@/****************************************************************************** @ * @ * 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 @* ih264_inter_pred_luma_horz_hpel_vert_hpel_a9q.s @* @* @brief @* Contains function definitions for inter prediction interpolation. @* @* @author @* Mohit @* @* @par List of Functions: @* @* - ih264_inter_pred_luma_horz_hpel_vert_hpel_a9q() @* @* @remarks @* None @* @******************************************************************************* @* @* All the functions here are replicated from ih264_inter_pred_filters.c @ @** @** @** @******************************************************************************* @* @* @brief @* This function implements a two stage cascaded six tap filter. It @* applies the six tap filter in the vertical direction on the @* predictor values, followed by applying the same filter in the @* horizontal direction on the output of the first stage. The six tap @* filtering operation is described in sec 8.4.2.2.1 titled "Luma sample @* interpolation process" @* @* @par Description: @* This function is called to obtain pixels lying at the following @* location (1/2,1/2). The function interpolates @* the predictors first in the horizontal direction and then in the @* vertical direction to output the (1/2,1/2). @* @* @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] ht @* integer height of the array @* @* @param[in] wd @* integer width of the array @* @* @param[in] pu1_tmp: temporary buffer @* @* @param[in] dydx: x and y reference offset for qpel calculations: UNUSED in this function. @* @* @returns @* @* @remarks @* None @* @******************************************************************************* @*; @void ih264_inter_pred_luma_horz_hpel_vert_hpel(UWORD8 *pu1_src, @ UWORD8 *pu1_dst, @ WORD32 src_strd,, @ WORD32 dst_strd, @ WORD32 ht, @ WORD32 wd, @ UWORD8* pu1_tmp, @ UWORD32 dydx) @**************Variables Vs Registers***************************************** @ r0 => *pu1_src @ r1 => *pu1_dst @ r2 => src_strd @ r3 => dst_strd @ r8 => ht @ r9 => wd .text .p2align 2 .global ih264_inter_pred_luma_horz_hpel_vert_hpel_a9q ih264_inter_pred_luma_horz_hpel_vert_hpel_a9q: stmfd sp!, {r4-r12, r14} @store register values to stack vstmdb sp!, {d8-d15} @push neon registers to stack ldr r8, [sp, #104] @ loads ht sub r0, r0, r2, lsl #1 @pu1_src-2*src_strd sub r0, r0, #2 @pu1_src-2 ldr r9, [sp, #108] @ loads wd vmov.s16 d0, #20 @ Filter coeff 20 vmov.s16 d1, #5 @ Filter coeff 5 subs r12, r9, #4 @if wd=4 branch to loop_4 beq loop_4 subs r12, r9, #8 @if wd=8 branch to loop_8 beq loop_8 mov r10, #8 sub r7, r3, r10 @when wd=16 loop_16: vld1.u32 {d2, d3, d4}, [r0], r2 @ Vector load from src[0_0] vld1.u32 {d5, d6, d7}, [r0], r2 @ Vector load from src[1_0] vld1.u32 {d8, d9, d10}, [r0], r2 @ Vector load from src[2_0] vld1.u32 {d11, d12, d13}, [r0], r2 @ Vector load from src[3_0] vld1.u32 {d14, d15, d16}, [r0], r2 @ Vector load from src[4_0] vld1.u32 {d17, d18, d19}, [r0], r2 @ Vector load from src[5_0] @ vERTICAL FILTERING FOR ROW 0 vaddl.u8 q10, d8, d11 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q12, d2, d17 @ temp2 = src[0_0] + src[5_0] vaddl.u8 q11, d5, d14 @ temp = src[1_0] + src[4_0] vaddl.u8 q13, d3, d18 @ temp2 = src[0_0] + src[5_0] vmla.u16 q12, q10, d0[0] @ temp += temp1 * 20 vmls.s16 q12, q11, d1[0] @ temp -= temp2 * 5 vaddl.u8 q10, d6, d15 @ temp = src[1_0] + src[4_0] vaddl.u8 q11, d9, d12 @ temp3 = src[2_0] + src[3_0] vaddl.u8 q14, d4, d19 @ temp2 = src[0_0] + src[5_0] vmla.u16 q13, q11, d0[0] @ temp4 += temp3 * 20 vmls.s16 q13, q10, d1[0] @ temp -= temp2 * 5 vaddl.u8 q11, d10, d13 @ temp3 = src[2_0] + src[3_0] vaddl.u8 q10, d7, d16 @ temp = src[1_0] + src[4_0] vmla.u16 q14, q11, d0[0] @ temp4 += temp3 * 20 vmls.s16 q14, q10, d1[0] @ temp -= temp2 * 5 vext.16 q10, q12, q13, #5 @//extract a[5] (column1) @Q12,Q13,Q14 HAVE VERTICAL FILTERED VALUES @CASCADED FILTERING FOR ROW 0 vext.16 q11, q12, q13, #2 @//extract a[2] (column1) vaddl.s16 q1, d20, d24 @// a0 + a5 (column1) vaddl.s16 q15, d21, d25 @// a0 + a5 (column1) vmlal.s16 q1, d22, d0[0] @// a0 + a5 + 20a2 (column1) vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column1) vext.16 q11, q12, q13, #1 @//extract a[1] (column1) vext.16 q10, q12, q13, #3 @//extract a[3] (column1) vmlsl.s16 q1, d22, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlsl.s16 q15, d23, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlal.s16 q1, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vext.16 q11, q12, q13, #4 @//extract a[4] (column1) vext.16 q10, q13, q14, #5 @//extract a[5] (column2) vmlsl.s16 q1, d22, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vmlsl.s16 q15, d23, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vqrshrun.s32 d22, q1, #10 vqrshrun.s32 d23, q15, #10 vqmovun.s16 d22, q11 vst1.u8 {d22}, [r1], r10 @//Store dest row0, column 1; (1/2,1/2) vext.16 q11, q13, q14, #2 @//extract a[2] (column2) vaddl.s16 q1, d20, d26 @// a0 + a5 (column2) vaddl.s16 q15, d21, d27 @// a0 + a5 (column2) vmlal.s16 q1, d22, d0[0] @// a0 + a5 + 20a2 (column2) vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column2) vext.16 q10, q13, q14, #3 @//extract a[3] (column2) vext.16 q11, q13, q14, #1 @//extract a[1] (column2) vmlal.s16 q1, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column2) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column2) vext.16 q10, q13, q14, #4 @//extract a[4] (column2) vmlsl.s16 q1, d22, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column2) vmlsl.s16 q15, d23, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column2) vmlsl.s16 q1, d20, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column2) vmlsl.s16 q15, d21, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column2) vqrshrun.s32 d20, q1, #10 vqrshrun.s32 d21, q15, #10 vld1.u32 {d2, d3, d4}, [r0], r2 @ Vector load from src[6_0] vqmovun.s16 d22, q10 vst1.u8 {d22}, [r1], r7 @//Store dest row0 ,column 2; (1/2,1/2) @ vERTICAL FILTERING FOR ROW 1 vaddl.u8 q10, d11, d14 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q12, d5, d2 @ temp2 = src[0_0] + src[5_0] vaddl.u8 q11, d8, d17 @ temp = src[1_0] + src[4_0] vaddl.u8 q13, d6, d3 @ temp2 = src[0_0] + src[5_0] vmla.u16 q12, q10, d0[0] @ temp += temp1 * 20 vaddl.u8 q10, d9, d18 @ temp = src[1_0] + src[4_0] vmls.s16 q12, q11, d1[0] @ temp -= temp2 * 5 vaddl.u8 q11, d12, d15 @ temp3 = src[2_0] + src[3_0] vaddl.u8 q14, d7, d4 @ temp2 = src[0_0] + src[5_0] vmla.u16 q13, q11, d0[0] @ temp4 += temp3 * 20 vaddl.u8 q11, d13, d16 @ temp3 = src[2_0] + src[3_0] vmls.s16 q13, q10, d1[0] @ temp -= temp2 * 5 vmla.u16 q14, q11, d0[0] @ temp4 += temp3 * 20 vaddl.u8 q10, d10, d19 @ temp = src[1_0] + src[4_0] vmls.s16 q14, q10, d1[0] @ temp -= temp2 * 5 vext.16 q10, q12, q13, #5 @//extract a[5] (column1) @Q12,Q13,Q14 HAVE VERTICAL FILTERED VALUES @CASCADED FILTERING FOR ROW 1 vext.16 q11, q12, q13, #2 @//extract a[2] (column1) vaddl.s16 q3, d20, d24 @// a0 + a5 (column1) vaddl.s16 q15, d21, d25 @// a0 + a5 (column1) vmlal.s16 q3, d22, d0[0] @// a0 + a5 + 20a2 (column1) vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column1) vext.16 q11, q12, q13, #1 @//extract a[1] (column1) vext.16 q10, q12, q13, #3 @//extract a[3] (column1) vmlsl.s16 q3, d22, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlsl.s16 q15, d23, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlal.s16 q3, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vext.16 q11, q12, q13, #4 @//extract a[4] (column1) vext.16 q10, q13, q14, #5 @//extract a[5] (column2) vmlsl.s16 q3, d22, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vmlsl.s16 q15, d23, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vqrshrun.s32 d22, q3, #10 vqrshrun.s32 d23, q15, #10 vqmovun.s16 d22, q11 vst1.u8 {d22}, [r1], r10 @//Store dest row1, column 1; (1/2,1/2) vext.16 q11, q13, q14, #2 @//extract a[2] (column2) vaddl.s16 q3, d20, d26 @// a0 + a5 (column2) vaddl.s16 q15, d21, d27 @// a0 + a5 (column2) vmlal.s16 q3, d22, d0[0] @// a0 + a5 + 20a2 (column2) vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column2) vext.16 q10, q13, q14, #3 @//extract a[3] (column2) vext.16 q11, q13, q14, #1 @//extract a[1] (column2) vmlal.s16 q3, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column2) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column2) vext.16 q10, q13, q14, #4 @//extract a[4] (column2) vmlsl.s16 q3, d22, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column2) vmlsl.s16 q15, d23, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column2) vmlsl.s16 q3, d20, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column2) vmlsl.s16 q15, d21, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column2) vqrshrun.s32 d20, q3, #10 vqrshrun.s32 d21, q15, #10 vqmovun.s16 d22, q10 vst1.u8 {d22}, [r1], r7 @//Store dest row1 ,column 2; (1/2,1/2) subs r8, r8, #2 @ 2 rows processed, decrement by 2 subne r0, r0 , r2, lsl #2 subne r0, r0, r2 beq end_func @ Branch if height==4 b loop_16 @ looping if height = 8 or 16 loop_8: vld1.u32 {d2, d3}, [r0], r2 @ Vector load from src[0_0] vld1.u32 {d4, d5}, [r0], r2 @ Vector load from src[1_0] vld1.u32 {d6, d7}, [r0], r2 @ Vector load from src[2_0] vld1.u32 {d8, d9}, [r0], r2 @ Vector load from src[3_0] vld1.u32 {d10, d11}, [r0], r2 @ Vector load from src[4_0] vld1.u32 {d12, d13}, [r0], r2 @ Vector load from src[5_0] @ vERTICAL FILTERING FOR ROW 0 vaddl.u8 q10, d6, d8 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q11, d4, d10 @ temp2 = src[1_0] + src4_0] vaddl.u8 q12, d2, d12 @ temp = src[0_0] + src[5_0] vaddl.u8 q13, d3, d13 @ temp = src[0_0] + src[5_0] vaddl.u8 q14, d7, d9 @ temp1 = src[2_0] + src[3_0] vmla.u16 q12, q10, d0[0] @ temp += temp1 * 20 vmls.s16 q12, q11, d1[0] @ temp -= temp2 * 5 vaddl.u8 q15, d5, d11 @ temp2 = src[1_0] + src4_0] vmla.u16 q13, q14, d0[0] @ temp += temp1 * 20 vmls.s16 q13, q15, d1[0] @ temp -= temp2 * 5 @Q12,Q13 HAVE VERTICAL FILTERED VALUES @CASCADED FILTERING FOR ROW 0 vext.16 q10, q12, q13, #5 @//extract a[5] (column1) vext.16 q11, q12, q13, #2 @//extract a[2] (column1) vaddl.s16 q14, d20, d24 @// a0 + a5 (column1) vaddl.s16 q15, d21, d25 @// a0 + a5 (column1) vext.16 q9, q12, q13, #1 @//extract a[1] (column1) vext.16 q10, q12, q13, #3 @//extract a[3] (column1) vext.16 q1, q12, q13, #4 @//extract a[4] (column1) vmlal.s16 q14, d22, d0[0] @// a0 + a5 + 20a2 (column1) vmlsl.s16 q14, d18, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlal.s16 q14, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q14, d2, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vld1.u32 {d14, d15}, [r0], r2 @ Vector load from src[6_0] vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column1) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q15, d19, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlsl.s16 q15, d3, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vaddl.u8 q12, d4, d14 @ temp = src[0_0] + src[5_0] vaddl.u8 q13, d5, d15 @ temp = src[0_0] + src[5_0] vqrshrun.s32 d18, q14, #10 vaddl.u8 q14, d9, d11 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q10, d8, d10 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q11, d6, d12 @ temp2 = src[1_0] + src4_0] vqrshrun.s32 d19, q15, #10 vmla.u16 q12, q10, d0[0] @ temp += temp1 * 20 vmls.s16 q12, q11, d1[0] @ temp -= temp2 * 5 vaddl.u8 q15, d7, d13 @ temp2 = src[1_0] + src4_0] vmla.u16 q13, q14, d0[0] @ temp += temp1 * 20 vmls.s16 q13, q15, d1[0] @ temp -= temp2 * 5 vqmovun.s16 d2, q9 @ vERTICAL FILTERING FOR ROW 1 @Q12,Q13 HAVE VERTICAL FILTERED VALUES @CASCADED FILTERING FOR ROW 1 vext.16 q10, q12, q13, #5 @//extract a[5] (column1) vext.16 q11, q12, q13, #2 @//extract a[2] (column1) vaddl.s16 q14, d20, d24 @// a0 + a5 (column1) vaddl.s16 q15, d21, d25 @// a0 + a5 (column1) vst1.u8 {d2}, [r1], r3 @//Store dest row0, column 1; (1/2,1/2) vext.16 q9, q12, q13, #1 @//extract a[1] (column1) vext.16 q10, q12, q13, #3 @//extract a[3] (column1) vext.16 q2, q12, q13, #4 @//extract a[4] (column1) vmlal.s16 q14, d22, d0[0] @// a0 + a5 + 20a2 (column1) vmlsl.s16 q14, d18, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlal.s16 q14, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q14, d4, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column1) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q15, d19, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlsl.s16 q15, d5, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vqrshrun.s32 d18, q14, #10 vqrshrun.s32 d19, q15, #10 vqmovun.s16 d3, q9 vst1.u8 {d3}, [r1], r3 @//Store dest row1, column 1; (1/2,1/2) subs r8, r8, #2 @ 2 rows processed, decrement by 2 subne r0, r0 , r2, lsl #2 subne r0, r0, r2 beq end_func @ Branch if height==4 b loop_8 @looping if height == 8 or 16 loop_4: vld1.u32 {d2, d3}, [r0], r2 @ Vector load from src[0_0] vld1.u32 {d4, d5}, [r0], r2 @ Vector load from src[1_0] vld1.u32 {d6, d7}, [r0], r2 @ Vector load from src[2_0] vld1.u32 {d8, d9}, [r0], r2 @ Vector load from src[3_0] vld1.u32 {d10, d11}, [r0], r2 @ Vector load from src[4_0] vld1.u32 {d12, d13}, [r0], r2 @ Vector load from src[5_0] @ vERTICAL FILTERING FOR ROW 0 vaddl.u8 q10, d6, d8 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q11, d4, d10 @ temp2 = src[1_0] + src4_0] vaddl.u8 q12, d2, d12 @ temp = src[0_0] + src[5_0] vaddl.u8 q13, d3, d13 @ temp = src[0_0] + src[5_0] vaddl.u8 q14, d7, d9 @ temp1 = src[2_0] + src[3_0] vmla.u16 q12, q10, d0[0] @ temp += temp1 * 20 vmls.s16 q12, q11, d1[0] @ temp -= temp2 * 5 vaddl.u8 q15, d5, d11 @ temp2 = src[1_0] + src4_0] vmla.u16 q13, q14, d0[0] @ temp += temp1 * 20 vmls.s16 q13, q15, d1[0] @ temp -= temp2 * 5 @Q12,Q13 HAVE VERTICAL FILTERED VALUES @CASCADED FILTERING FOR ROW 0 vext.16 q10, q12, q13, #5 @//extract a[5] (column1) vext.16 q11, q12, q13, #2 @//extract a[2] (column1) vaddl.s16 q14, d20, d24 @// a0 + a5 (column1) vaddl.s16 q15, d21, d25 @// a0 + a5 (column1) vext.16 q1, q12, q13, #4 @//extract a[4] (column1) vext.16 q9, q12, q13, #1 @//extract a[1] (column1) vext.16 q10, q12, q13, #3 @//extract a[3] (column1) vmlal.s16 q14, d22, d0[0] @// a0 + a5 + 20a2 (column1) vmlsl.s16 q14, d18, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlal.s16 q14, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q14, d2, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vld1.u32 {d14, d15}, [r0], r2 @ Vector load from src[6_0] vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column1) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q15, d19, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlsl.s16 q15, d3, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vaddl.u8 q12, d4, d14 @ temp = src[0_0] + src[5_0] vaddl.u8 q13, d5, d15 @ temp = src[0_0] + src[5_0] vqrshrun.s32 d18, q14, #10 vaddl.u8 q14, d9, d11 @ temp1 = src[2_0] + src[3_0] vaddl.u8 q11, d6, d12 @ temp2 = src[1_0] + src4_0] vaddl.u8 q10, d8, d10 @ temp1 = src[2_0] + src[3_0] vqrshrun.s32 d19, q15, #10 vmla.u16 q12, q10, d0[0] @ temp += temp1 * 20 vmls.s16 q12, q11, d1[0] @ temp -= temp2 * 5 vaddl.u8 q15, d7, d13 @ temp2 = src[1_0] + src4_0] vqmovun.s16 d2, q9 vmla.u16 q13, q14, d0[0] @ temp += temp1 * 20 vmls.s16 q13, q15, d1[0] @ temp -= temp2 * 5 @ vERTICAL FILTERING FOR ROW 1 @Q12,Q13 HAVE VERTICAL FILTERED VALUES @CASCADED FILTERING FOR ROW 1 vext.16 q10, q12, q13, #5 @//extract a[5] (column1) vext.16 q11, q12, q13, #2 @//extract a[2] (column1) vst1.u32 {d2[0]}, [r1], r3 @//Store dest row0, column 1; (1/2,1/2) vaddl.s16 q14, d20, d24 @// a0 + a5 (column1) vaddl.s16 q15, d21, d25 @// a0 + a5 (column1) vext.16 q9, q12, q13, #1 @//extract a[1] (column1) vext.16 q10, q12, q13, #3 @//extract a[3] (column1) vext.16 q2, q12, q13, #4 @//extract a[4] (column1) vmlal.s16 q14, d22, d0[0] @// a0 + a5 + 20a2 (column1) vmlsl.s16 q14, d18, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlal.s16 q14, d20, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q14, d4, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vmlal.s16 q15, d23, d0[0] @// a0 + a5 + 20a2 (column1) vmlal.s16 q15, d21, d0[0] @// a0 + a5 + 20a2 + 20a3 (column1) vmlsl.s16 q15, d19, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 (column1) vmlsl.s16 q15, d5, d1[0] @// a0 + a5 + 20a2 + 20a3 - 5a1 - 5a4 (column1) vqrshrun.s32 d18, q14, #10 vqrshrun.s32 d19, q15, #10 vqmovun.s16 d4, q9 vst1.u32 {d4[0]}, [r1], r3 @//Store dest row1, column 1; (1/2,1/2) subs r8, r8, #2 @ 2 rows processed, decrement by 2 subne r0, r0 , r2, lsl #2 subne r0, r0, r2 beq end_func @ Branch if height==4 b loop_4 @looping if height == 8 or 16 end_func: vldmia sp!, {d8-d15} @ Restore neon registers that were saved ldmfd sp!, {r4-r12, pc} @Restoring registers from stack