//****************************************************************************** //* //* 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_qpel_vert_hpel_av8.s //* //* @brief //* Contains function definitions for inter prediction interpolation. //* //* @author //* Mohit //* //* @par List of Functions: //* //* - ih264_inter_pred_luma_horz_qpel_vert_hpel_av8() //* //* @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. It then averages //* the output of the 1st stage and the final stage to obtain the quarter //* pel values.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 (1/4,1/2) or (3/4,1/2). The function interpolates //* the predictors first in the verical direction and then in the //* horizontal direction to output the (1/2,1/2). It then averages //* the output of the 2nd stage and (1/2,1/2) value to obtain (1/4,1/2) //* or (3/4,1/2) depending on the offset. //* //* @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 //* //* @returns //* //* @remarks //* None //* //******************************************************************************* //*/; //void ih264_inter_pred_luma_horz_qpel_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***************************************** // x0 => *pu1_src // x1 => *pu1_dst // w2 => src_strd // w3 => dst_strd // w4 => ht // w5 => wd // x6 => *pu1_tmp // w7 => dydx .text .p2align 2 .include "ih264_neon_macros.s" .global ih264_inter_pred_luma_horz_qpel_vert_hpel_av8 ih264_inter_pred_luma_horz_qpel_vert_hpel_av8: // STMFD sp!, {x4-x12, x14} //store register values to stack push_v_regs stp x19, x20, [sp, #-16]! sxtw x2, w2 sxtw x3, w3 sxtw x4, w4 sxtw x5, w5 sub x0, x0, x2, lsl #1 //pu1_src-2*src_strd sub x0, x0, #2 //pu1_src-2 mov x9, x6 mov w6, w7 and x6, x6, #2 // dydx & 0x3 followed by dydx>>1 and dydx<<1 add x7, x9, #4 add x6, x7, x6 // pi16_pred1_temp += (x_offset>>1) movi v26.8h, #0x14 // Filter coeff 20 into Q13 movi v24.8h, #0x5 // Filter coeff 5 into Q12 movi v27.8h, #0x14 // Filter coeff 20 into Q13 movi v25.8h, #0x5 // Filter coeff 5 into Q12 mov x7, #0x20 mov x8, #0x30 subs x12, x5, #4 //if wd=4 branch to loop_4 beq loop_4_start subs x12, x5, #8 //if wd=8 branch to loop_8 beq loop_8_start //when wd=16 movi v28.8h, #0x14 // Filter coeff 20 into Q13 movi v30.8h, #0x5 // Filter coeff 5 into Q12 sub x2, x2, #16 ld1 {v0.2s, v1.2s}, [x0], #16 // Vector load from src[0_0] ld1 {v12.2s}, [x0], x2 // Vector load from src[0_0] ld1 {v2.2s, v3.2s}, [x0], #16 // Vector load from src[1_0] ld1 {v13.2s}, [x0], x2 // Vector load from src[1_0] ld1 {v4.2s, v5.2s}, [x0], #16 // Vector load from src[2_0] ld1 {v14.2s}, [x0], x2 // Vector load from src[2_0] ld1 {v6.2s, v7.2s}, [x0], #16 // Vector load from src[3_0] ld1 {v15.2s}, [x0], x2 // Vector load from src[3_0] ld1 {v8.2s, v9.2s}, [x0], #16 // Vector load from src[4_0] ld1 {v16.2s}, [x0], x2 // Vector load from src[4_0] loop_16: ld1 {v10.2s, v11.2s}, [x0], #16 // Vector load from src[5_0] ld1 {v17.2s}, [x0], x2 // Vector load from src[5_0] uaddl v20.8h, v4.8b, v6.8b uaddl v18.8h, v0.8b, v10.8b uaddl v22.8h, v2.8b, v8.8b mla v18.8h, v20.8h , v28.8h uaddl v24.8h, v5.8b, v7.8b uaddl v20.8h, v1.8b, v11.8b uaddl v26.8h, v3.8b, v9.8b mla v20.8h, v24.8h , v28.8h uaddl v24.8h, v14.8b, v15.8b mls v18.8h, v22.8h , v30.8h uaddl v22.8h, v12.8b, v17.8b mls v20.8h, v26.8h , v30.8h uaddl v26.8h, v13.8b, v16.8b mla v22.8h, v24.8h , v28.8h mls v22.8h, v26.8h , v30.8h st1 {v18.4s }, [x9], #16 st1 {v20.4s}, [x9], #16 ext v24.16b, v18.16b , v20.16b , #4 ext v26.16b, v18.16b , v20.16b , #6 st1 {v22.4s}, [x9] ext v22.16b, v18.16b , v20.16b , #10 add v0.8h, v24.8h , v26.8h ext v24.16b, v18.16b , v20.16b , #2 ext v26.16b, v18.16b , v20.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v18.4h, v22.4h smlal v26.4s, v0.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v18.8h, v22.8h smlal2 v22.4s, v0.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v18.4h, v26.4s, #10 sqrshrun v19.4h, v22.4s, #10 ld1 {v22.4s}, [x9], #16 uqxtn v18.8b, v18.8h uqxtn v19.8b, v19.8h mov v18.s[1], v19.s[0] ext v24.16b, v20.16b , v22.16b , #4 ext v26.16b, v20.16b , v22.16b , #6 ext v0.16b, v20.16b , v22.16b , #10 st1 {v18.2s}, [x1] add v18.8h, v24.8h , v26.8h ext v24.16b, v20.16b , v22.16b , #2 ext v26.16b, v20.16b , v22.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v0.4h, v20.4h smlal v26.4s, v18.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v0.8h, v20.8h smlal2 v22.4s, v18.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v19.4h, v26.4s, #10 sqrshrun v18.4h, v22.4s, #10 uaddl v24.8h, v7.8b, v9.8b ld1 {v20.4s}, [x6], #16 ld1 {v22.4s}, [x6], x7 uqxtn v19.8b, v19.8h uqxtn v18.8b, v18.8h mov v19.s[1], v18.s[0] ld1 {v18.2s}, [x1] sqrshrun v20.8b, v20.8h, #5 sqrshrun v21.8b, v22.8h, #5 uaddl v22.8h, v4.8b, v10.8b ld1 {v0.2s, v1.2s}, [x0], #16 // Vector load from src[6_0] urhadd v18.16b, v18.16b , v20.16b urhadd v19.16b, v19.16b , v21.16b ld1 {v12.2s}, [x0], x2 // Vector load from src[6_0] uaddl v20.8h, v6.8b, v8.8b uaddl v26.8h, v5.8b, v11.8b st1 {v18.2s, v19.2s}, [x1], x3 // store row 0 //ROW_2 uaddl v18.8h, v2.8b, v0.8b mla v18.8h, v20.8h , v28.8h uaddl v20.8h, v3.8b, v1.8b mla v20.8h, v24.8h , v28.8h uaddl v24.8h, v15.8b, v16.8b mls v18.8h, v22.8h , v30.8h uaddl v22.8h, v13.8b, v12.8b mls v20.8h, v26.8h , v30.8h uaddl v26.8h, v14.8b, v17.8b mla v22.8h, v24.8h , v28.8h mls v22.8h, v26.8h , v30.8h st1 {v18.4s}, [x9], #16 st1 {v20.4s}, [x9], #16 ext v24.16b, v18.16b , v20.16b , #4 ext v26.16b, v18.16b , v20.16b , #6 st1 {v22.4s}, [x9] ext v22.16b, v18.16b , v20.16b , #10 add v2.8h, v24.8h , v26.8h ext v24.16b, v18.16b , v20.16b , #2 ext v26.16b, v18.16b , v20.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v18.4h, v22.4h smlal v26.4s, v2.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v18.8h, v22.8h smlal2 v22.4s, v2.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v18.4h, v26.4s, #10 sqrshrun v19.4h, v22.4s, #10 ld1 {v22.4s}, [x9], #16 uqxtn v18.8b, v18.8h uqxtn v19.8b, v19.8h mov v18.s[1], v19.s[0] ext v24.16b, v20.16b , v22.16b , #4 ext v26.16b, v20.16b , v22.16b , #6 ext v2.16b, v20.16b , v22.16b , #10 st1 {v18.2s}, [x1] add v18.8h, v24.8h , v26.8h ext v24.16b, v20.16b , v22.16b , #2 ext v26.16b, v20.16b , v22.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v2.4h, v20.4h smlal v26.4s, v18.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v2.8h, v20.8h smlal2 v22.4s, v18.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v19.4h, v26.4s, #10 sqrshrun v18.4h, v22.4s, #10 uaddl v24.8h, v9.8b, v11.8b ld1 {v20.4s}, [x6], #16 ld1 {v22.4s}, [x6], x7 uqxtn v19.8b, v19.8h uqxtn v18.8b, v18.8h mov v19.s[1], v18.s[0] ld1 {v18.4s}, [x1] sqrshrun v20.8b, v20.8h, #5 sqrshrun v21.8b, v22.8h, #5 uaddl v22.8h, v6.8b, v0.8b ld1 {v2.2s, v3.2s}, [x0], #16 // Vector load from src[7_0] urhadd v18.16b, v18.16b , v20.16b urhadd v19.16b, v19.16b , v21.16b ld1 {v13.2s}, [x0], x2 // Vector load from src[7_0] uaddl v20.8h, v8.8b, v10.8b uaddl v26.8h, v7.8b, v1.8b st1 {v18.2s, v19.2s}, [x1], x3 // store row 1 //ROW_3 uaddl v18.8h, v4.8b, v2.8b mla v18.8h, v20.8h , v28.8h uaddl v20.8h, v5.8b, v3.8b mla v20.8h, v24.8h , v28.8h uaddl v24.8h, v16.8b, v17.8b mls v18.8h, v22.8h , v30.8h uaddl v22.8h, v14.8b, v13.8b mls v20.8h, v26.8h , v30.8h uaddl v26.8h, v15.8b, v12.8b mla v22.8h, v24.8h , v28.8h mls v22.8h, v26.8h , v30.8h st1 {v18.4s}, [x9], #16 st1 {v20.4s}, [x9], #16 ext v24.16b, v18.16b , v20.16b , #4 ext v26.16b, v18.16b , v20.16b , #6 st1 {v22.4s}, [x9] ext v22.16b, v18.16b , v20.16b , #10 add v4.8h, v24.8h , v26.8h ext v24.16b, v18.16b , v20.16b , #2 ext v26.16b, v18.16b , v20.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v18.4h, v22.4h smlal v26.4s, v4.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v18.8h, v22.8h smlal2 v22.4s, v4.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v18.4h, v26.4s, #10 sqrshrun v19.4h, v22.4s, #10 ld1 {v22.4s}, [x9], #16 uqxtn v18.8b, v18.8h uqxtn v19.8b, v19.8h mov v18.s[1], v19.s[0] ext v24.16b, v20.16b , v22.16b , #4 ext v26.16b, v20.16b , v22.16b , #6 ext v4.16b, v20.16b , v22.16b , #10 st1 {v18.2s}, [x1] add v18.8h, v24.8h , v26.8h ext v24.16b, v20.16b , v22.16b , #2 ext v26.16b, v20.16b , v22.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v4.4h, v20.4h smlal v26.4s, v18.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v4.8h, v20.8h smlal2 v22.4s, v18.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v19.4h, v26.4s, #10 sqrshrun v18.4h, v22.4s, #10 uaddl v24.8h, v11.8b, v1.8b ld1 {v20.4s}, [x6], #16 ld1 {v22.4s}, [x6], x7 uqxtn v19.8b, v19.8h uqxtn v18.8b, v18.8h mov v19.s[1], v18.s[0] ld1 {v18.2s}, [x1] sqrshrun v20.8b, v20.8h, #5 sqrshrun v21.8b, v22.8h, #5 uaddl v22.8h, v8.8b, v2.8b ld1 {v4.2s, v5.2s}, [x0], #16 // Vector load from src[8_0] urhadd v18.16b, v18.16b , v20.16b urhadd v19.16b, v19.16b , v21.16b ld1 {v14.2s}, [x0], x2 // Vector load from src[8_0] uaddl v20.8h, v10.8b, v0.8b uaddl v26.8h, v9.8b, v3.8b st1 {v18.2s, v19.2s}, [x1], x3 // store row 2 //ROW_4 uaddl v18.8h, v6.8b, v4.8b mla v18.8h, v20.8h , v28.8h uaddl v20.8h, v7.8b, v5.8b mla v20.8h, v24.8h , v28.8h uaddl v24.8h, v17.8b, v12.8b mls v18.8h, v22.8h , v30.8h uaddl v22.8h, v15.8b, v14.8b mls v20.8h, v26.8h , v30.8h uaddl v26.8h, v16.8b, v13.8b mla v22.8h, v24.8h , v28.8h mls v22.8h, v26.8h , v30.8h st1 {v18.4s}, [x9], #16 st1 {v20.4s}, [x9], #16 ext v24.16b, v18.16b , v20.16b , #4 ext v26.16b, v18.16b , v20.16b , #6 st1 {v22.4s}, [x9] ext v22.16b, v18.16b , v20.16b , #10 add v6.8h, v24.8h , v26.8h ext v24.16b, v18.16b , v20.16b , #2 ext v26.16b, v18.16b , v20.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v18.4h, v22.4h smlal v26.4s, v6.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v18.8h, v22.8h smlal2 v22.4s, v6.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h sqrshrun v18.4h, v26.4s, #10 sqrshrun v19.4h, v22.4s, #10 ld1 {v22.4s}, [x9], #16 uqxtn v18.8b, v18.8h uqxtn v19.8b, v19.8h mov v18.s[1], v19.s[0] ext v24.16b, v20.16b , v22.16b , #4 ext v26.16b, v20.16b , v22.16b , #6 ext v6.16b, v20.16b , v22.16b , #10 st1 {v18.2s}, [x1] add v18.8h, v24.8h , v26.8h ext v24.16b, v20.16b , v22.16b , #2 ext v26.16b, v20.16b , v22.16b , #8 add v24.8h, v24.8h , v26.8h saddl v26.4s, v6.4h, v20.4h smlal v26.4s, v18.4h, v28.4h smlsl v26.4s, v24.4h, v30.4h saddl2 v22.4s, v6.8h, v20.8h smlal2 v22.4s, v18.8h, v28.8h smlsl2 v22.4s, v24.8h, v30.8h mov v6.16b, v2.16b mov v7.16b, v3.16b mov v2.16b, v10.16b mov v3.16b, v11.16b subs x4, x4, #4 sqrshrun v19.4h, v26.4s, #10 sqrshrun v18.4h, v22.4s, #10 mov v10.16b, v0.16b mov v11.16b, v1.16b mov v24.8b, v14.8b mov v14.16b, v12.16b mov v15.16b, v13.16b uqxtn v19.8b, v19.8h uqxtn v18.8b, v18.8h mov v19.s[1], v18.s[0] ld1 {v20.4s}, [x6], #16 ld1 {v22.4s}, [x6], x7 ld1 {v18.2s}, [x1] sqrshrun v20.8b, v20.8h, #5 sqrshrun v21.8b, v22.8h, #5 mov v0.16b, v8.16b mov v1.16b, v9.16b mov v8.16b, v4.16b mov v9.16b, v5.16b mov v12.16b, v16.16b mov v13.16b, v17.16b urhadd v18.16b, v18.16b , v20.16b urhadd v19.16b, v19.16b , v21.16b mov v4.16b, v10.16b mov v5.16b, v11.16b mov v16.8b, v24.8b st1 {v18.2s, v19.2s}, [x1], x3 // store row 3 bgt loop_16 // looping if height =16 b end_func loop_8_start: ld1 {v0.2s, v1.2s}, [x0], x2 // Vector load from src[0_0] ld1 {v2.2s, v3.2s}, [x0], x2 // Vector load from src[1_0] ld1 {v4.2s, v5.2s}, [x0], x2 // Vector load from src[2_0] ld1 {v6.2s, v7.2s}, [x0], x2 // Vector load from src[3_0] ld1 {v8.2s, v9.2s}, [x0], x2 // Vector load from src[4_0] loop_8: ld1 {v10.2s, v11.2s}, [x0], x2 // Vector load from src[5_0] uaddl v14.8h, v4.8b, v6.8b uaddl v12.8h, v0.8b, v10.8b uaddl v16.8h, v2.8b, v8.8b mla v12.8h, v14.8h , v26.8h uaddl v18.8h, v5.8b, v7.8b uaddl v14.8h, v1.8b, v11.8b uaddl v22.8h, v3.8b, v9.8b mla v14.8h, v18.8h , v26.8h mls v12.8h, v16.8h , v24.8h ld1 {v0.2s, v1.2s}, [x0], x2 // Vector load from src[6_0] uaddl v16.8h, v6.8b, v8.8b mls v14.8h, v22.8h , v24.8h uaddl v28.8h, v2.8b, v0.8b st1 {v12.4s}, [x9], #16 // store row 0 to temp buffer: col 0 ext v22.16b, v12.16b , v14.16b , #10 uaddl v18.8h, v4.8b, v10.8b mla v28.8h, v16.8h , v26.8h saddl v30.4s, v12.4h, v22.4h st1 {v14.4s}, [x9], x7 // store row 0 to temp buffer: col 1 saddl2 v22.4s, v12.8h, v22.8h ext v16.16b, v12.16b , v14.16b , #4 mls v28.8h, v18.8h , v24.8h ext v18.16b, v12.16b , v14.16b , #6 ext v20.16b, v12.16b , v14.16b , #8 ext v14.16b, v12.16b , v14.16b , #2 add v16.8h, v16.8h , v18.8h add v18.8h, v14.8h , v20.8h uaddl v20.8h, v7.8b, v9.8b smlal v30.4s, v16.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal2 v22.4s, v16.8h, v26.8h smlsl2 v22.4s, v18.8h, v24.8h uaddl v14.8h, v3.8b, v1.8b st1 {v28.4s}, [x9], #16 // store row 1 to temp buffer: col 0 mla v14.8h, v20.8h , v26.8h sqrshrun v12.4h, v30.4s, #10 uaddl v16.8h, v5.8b, v11.8b sqrshrun v13.4h, v22.4s, #10 mls v14.8h, v16.8h , v24.8h ld1 {v2.2s, v3.2s}, [x0], x2 // Vector load from src[7_0] uqxtn v25.8b, v12.8h uqxtn v13.8b, v13.8h mov v25.s[1], v13.s[0] uaddl v16.8h, v8.8b, v10.8b ext v22.16b, v28.16b , v14.16b , #10 uaddl v20.8h, v4.8b, v2.8b saddl v30.4s, v28.4h, v22.4h mla v20.8h, v16.8h , v26.8h st1 {v14.4s}, [x9], x7 // store row 1 to temp buffer: col 1 saddl2 v22.4s, v28.8h, v22.8h ext v16.16b, v28.16b , v14.16b , #4 ext v18.16b, v28.16b , v14.16b , #6 ext v12.16b, v28.16b , v14.16b , #8 ext v14.16b, v28.16b , v14.16b , #2 add v16.8h, v16.8h , v18.8h add v18.8h, v12.8h , v14.8h ld1 {v14.4s, v15.4s}, [x6], x8 // load row 0 from temp buffer smlal v30.4s, v16.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal2 v22.4s, v16.8h, v26.8h smlsl2 v22.4s, v18.8h, v24.8h sqrshrun v14.8b, v14.8h, #0x5 ld1 {v28.4s, v29.4s}, [x6], x8 // load row 1 from temp buffer uaddl v18.8h, v6.8b, v0.8b sqrshrun v16.4h, v30.4s, #10 sqrshrun v15.8b, v28.8h, #0x5 sqrshrun v17.4h, v22.4s, #10 mov v12.8b, v25.8b mov v25.8b, v24.8b uaddl v28.8h, v9.8b, v11.8b uqxtn v13.8b, v16.8h uqxtn v17.8b, v17.8h mov v13.s[1], v17.s[0] urhadd v12.16b, v12.16b , v14.16b urhadd v13.16b, v13.16b , v15.16b uaddl v14.8h, v5.8b, v3.8b uaddl v22.8h, v7.8b, v1.8b mls v20.8h, v18.8h , v24.8h st1 {v12.2s}, [x1], x3 // store row 0 mla v14.8h, v28.8h , v26.8h ld1 {v4.2s, v5.2s}, [x0], x2 // Vector load from src[8_0] uaddl v30.8h, v10.8b, v0.8b uaddl v28.8h, v6.8b, v4.8b mls v14.8h, v22.8h , v24.8h st1 {v13.2s}, [x1], x3 // store row 1 mla v28.8h, v30.8h , v26.8h st1 {v20.4s}, [x9], #16 // store row 2 to temp buffer: col 0 ext v22.16b, v20.16b , v14.16b , #10 saddl v30.4s, v20.4h, v22.4h st1 {v14.2s, v15.2s}, [x9], x7 // store row 2 to temp buffer: col 0 saddl2 v22.4s, v20.8h, v22.8h ext v16.16b, v20.16b , v14.16b , #4 ext v18.16b, v20.16b , v14.16b , #6 ext v12.16b, v20.16b , v14.16b , #8 ext v14.16b, v20.16b , v14.16b , #2 add v16.8h, v16.8h , v18.8h add v18.8h, v14.8h , v12.8h uaddl v20.8h, v8.8b, v2.8b smlal v30.4s, v16.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal2 v22.4s, v16.8h, v26.8h smlsl2 v22.4s, v18.8h, v24.8h uaddl v18.8h, v11.8b, v1.8b uaddl v16.8h, v7.8b, v5.8b sqrshrun v12.4h, v30.4s, #10 uaddl v30.8h, v9.8b, v3.8b mla v16.8h, v18.8h , v26.8h sqrshrun v13.4h, v22.4s, #10 mls v28.8h, v20.8h , v24.8h ld1 {v14.4s, v15.4s}, [x6], x8 // load row 2 from temp buffer mls v16.8h, v30.8h , v24.8h uqxtn v27.8b, v12.8h uqxtn v13.8b, v13.8h mov v27.s[1], v13.s[0] sqrshrun v14.8b, v14.8h, #5 ext v22.16b, v28.16b , v16.16b , #10 st1 {v28.4s}, [x9], #16 // store row 3 to temp buffer: col 0 saddl v30.4s, v28.4h, v22.4h st1 {v16.2s, v17.2s}, [x9], x7 // store row 3 to temp buffer: col 1 saddl2 v22.4s, v28.8h, v22.8h ext v12.16b, v28.16b , v16.16b , #4 ext v18.16b, v28.16b , v16.16b , #6 ext v20.16b, v28.16b , v16.16b , #8 ext v28.16b, v28.16b , v16.16b , #2 add v12.8h, v12.8h , v18.8h add v18.8h, v28.8h , v20.8h ld1 {v16.4s, v17.4s}, [x6], x8 // load row 3 from temp buffer smlal v30.4s, v12.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal2 v22.4s, v12.8h, v26.8h smlsl2 v22.4s, v18.8h, v24.8h sqrshrun v15.8b, v16.8h, #0x5 mov v12.8b, v27.8b mov v27.8b, v26.8b sqrshrun v16.4h, v30.4s, #10 mov v6.16b, v2.16b mov v7.16b, v3.16b sqrshrun v17.4h, v22.4s, #10 mov v2.16b, v10.16b mov v3.16b, v11.16b mov v10.16b, v0.16b mov v11.16b, v1.16b subs x4, x4, #4 uqxtn v13.8b, v16.8h uqxtn v17.8b, v17.8h mov v13.s[1], v17.s[0] urhadd v12.16b, v12.16b , v14.16b urhadd v13.16b, v13.16b , v15.16b mov v0.16b, v8.16b mov v1.16b, v9.16b mov v8.16b, v4.16b mov v9.16b, v5.16b mov v4.16b, v10.16b mov v5.16b, v11.16b st1 {v12.2s}, [x1], x3 // store row 2 st1 {v13.2s}, [x1], x3 // store row 3 bgt loop_8 //if height =8 loop b end_func loop_4_start: ld1 {v0.2s, v1.2s}, [x0], x2 // Vector load from src[0_0] ld1 {v2.2s, v3.2s}, [x0], x2 // Vector load from src[1_0] ld1 {v4.2s, v5.2s}, [x0], x2 // Vector load from src[2_0] ld1 {v6.2s, v7.2s}, [x0], x2 // Vector load from src[3_0] ld1 {v8.2s, v9.2s}, [x0], x2 // Vector load from src[4_0] loop_4: ld1 {v10.2s, v11.2s}, [x0], x2 // Vector load from src[5_0] uaddl v14.8h, v4.8b, v6.8b // temp1 = src[2_0] + src[3_0] uaddl v12.8h, v0.8b, v10.8b // temp = src[0_0] + src[5_0] uaddl v16.8h, v2.8b, v8.8b // temp2 = src[1_0] + src[4_0] mla v12.8h, v14.8h , v26.8h // temp += temp1 * 20 uaddl v18.8h, v5.8b, v7.8b // temp1 = src[2_0] + src[3_0] uaddl v14.8h, v1.8b, v11.8b // temp = src[0_0] + src[5_0] uaddl v22.8h, v3.8b, v9.8b // temp2 = src[1_0] + src[4_0] mla v14.8h, v18.8h , v26.8h // temp += temp1 * 20 mls v12.8h, v16.8h , v24.8h // temp -= temp2 * 5 ld1 {v0.2s, v1.2s}, [x0], x2 // Vector load from src[6_0] uaddl v16.8h, v6.8b, v8.8b mls v14.8h, v22.8h , v24.8h // temp -= temp2 * 5 //Q6 and Q7 have filtered values uaddl v28.8h, v2.8b, v0.8b st1 {v12.4s}, [x9], #16 // store row 0 to temp buffer: col 0 ext v22.16b, v12.16b , v14.16b , #10 uaddl v18.8h, v4.8b, v10.8b mla v28.8h, v16.8h , v26.8h saddl v30.4s, v12.4h, v22.4h st1 {v14.4s}, [x9], x7 // store row 0 to temp buffer: col 1 saddl v22.4s, v13.4h, v23.4h ext v16.16b, v12.16b , v14.16b , #4 mls v28.8h, v18.8h , v24.8h ext v18.16b, v12.16b , v14.16b , #6 ext v20.16b, v12.16b , v14.16b , #8 ext v14.16b, v12.16b , v14.16b , #2 add v16.8h, v16.8h , v18.8h add v18.8h, v14.8h , v20.8h uaddl v20.8h, v7.8b, v9.8b smlal v30.4s, v16.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal v22.4s, v17.4h, v26.4h smlsl v22.4s, v19.4h, v24.4h uaddl v14.8h, v3.8b, v1.8b st1 {v28.4s}, [x9], #16 // store row 1 to temp buffer: col 0 mla v14.8h, v20.8h , v26.8h sqrshrun v12.4h, v30.4s, #10 uaddl v16.8h, v5.8b, v11.8b sqrshrun v13.4h, v22.4s, #10 mls v14.8h, v16.8h , v24.8h ld1 {v2.2s, v3.2s}, [x0], x2 // Vector load from src[7_0] uqxtn v25.8b, v12.8h uaddl v16.8h, v8.8b, v10.8b ext v22.16b, v28.16b , v14.16b , #10 uaddl v20.8h, v4.8b, v2.8b saddl v30.4s, v28.4h, v22.4h mla v20.8h, v16.8h , v26.8h st1 {v14.4s}, [x9], x7 // store row 1 to temp buffer: col 1 saddl v22.4s, v29.4h, v23.4h ext v16.16b, v28.16b , v14.16b , #4 ext v18.16b, v28.16b , v14.16b , #6 ext v12.16b, v28.16b , v14.16b , #8 ext v14.16b, v28.16b , v14.16b , #2 add v16.8h, v16.8h , v18.8h add v18.8h, v12.8h , v14.8h ld1 {v14.2s}, [x6], x8 //load row 0 from temp buffer smlal v30.4s, v16.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal v22.4s, v17.4h, v26.4h smlsl v22.4s, v19.4h, v24.4h sqrshrun v14.8b, v14.8h, #0x5 ld1 {v28.2s}, [x6], x8 //load row 1 from temp buffer uaddl v18.8h, v6.8b, v0.8b sqrshrun v16.4h, v30.4s, #10 sqrshrun v15.8b, v28.8h, #0x5 sqrshrun v17.4h, v22.4s, #10 mov v12.8b, v25.8b mov v25.8b, v24.8b uaddl v28.8h, v9.8b, v11.8b uqxtn v13.8b, v16.8h urhadd v12.16b, v12.16b , v14.16b urhadd v13.16b, v13.16b , v15.16b uaddl v14.8h, v5.8b, v3.8b uaddl v22.8h, v7.8b, v1.8b mls v20.8h, v18.8h , v24.8h st1 {v12.s}[0], [x1], x3 // store row 0 mla v14.8h, v28.8h , v26.8h ld1 {v4.2s, v5.2s}, [x0], x2 // Vector load from src[8_0] uaddl v30.8h, v10.8b, v0.8b uaddl v28.8h, v6.8b, v4.8b mls v14.8h, v22.8h , v24.8h st1 {v13.s}[0], [x1], x3 //store row 1 mla v28.8h, v30.8h , v26.8h st1 {v20.4s}, [x9], #16 // store row 2 to temp buffer: col 0 ext v22.16b, v20.16b , v14.16b , #10 saddl v30.4s, v20.4h, v22.4h st1 {v14.4s}, [x9], x7 // store row 2 to temp buffer: col 1 saddl v22.4s, v21.4h, v23.4h ext v16.16b, v20.16b , v14.16b , #4 ext v18.16b, v20.16b , v14.16b , #6 ext v12.16b, v20.16b , v14.16b , #8 ext v14.16b, v20.16b , v14.16b , #2 add v16.8h, v16.8h , v18.8h add v18.8h, v14.8h , v12.8h uaddl v20.8h, v8.8b, v2.8b smlal v30.4s, v16.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal v22.4s, v17.4h, v26.4h smlsl v22.4s, v19.4h, v24.4h uaddl v18.8h, v11.8b, v1.8b uaddl v16.8h, v7.8b, v5.8b sqrshrun v12.4h, v30.4s, #10 uaddl v30.8h, v9.8b, v3.8b mla v16.8h, v18.8h , v26.8h sqrshrun v13.4h, v22.4s, #10 mls v28.8h, v20.8h , v24.8h ld1 {v14.2s}, [x6], x8 //load row 3 from temp buffer mls v16.8h, v30.8h , v24.8h uqxtn v27.8b, v12.8h sqrshrun v14.8b, v14.8h, #5 ext v22.16b, v28.16b , v16.16b , #10 st1 {v28.4s}, [x9], #16 // store row 3 to temp buffer: col 0 saddl v30.4s, v28.4h, v22.4h st1 {v16.4s}, [x9], x7 // store row 3 to temp buffer: col 1 saddl v22.4s, v29.4h, v23.4h ext v12.16b, v28.16b , v16.16b , #4 ext v18.16b, v28.16b , v16.16b , #6 ext v20.16b, v28.16b , v16.16b , #8 ext v28.16b, v28.16b , v16.16b , #2 add v12.8h, v12.8h , v18.8h add v18.8h, v28.8h , v20.8h ld1 {v16.2s}, [x6], x8 //load row 4 from temp buffer smlal v30.4s, v12.4h, v26.4h smlsl v30.4s, v18.4h, v24.4h smlal v22.4s, v13.4h, v26.4h smlsl v22.4s, v19.4h, v24.4h sqrshrun v15.8b, v16.8h, #0x5 mov v12.8b, v27.8b mov v27.8b, v26.8b sqrshrun v16.4h, v30.4s, #10 mov v6.16b, v2.16b mov v7.16b, v3.16b sqrshrun v17.4h, v22.4s, #10 mov v2.16b, v10.16b mov v3.16b, v11.16b mov v10.16b, v0.16b mov v11.16b, v1.16b subs x4, x4, #4 uqxtn v13.8b, v16.8h urhadd v12.16b, v12.16b , v14.16b urhadd v13.16b, v13.16b , v15.16b mov v0.16b, v8.16b mov v1.16b, v9.16b mov v8.16b, v4.16b mov v9.16b, v5.16b mov v4.16b, v10.16b mov v5.16b, v11.16b st1 {v12.s}[0], [x1], x3 // store row 2 st1 {v13.s}[0], [x1], x3 // store row 3 bgt loop_4 end_func: // LDMFD sp!,{x4-x12,PC} //Restoring registers from stack ldp x19, x20, [sp], #16 pop_v_regs ret