@/***************************************************************************** @* @* 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_horz_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 / akshaya mukund @* @* @par list of functions: @* @* @* @remarks @* none @* @******************************************************************************* @*/ @/** @******************************************************************************* @* @* @brief @* chroma interprediction filter for horizontal input @* @* @par description: @* applies a horizontal filter with coefficients pointed to by 'pi1_coeff' @* to the elements pointed by 'pu1_src' and writes to the location pointed @* by 'pu1_dst' the output is downshifted by 6 and clipped to 8 bits @* assumptions : the function is optimized considering the fact width is @* multiple of 2,4 or 8. if width is 2, then height should be multiple of 2. @* width 4,8 is optimized further @* @* @param[in] pu1_src @* uword8 pointer to the source @* @* @param[out] pu1_dst @* uword8 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_horz(uword8 *pu1_src, @ uword8 *pu1_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_horz_a9q .type ihevc_inter_pred_chroma_horz_a9q, %function ihevc_inter_pred_chroma_horz_a9q: stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments ldr r4,[sp,#40] @loads pi1_coeff ldr r7,[sp,#44] @loads ht ldr r10,[sp,#48] @loads wd vld1.8 {d0},[r4] @coeff = vld1_s8(pi1_coeff) subs r14,r7,#0 @checks for ht == 0 vabs.s8 d2,d0 @vabs_s8(coeff) mov r11,#2 ble end_loops vdup.8 d24,d2[0] @coeffabs_0 = vdup_lane_u8(coeffabs, 0) sub r12,r0,#2 @pu1_src - 2 vdup.8 d25,d2[1] @coeffabs_1 = vdup_lane_u8(coeffabs, 1) add r4,r12,r2 @pu1_src_tmp2_8 = pu1_src + src_strd vdup.8 d26,d2[2] @coeffabs_2 = vdup_lane_u8(coeffabs, 2) tst r10,#3 @checks wd for multiples mov r5,r10,lsl #1 vdup.8 d27,d2[3] @coeffabs_3 = vdup_lane_u8(coeffabs, 3) bne outer_loop_4 cmp r10,#12 beq skip_16 cmp r10,#8 bge outer_loop_16 skip_16: tst r7,#3 sub r9,r0,#2 beq outer_loop_ht_4 @jumps to else condition b outer_loop_8 outer_loop_16: mov r10,r5 @2wd mul r14,r14,r10 rsb r6,r3,#16 add r4,r12,r2 mov r9,#10 and r0, r12, #31 rsb r8,r5,r3,lsl #1 pld [r12, r2, lsl #1] vld1.u32 {q0},[r12],r11 @vector load pu1_src pld [r4, r2, lsl #1] vld1.u32 {q1},[r12],r11 @vector load pu1_src vld1.u32 {q2},[r12],r11 @vector load pu1_src vld1.u32 {q3},[r12],r9 @vector load pu1_src vmull.u8 q15,d2,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {q4},[r4],r11 @vector load pu1_src vmlsl.u8 q15,d0,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {q5},[r4],r11 @vector load pu1_src vmlal.u8 q15,d4,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vld1.u32 {q6},[r4],r11 @vector load pu1_src vmlsl.u8 q15,d6,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {q7},[r4],r9 @vector load pu1_src vmull.u8 q14,d3,d25 vmlsl.u8 q14,d1,d24 vmlal.u8 q14,d5,d26 vmlsl.u8 q14,d7,d27 cmp r14,#32 beq epilog_end sub r14,#64 inner_loop_16: @ bgt l_2 @ pld [r12, r2, lsl #1] @ pld [r4, r2, lsl #1] pld [r12, r2, lsl #2] pld [r4, r2, lsl #2] subs r10,r10,#16 vmull.u8 q11,d10,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ addeq r12,r12,r8 addeq r4,r12,r2 vmlsl.u8 q11,d8,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vqrshrun.s16 d30,q15,#6 vld1.u32 {q0},[r12],r11 @vector load pu1_src vqrshrun.s16 d31,q14,#6 vld1.u32 {q1},[r12],r11 @vector load pu1_src vmlal.u8 q11,d12,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vld1.u32 {q2},[r12],r11 @vector load pu1_src vmlsl.u8 q11,d14,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {q3},[r12],r9 @vector load pu1_src vmull.u8 q10,d11,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q10,d9,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vst1.16 {q15}, [r1],r3 vmlal.u8 q10,d13,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vld1.u32 {q4},[r4],r11 @vector load pu1_src vmlsl.u8 q10,d15,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {q5},[r4],r11 @vector load pu1_src vmull.u8 q15,d2,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {q6},[r4],r11 @vector load pu1_src vmlsl.u8 q15,d0,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {q7},[r4],r9 @vector load pu1_src vmlal.u8 q15,d4,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ cmp r10,#0 vqrshrun.s16 d22,q11,#6 vqrshrun.s16 d23,q10,#6 vmlsl.u8 q15,d6,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ moveq r10,r5 @2wd vmull.u8 q14,d3,d25 vst1.16 {q11},[r1],r6 @store the result pu1_dst vmlsl.u8 q14,d1,d24 addeq r1,r1,r8 vmlal.u8 q14,d5,d26 subs r14,r14,#32 @decrement the ht loop vmlsl.u8 q14,d7,d27 @ mov r0, r7 bgt inner_loop_16 add r14,r14,#64 cmp r14,#32 beq epilog_end epilog: vqrshrun.s16 d30,q15,#6 vqrshrun.s16 d31,q14,#6 vst1.16 {q15}, [r1],r3 vmull.u8 q11,d10,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q11,d8,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ subs r10,r10,#16 @decrement the wd loop vmlal.u8 q11,d12,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ addeq r12,r12,r8 vmlsl.u8 q11,d14,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ moveq r10,r5 @2wd addeq r4,r12,r2 vmull.u8 q10,d11,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {q0},[r12],r11 @vector load pu1_src vmlsl.u8 q10,d9,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {q1},[r12],r11 @vector load pu1_src vmlal.u8 q10,d13,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vld1.u32 {q2},[r12],r11 @vector load pu1_src vmlsl.u8 q10,d15,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {q3},[r12],r9 @vector load pu1_src vmull.u8 q15,d2,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {q4},[r4],r11 @vector load pu1_src vmlsl.u8 q15,d0,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {q5},[r4],r11 @vector load pu1_src vmlal.u8 q15,d4,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q15,d6,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {q6},[r4],r11 @vector load pu1_src vmull.u8 q14,d3,d25 vld1.u32 {q7},[r4],r9 @vector load pu1_src vmlsl.u8 q14,d1,d24 vqrshrun.s16 d22,q11,#6 vqrshrun.s16 d23,q10,#6 vst1.16 {q11},[r1],r6 @store the result pu1_dst vmlal.u8 q14,d5,d26 vmlsl.u8 q14,d7,d27 addeq r1,r1,r8 epilog_end: vqrshrun.s16 d30,q15,#6 vqrshrun.s16 d31,q14,#6 vmull.u8 q11,d10,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q11,d8,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vmlal.u8 q11,d12,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q11,d14,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vmull.u8 q10,d11,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q10,d9,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vmlal.u8 q10,d13,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q10,d15,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vqrshrun.s16 d22,q11,#6 vqrshrun.s16 d23,q10,#6 vst1.16 {q15}, [r1],r3 vst1.16 {q11},[r1] @store the result pu1_dst b end_loops outer_loop_8: add r6,r1,r3 @pu1_dst + dst_strd mov r7,r5 add r4,r12,r2 @pu1_src + src_strd inner_loop_8: @vld1.u32 {d0,d1},[r12],r11 @vector load pu1_src vld1.u32 {d0},[r12],r11 @vector load pu1_src vld1.u32 {d1},[r12],r11 @vector load pu1_src vld1.u32 {d2},[r12],r11 @vector load pu1_src vld1.u32 {d3},[r12],r11 @vector load pu1_src @vext.u8 d2,d0,d1,#2 @vector extract of src[0_2] vmull.u8 q4,d1,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q4,d0,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ @vext.u8 d4,d0,d1,#4 @vector extract of src[0_4] @vext.u8 d6,d0,d1,#6 @vector extract of src[0_6] vmlal.u8 q4,d2,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q4,d3,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {d4},[r4],r11 @vector load pu1_src vld1.u32 {d5},[r4],r11 @vector load pu1_src vld1.u32 {d6},[r4],r11 @vector load pu1_src vld1.u32 {d7},[r4],r11 @vector load pu1_src @vld1.u32 {d12,d13},[r4],r11 @vector load pu1_src + src_strd @vext.u8 d14,d12,d13,#2 @vector extract of src[0_2] vmull.u8 q5,d5,d25 @mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q5,d4,d24 @mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ @vext.u8 d16,d12,d13,#4 @vector extract of src[0_4] @vext.u8 d18,d12,d13,#6 @vector extract of src[0_6] vqrshrun.s16 d8,q4,#6 @right shift and saturating narrow result 1 vmlal.u8 q5,d6,d26 @mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q5,d7,d27 @mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vst1.8 {d8},[r1]! @store the result pu1_dst vqrshrun.s16 d10,q5,#6 @right shift and saturating narrow result 2 subs r7,r7,#8 @decrement the wd loop vst1.8 {d10},[r6]! @store the result pu1_dst bgt inner_loop_8 sub r12,r12,r5 subs r14,r14,#2 @decrement the ht loop sub r1,r1,r5 add r12,r12,r2,lsl #1 add r1,r1,r3,lsl #1 bgt outer_loop_8 b end_loops @height if 4 comes outer_loop_ht_4: mov r7,r5 prologue_ht_4: inner_loop_ht_4: mov r12,r9 mov r4,r1 sub r8, r2, #6 vld1.u32 {d0},[r12],r11 @(1)vector load pu1_src vld1.u32 {d1},[r12],r11 @(1)vector load pu1_src vld1.u32 {d2},[r12],r11 @(1)vector load pu1_src @vld1.u32 {d3},[r12],r2 @(1)vector load pu1_src vld1.u32 {d3},[r12],r8 @(1)vector load pu1_src @sub r12, r12, #6 @(1) vld1.u32 {d4},[r12],r11 @(2)vector load pu1_src vld1.u32 {d5},[r12],r11 @(2)vector load pu1_src vld1.u32 {d6},[r12],r11 @(2)vector load pu1_src @vld1.u32 {d7},[r12],r2 @(2)vector load pu1_src vld1.u32 {d7},[r12],r8 @(2)vector load pu1_src @sub r12, r12, #6 @(2) vld1.u32 {d14},[r12],r11 @(3)vector load pu1_src vmull.u8 q4,d1,d25 @(1)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {d15},[r12],r11 @(3)vector load pu1_src vmlsl.u8 q4,d0,d24 @(1)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {d16},[r12],r11 @(3)vector load pu1_src vmlal.u8 q4,d2,d26 @(1)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ @vld1.u32 {d17},[r12],r2 @(3)vector load pu1_src vld1.u32 {d17},[r12],r8 @(3)vector load pu1_src vmlsl.u8 q4,d3,d27 @(1)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ @sub r12, r12, #6 @(3) vmull.u8 q5,d5,d25 @(2)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {d18},[r12],r11 @(4)vector load pu1_src vmlsl.u8 q5,d4,d24 @(2)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {d19},[r12],r11 @(4)vector load pu1_src vmlal.u8 q5,d6,d26 @(2)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vld1.u32 {d20},[r12],r11 @(4)vector load pu1_src vmlsl.u8 q5,d7,d27 @(2)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vld1.u32 {d21},[r12],r2 @(4)vector load pu1_src vqrshrun.s16 d8,q4,#6 @(1)right shift and saturating narrow result 1 add r9,r9,#8 @(core loop) subs r7,r7,#8 @(prologue)decrement the wd loop beq epilogue core_loop: mov r12,r9 vld1.u32 {d0},[r12],r11 @(1_1)vector load pu1_src vmull.u8 q6,d15,d25 @(3)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {d1},[r12],r11 @(1_1)vector load pu1_src vmlsl.u8 q6,d14,d24 @(3)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {d2},[r12],r11 @(1_1)vector load pu1_src vmlal.u8 q6,d16,d26 @(3)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ @vld1.u32 {d3},[r12],r2 @(1_1)vector load pu1_src vld1.u32 {d3},[r12],r8 @(1_1)vector load pu1_src vmlsl.u8 q6,d17,d27 @(3)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ @sub r12, r12, #6 @(1_1) vst1.8 {d8},[r4],r3 @(1)store the result pu1_dst vqrshrun.s16 d10,q5,#6 @(2)right shift and saturating narrow result 2 vld1.u32 {d4},[r12],r11 @(2_1)vector load pu1_src vmull.u8 q11,d19,d25 @(4)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {d5},[r12],r11 @(2_1)vector load pu1_src vmlsl.u8 q11,d18,d24 @(4)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {d6},[r12],r11 @(2_1)vector load pu1_src vmlal.u8 q11,d20,d26 @(4)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ @vld1.u32 {d7},[r12],r2 @(2_1)vector load pu1_src vld1.u32 {d7},[r12],r8 @(2_1)vector load pu1_src vmlsl.u8 q11,d21,d27 @(4)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ @sub r12, r12, #6 @(2_1) vst1.8 {d10},[r4],r3 @(2)store the result pu1_dst vqrshrun.s16 d12,q6,#6 @(3)right shift and saturating narrow result 1 vld1.u32 {d14},[r12],r11 @(3_1)vector load pu1_src vmull.u8 q4,d1,d25 @(1_1)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {d15},[r12],r11 @(3_1)vector load pu1_src vmlsl.u8 q4,d0,d24 @(1_1)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {d16},[r12],r11 @(3_1)vector load pu1_src vmlal.u8 q4,d2,d26 @(1_1)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ @vld1.u32 {d17},[r12],r2 @(3_1)vector load pu1_src vld1.u32 {d17},[r12],r8 @(3_1)vector load pu1_src vmlsl.u8 q4,d3,d27 @(1_1)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ @sub r12, r12, #6 @(3_1) vst1.8 {d12},[r4],r3 @(3)store the result pu1_dst vqrshrun.s16 d22,q11,#6 @(4)right shift and saturating narrow result 2 add r9,r9,#8 @(core loop) vmull.u8 q5,d5,d25 @(2_1)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vld1.u32 {d18},[r12],r11 @(4_1)vector load pu1_src vld1.u32 {d19},[r12],r11 @(4_1)vector load pu1_src vmlsl.u8 q5,d4,d24 @(2_1)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vld1.u32 {d20},[r12],r11 @(4_1)vector load pu1_src vmlal.u8 q5,d6,d26 @(2_1)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vld1.u32 {d21},[r12],r2 @(4_1)vector load pu1_src vmlsl.u8 q5,d7,d27 @(2_1)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ add r1,r1,#8 @(core loop) subs r7,r7,#8 @(core loop) vst1.8 {d22}, [r4], r3 @(4)store the result pu1_dst vqrshrun.s16 d8,q4,#6 @(1_1)right shift and saturating narrow result 1 mov r4, r1 @(core loop) bgt core_loop @loopback epilogue: vmull.u8 q6,d15,d25 @(3)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q6,d14,d24 @(3)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vmlal.u8 q6,d16,d26 @(3)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q6,d17,d27 @(3)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vst1.8 {d8},[r4],r3 @(1)store the result pu1_dst vqrshrun.s16 d10,q5,#6 @(2)right shift and saturating narrow result 2 vmull.u8 q11,d19,d25 @(4)mul_res = vmull_u8(src[0_3], coeffabs_3)@ vmlsl.u8 q11,d18,d24 @(4)mul_res = vmlsl_u8(src[0_2], coeffabs_2)@ vmlal.u8 q11,d20,d26 @(4)mul_res = vmlsl_u8(src[0_0], coeffabs_0)@ vmlsl.u8 q11,d21,d27 @(4)mul_res = vmlal_u8(src[0_1], coeffabs_1)@ vst1.8 {d10},[r4],r3 @(2)store the result pu1_dst vqrshrun.s16 d12,q6,#6 @(3)right shift and saturating narrow result 1 vst1.8 {d12},[r4],r3 @(3)store the result pu1_dst add r1,r1,#8 @(core loop) vqrshrun.s16 d22,q11,#6 @(4)right shift and saturating narrow result 2 vst1.8 {d22}, [r4], r3 @(4)store the result pu1_dst sub r9,r9,r5 subs r14,r14,#4 @decrement the ht loop sub r1,r1,r5 add r9,r9,r2,lsl #2 add r1,r1,r3,lsl #2 bgt outer_loop_ht_4 b end_loops outer_loop_4: add r6,r1,r3 @pu1_dst + dst_strd mov r7,r5 add r4,r12,r2 @pu1_src + src_strd inner_loop_4: @vld1.u32 {d0,d1},[r12] @vector load pu1_src vld1.u32 {d0},[r12],r11 @vector load pu1_src vld1.u32 {d1},[r12],r11 @vector load pu1_src vld1.u32 {d2},[r12],r11 @vector load pu1_src vld1.u32 {d3},[r12] @vector load pu1_src sub r12,r12,#2 @increment the input pointer vld1.u32 {d4},[r4],r11 @vector load pu1_src vld1.u32 {d5},[r4],r11 @vector load pu1_src vld1.u32 {d6},[r4],r11 @vector load pu1_src vld1.u32 {d7},[r4] @vector load pu1_src @vext.u8 d2,d0,d1,#2 @vector extract of src[0_2] @vext.u8 d4,d0,d1,#4 @vector extract of src[0_4] @vld1.u32 {d12,d13},[r4] @vector load pu1_src + src_strd @vext.u8 d6,d0,d1,#6 @vector extract of src[0_6] sub r4,r4,#2 @increment the input pointer @vext.u8 d14,d12,d13,#2 @vector extract of src[0_2] @vext.u8 d16,d12,d13,#4 @vector extract of src[0_4] @vext.u8 d18,d12,d13,#6 @vector extract of src[0_6] vzip.32 d0,d4 @vector zip the i iteration and ii interation in single register vzip.32 d1,d5 vzip.32 d2,d6 vzip.32 d3,d7 vmull.u8 q4,d1,d25 @arithmetic operations for ii iteration in the same time vmlsl.u8 q4,d0,d24 vmlal.u8 q4,d2,d26 vmlsl.u8 q4,d3,d27 vqrshrun.s16 d8,q4,#6 @narrow right shift and saturating the result vst1.32 {d8[0]},[r1]! @store the i iteration result which is in upper part of the register subs r7,r7,#4 @decrement the wd by 4 vst1.32 {d8[1]},[r6]! @store the ii iteration result which is in lower part of the register bgt inner_loop_4 sub r12,r12,r5 subs r14,r14,#2 @decrement the ht by 2 sub r1,r1,r5 add r12,r12,r2,lsl #1 add r1,r1,r3,lsl #1 bgt outer_loop_4 end_loops: ldmfd sp!,{r4-r12,r15} @reload the registers from sp