vbif d3, d12, d11 @final p0
|
||
|
vbit d1, d8, d26 @final p2
|
||
|
vbif d10, d2, d26 @final p1
|
||
|
vaddl.u8 q6, d4, d2 @q0+p1
|
||
|
vmlal.u8 q6, d5, d14 @X2(q1) + q0 + p1
|
||
|
|
||
|
vaddl.u8 q11, d2, d6 @p1+q2
|
||
|
vmla.u16 q11, q10, q14 @p1 + X2(p0) + X2(q0) + X2(q1) + q2
|
||
|
vrshrn.u16 d12, q6, #2 @(X2(q1) + q0 + p1 + 2) >> 2; q0'
|
||
|
vaddw.u8 q10, q10, d6 @p0 + q0 + q1 + q2
|
||
|
vrshrn.u16 d8, q11, #3 @(p1 + X2(p0) + X2(q0) + X2(q1) + q2 + 4) >> 3 qo"
|
||
|
|
||
|
vrshrn.u16 d2, q10, #2 @p0 + q0 + q1 + q2 + 2)>>2 q1'
|
||
|
vbit d12, d8, d15 @q0' or q0"
|
||
|
vbic d15, d15, d11 @final condn for q's
|
||
|
vbit d5, d2, d15 @final q1
|
||
|
vaddl.u8 q12, d6, d7 @q2+q3
|
||
|
vmla.u16 q10, q12, q14 @X2(q3) + X3(q2) + q1 + q0 + p0
|
||
|
vbif d4, d12, d11 @final q0
|
||
|
vrshrn.u16 d9, q10, #3 @(X2(q3) + X3(q2) + q1 + q0 + p0 + 4) >> 3;
|
||
|
vbit d6, d9, d15 @final q2
|
||
|
vand d2, d10, d10 @D0->p3, D1->p2, D2->p1, D3->p0, D4->q0, D5->q1, D6->q2, D7->q3
|
||
|
|
||
|
vzip.8 d0, d1 @D0,D1 -> [p3:p2]
|
||
|
vzip.8 d2, d3 @D2,D3 -> [p1:p0]
|
||
|
vzip.8 d4, d5 @D4,D5 -> [q0:q1]
|
||
|
vzip.8 d6, d7 @D6,D7 -> [q2:q3]
|
||
|
|
||
|
sub r0, r0, r1, lsl#3 @restore pointer
|
||
|
|
||
|
@storing [p3:p2],[p1:p0]:[q0:q1]:[q2:q3] in every row
|
||
|
vst4.16 {d0[0], d2[0], d4[0], d6[0]}, [r0], r1
|
||
|
vst4.16 {d0[1], d2[1], d4[1], d6[1]}, [r0], r1
|
||
|
vst4.16 {d0[2], d2[2], d4[2], d6[2]}, [r0], r1
|
||
|
vst4.16 {d0[3], d2[3], d4[3], d6[3]}, [r0], r1
|
||
|
vst4.16 {d1[0], d3[0], d5[0], d7[0]}, [r0], r1
|
||
|
vst4.16 {d1[1], d3[1], d5[1], d7[1]}, [r0], r1
|
||
|
vst4.16 {d1[2], d3[2], d5[2], d7[2]}, [r0], r1
|
||
|
vst4.16 {d1[3], d3[3], d5[3], d7[3]}, [r0], r1
|
||
|
vpop {d8 - d15}
|
||
|
ldmfd sp!, {pc}
|
||
|
|
||
|
|
||
|
|
||
|
@**
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
@* @brief
|
||
|
@* Performs filtering of a luma block vertical edge for cases where the
|
||
|
@* boundary strength is less than 4 on calling twice
|
||
|
@*
|
||
|
@* @par Description:
|
||
|
@* This operation is described in Sec. 8.7.2.4 under the title
|
||
|
@* "Filtering process for edges for bS equal to 4" in ITU T Rec H.264.
|
||
|
@*
|
||
|
@* @param[in] r0 - pu1_src
|
||
|
@* Pointer to the src sample q0
|
||
|
@*
|
||
|
@* @param[in] r1 - src_strd
|
||
|
@* Source stride
|
||
|
@*
|
||
|
@* @param[in] r2 - alpha
|
||
|
@* Alpha Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] r3 - beta
|
||
|
@* Beta Value for the boundary
|
||
|
@*
|
||
|
@* @param[in] sp(0) - u4_bs
|
||
|
@* Packed Boundary strength array
|
||
|
@*
|
||
|
@* @param[in] sp(4) - pu1_cliptab
|
||
|
@* tc0_table
|
||
|
@*
|
||
|
@* @returns
|
||
|
@* None
|
||
|
@*
|
||
|
@* @remarks
|
||
|
@* None
|
||
|
@*
|
||
|
@*******************************************************************************
|
||
|
@*
|
||
|
|
||
|
.global ih264_deblk_luma_vert_bslt4_mbaff_a9
|
||
|
|
||
|
ih264_deblk_luma_vert_bslt4_mbaff_a9:
|
||
|
|
||
|
stmfd sp!, {r12, lr}
|
||
|
|
||
|
sub r0, r0, #4 @pointer uc_edgePixel-4
|
||
|
ldr r12, [sp, #8] @r12 = ui_Bs
|
||
|
ldr r14, [sp, #12] @r14 = pu1_ClipTab
|
||
|
vpush {d8 - d15}
|
||
|
@loading [p3:p2],[p1:p0]:[q0:q1]:[q2:q3] for every row
|
||
|
vld4.16 {d0[0], d2[0], d4[0], d6[0]}, [r0], r1
|
||
|
vld4.16 {d0[1], d2[1], d4[1], d6[1]}, [r0], r1
|
||
|
vld4.16 {d0[2], d2[2], d4[2], d6[2]}, [r0], r1
|
||
|
vld4.16 {d0[3], d2[3], d4[3], d6[3]}, [r0], r1
|
||
|
vld4.16 {d1[0], d3[0], d5[0], d7[0]}, [r0], r1
|
||
|
vld4.16 {d1[1], d3[1], d5[1], d7[1]}, [r0], r1
|
||
|
vld4.16 {d1[2], d3[2], d5[2], d7[2]}, [r0], r1
|
||
|
vld4.16 {d1[3], d3[3], d5[3], d7[3]}, [r0], r1
|
||
|
|
||
|
vuzp.8 d0, d1 @D0->p3, D1->p2
|
||
|
vuzp.8 d2, d3 @D2->p1, D3->p0
|
||
|
vuzp.8 d4, d5 @D4->q0, D5->q1
|
||
|
vuzp.8 d6, d7 @D6->q2, D7->q3
|
||
|
|
||
|
rev r12, r12 @reversing ui_bs
|
||
|
vmov.32 d8[0], r12 @D8[0] = ui_Bs
|
||
|
vld1.32 d9[0], [r14] @D9[0] contains cliptab
|
||
|
vmovl.u8 q15, d8 @D30 = ui_Bs in each 16 bt scalar
|
||
|
vtbl.8 d8, {d9}, d30 @puc_ClipTab[ui_Bs]
|
||
|
vsli.16 d8, d8, #8 @D8 = C0
|
||
|
|
||
|
vrhadd.u8 d10, d3, d4 @((p0 + q0 + 1) >> 1)
|
||
|
vmov.i8 d31, #2
|
||
|
vabd.u8 d11, d3, d4 @ABS(p0 - q0)
|
||
|
vaddl.u8 q6, d10, d1 @(p2 + ((p0 + q0 + 1) >> 1)
|
||
|
vmlsl.u8 q6, d2, d31 @(p2 + ((p0 + q0 + 1) >> 1) - (p1 << 1))
|
||
|
vdup.8 d14, r2 @alpha
|
||
|
vcle.u8 d11, d14, d11 @ABS(p0 - q0) >= Alpha(Alpha <=ABS(p0 - q0))
|
||
|
vdup.i8 d14, r3 @beta
|
||
|
vabd.u8 d15, d5, d4 @ABS(q1 - q0)
|
||
|
vqshrn.s16 d12, q6, #1 @((p2 + ((p0 + q0 + 1) >> 1) - (p1 << 1)) >> 1)
|
||
|
vcge.u8 d15, d15, d14 @ABS(q1 - q0) >= Beta
|
||
|
vabd.u8 d13, d2, d3 @ABS(p1 - p0)
|
||
|
vmin.s8 d12, d12, d8 @min(deltap1 ,C0)
|
||
|
vorr d11, d11, d15 @ABS(q1 - q0) >= Beta ||ABS(p0 - q0) >= Alpha
|
||
|
vneg.s8 d15, d8 @-C0
|
||
|
vcge.u8 d13, d13, d14 @ABS(p1 - p0) >= Beta
|
||
|
vmax.s8 d12, d12, d15 @max(deltap1,-C0)
|
||
|
vorr d11, d11, d13 @ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta)
|
||
|
vceq.u16 d13, d30, #0 @ui_bs == 0
|
||
|
vaddl.u8 q14, d10, d6 @q2 + ((p0 + q0 + 1) >> 1)
|
||
|
vsubw.u8 q14, q14, d5 @q2 + ((p0 + q0 + 1) >> 1) - q1
|
||
|
vsubw.u8 q14, q14, d5 @q2 + ((p0 + q0 + 1) >> 1) - 2*q1
|
||
|
vorr d13, d13, d11 @(ABS(p0 - q0) >= Alpha || ABS(q1 - q0) >= Beta || ABS(p1 - p0) >= Beta))
|
||
|
@|| (ui_bs == 0)
|
||
|
vqshrn.s16 d9, q14, #1 @(q2 + ((p0 + q0 + 1) >> 1) - (q1 << 1)) >> 1
|
||
|
vabd.u8 d11, d1, d3 @Ap = ABS(p2 - p0)
|
||
|
vabd.u8 d10, d6, d4 @Aq= ABS(q2 - q0)
|
||
|
vclt.u8 d11, d11, d14 @Ap < Beta
|
||
|
vmin.s8 d9, d9, d8 @min(deltaq1,C0)
|
||
|
vclt.u8 d10, d10, d14 @Aq < Beta
|
||
|
vmax.s8 d9, d9, d15 @max(deltaq1,-C0)
|
||
|
vsubl.u8 q7, d4, d3 @q0 - p0
|
||
|
vshl.s16 q7, q7, #2 @(q0 - p0) << 2
|
||
|
vsub.u8 d8, d8, d11 @C0 + (Ap < Beta)
|
||
|
vaddw.u8 q7, q7, d2 @((q0 - p0) << 2) + p1
|
||
|
vsubw.u8 q7, q7, d5 @((q0 - p0) << 2) + (p1 - q1)
|
||
|
vbic d11, d11, d13 @final condition for p1
|
||
|
vrshr.s16 q15, q7, #3 @delta = (((q0 - p0) << 2) + (p1 - q1) + 4) >> 3
|
||
|
vsub.u8 d8, d8, d10 @C0 + (Ap < Beta) + (Aq < Beta)
|
||
|
vbic d10, d10, d13 @final condition for q1
|
||
|
vabs.s16 q14, q15
|
||
|
vmovn.i16 d15, q14 @abs(delta)
|
||
|
vand d12, d12, d11 @delatp1
|
||
|
vand d9, d9, d10 @deltaq1
|
||
|
vmin.u8 d15, d15, d8 @min((abs(delta),C)
|
||
|
vadd.i8 d2, d2, d12 @p1+deltap1
|
||
|
vadd.i8 d5, d5, d9 @q1+deltaq1
|
||
|
vbic d15, d15, d13 @abs(delta) of pixels to be changed only
|
||
|
vcge.s16 q14, q15, #0
|
||
|
vmovn.i16 d14, q14 @sign(delta)
|
||
|
vqsub.u8 d11, d3, d15 @clip(p0-delta)
|
||
|
vqadd.u8 d3, d3, d15 @clip(p0+delta)
|
||
|
vqadd.u8 d12, d4, d15 @clip(q0+delta)
|
||
|
vqsub.u8 d4, d4, d15 @clip(q0-delta)
|
||
|
vbif d3, d11, d14 @p0
|
||
|
vbif d4, d12, d14 @q0
|
||
|
|
||
|
sub r0, r0, r1, lsl#3 @restore pointer
|
||
|
@D0->p3, D1->p2, D2->p1, D3->p0, D4->q0, D5->q1, D6->q2, D7->q3
|
||
|
vzip.8 d0, d1 @D0,D1 -> [p3:p2]
|
||
|
vzip.8 d2, d3 @D2,D3 -> [p1:p0]
|
||
|
vzip.8 d4, d5 @D4,D5 -> [q0:q1]
|
||
|
vzip.8 d6, d7 @D6,D7 -> [q2:q3]
|
||
|
|
||
|
@storing [p3:p2],[p1:p0]:[q0:q1]:[q2:q3] in every row
|
||
|
vst4.16 {d0[0], d2[0], d4[0], d6[0]}, [r0], r1
|
||
|
vst4.16 {d0[1], d2[1], d4[1], d6[1]}, [r0], r1
|
||
|
vst4.16 {d0[2], d2[2], d4[2], d6[2]}, [r0], r1
|
||
|
vst4.16 {d0[3], d2[3], d4[3], d6[3]}, [r0], r1
|
||
|
vst4.16 {d1[0], d3[0], d5[0], d7[0]}, [r0], r1
|
||
|
vst4.16 {d1[1], d3[1], d5[1], d7[1]}, [r0], r1
|
||
|
vst4.16 {d1[2], d3[2], d5[2], d7[2]}, [r0], r1
|
||
|
vst4.16 {d1[3], d3[3], d5[3], d7[3]}, [r0], r1
|
||
|
vpop {d8 - d15}
|
||
|
ldmfd sp!, {r12, pc}
|
||
|
|
||
|
|
||
|
|