//****************************************************************************** //* //* 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_chroma_av8.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_chroma_vert_bs4_av8() */ ///* ih264_deblk_chroma_vert_bslt4_av8() */ ///* ih264_deblk_chroma_horz_bs4_av8() */ ///* ih264_deblk_chroma_horz_bslt4_av8() */ ///* Issues / Problems : None */ ///* */ ///* Revision History : */ ///* */ ///* DD MM YYYY Author(s) Changes (Describe the changes made) */ ///* 28 11 2013 Ittiam Draft */ ///*****************************************************************************/ .text .p2align 2 .include "ih264_neon_macros.s" ///** //******************************************************************************* //* //* @brief //* Performs filtering of a chroma block horizontal edge when the //* boundary strength is set to 4 in high profile //* //* @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] x0 - pu1_src //* Pointer to the src sample q0 //* //* @param[in] w1 - src_strd //* Source stride //* //* @param[in] w2 - alpha_cb //* Alpha Value for the boundary in U //* //* @param[in] w3 - beta_cb //* Beta Value for the boundary in U //* //* @param[in] w4 - alpha_cr //* Alpha Value for the boundary in V //* //* @param[in] w5 - beta_cr //* Beta Value for the boundary in V //* //* @returns //* None //* //* @remarks //* None //* //******************************************************************************* //*/ .global ih264_deblk_chroma_horz_bs4_av8 ih264_deblk_chroma_horz_bs4_av8: // STMFD sp!,{x4-x6,x14} // push_v_regs stp x19, x20, [sp, #-16]! sxtw x1, w1 mov x6, x5 mov x5, x4 sub x0, x0, x1, lsl #1 //x0 = uc_edgePixel pointing to p1 of chroma ld2 {v6.8b, v7.8b}, [x0], x1 //D6 = p1u , D7 = p1v mov x4, x0 //Keeping a backup of the pointer p0 of chroma ld2 {v4.8b, v5.8b}, [x0], x1 //D4 = p0u , D5 = p0v dup v20.8b, w2 //D20 contains alpha_cb dup v21.8b, w5 //D21 contains alpha_cr mov v20.d[1], v21.d[0] ld2 {v0.8b, v1.8b}, [x0], x1 //D0 = q0u , D1 = q0v uaddl v8.8h, v6.8b, v0.8b // uaddl v10.8h, v7.8b, v1.8b //Q4,Q5 = q0 + p1 movi v31.8b, #2 // ld2 {v2.8b, v3.8b}, [x0] //D2 = q1u , D3 = q1v mov v0.d[1], v1.d[0] mov v2.d[1], v3.d[0] mov v4.d[1], v5.d[0] mov v6.d[1], v7.d[0] uabd v26.16b, v6.16b , v4.16b //Q13 = ABS(p1 - p0) umlal v8.8h, v2.8b, v31.8b // umlal v10.8h, v3.8b, v31.8b //Q5,Q4 = (X2(q1U) + q0U + p1U) uabd v22.16b, v4.16b , v0.16b //Q11 = ABS(p0 - q0) uabd v24.16b, v2.16b , v0.16b //Q12 = ABS(q1 - q0) uaddl v14.8h, v4.8b, v2.8b // uaddl v28.8h, v5.8b, v3.8b //Q14,Q7 = P0 + Q1 dup v16.8b, w3 //D16 contains beta_cb dup v17.8b, w6 //D17 contains beta_cr mov v16.d[1], v17.d[0] umlal v14.8h, v6.8b, v31.8b // umlal v28.8h, v7.8b, v31.8b //Q14,Q7 = (X2(p1U) + p0U + q1U) cmhs v18.16b, v22.16b, v20.16b cmhs v24.16b, v24.16b, v16.16b cmhs v26.16b, v26.16b, v16.16b rshrn v8.8b, v8.8h, #2 // rshrn v9.8b, v10.8h, #2 //Q4 = (X2(q1U) + q0U + p1U + 2) >> 2 mov v8.d[1], v9.d[0] orr v18.16b, v18.16b , v24.16b //Q9 = ( ABS(p0 - q0) >= Alpha ) | ( ABS(q1 - q0) >= Beta ) rshrn v10.8b, v14.8h, #2 // rshrn v11.8b, v28.8h, #2 //Q5 = (X2(p1U) + p0U + q1U + 2) >> 2 mov v10.d[1], v11.d[0] orr v18.16b, v18.16b , v26.16b //Q9 = ( ABS(p0 - q0) >= Alpha ) | ( ABS(q1 - q0) >= Beta ) | ( ABS(p1 - p0) >= Beta ) bit v10.16b, v4.16b , v18.16b // bit v8.16b, v0.16b , v18.16b // mov v11.d[0], v10.d[1] mov v9.d[0], v8.d[1] st2 {v10.8b, v11.8b}, [x4], x1 // st2 {v8.8b, v9.8b}, [x4] // // LDMFD sp!,{x4-x6,pc} // ldp x19, x20, [sp], #16 pop_v_regs ret ///** //******************************************************************************* //* //* @brief //* Performs filtering of a chroma block vertical edge when the //* boundary strength is set to 4 in high profile //* //* @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] x0 - pu1_src //* Pointer to the src sample q0 //* //* @param[in] w1 - src_strd //* Source stride //* //* @param[in] w2 - alpha_cb //* Alpha Value for the boundary in U //* //* @param[in] w3 - beta_cb //* Beta Value for the boundary in U //* //* @param[in] w4 - alpha_cr //* Alpha Value for the boundary in V //* //* @param[in] w5 - beta_cr //* Beta Value for the boundary in V //* //* @returns //* None //* //* @remarks //* None //* //******************************************************************************* //*/ .global ih264_deblk_chroma_vert_bs4_av8 ih264_deblk_chroma_vert_bs4_av8: // STMFD sp!,{x4,x5,x12,x14} push_v_regs stp x19, x20, [sp, #-16]! sxtw x1, w1 sub x0, x0, #4 //point x0 to p1u of row0. mov x12, x0 //keep a back up of x0 for buffer write add w2, w2, w4, lsl #8 //w2 = (alpha_cr,alpha_cb) add w3, w3, w5, lsl #8 //w3 = (beta_cr,beta_cb) ld4 {v0.h, v1.h, v2.h, v3.h}[0], [x0], x1 ld4 {v0.h, v1.h, v2.h, v3.h}[1], [x0], x1 ld4 {v0.h, v1.h, v2.h, v3.h}[2], [x0], x1 ld4 {v0.h, v1.h, v2.h, v3.h}[3], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[0], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[1], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[2], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[3], [x0], x1 mov v10.16b, v2.16b mov v2.16b, v1.16b mov v1.16b, v4.16b mov v4.16b, v10.16b mov v10.16b, v6.16b mov v6.16b, v3.16b mov v3.16b, v5.16b mov v5.16b, v10.16b dup v22.8h, w2 //Q11 = alpha dup v24.8h, w3 //Q12 = beta movi v31.8b, #2 mov v0.d[1], v1.d[0] mov v2.d[1], v3.d[0] mov v4.d[1], v5.d[0] mov v6.d[1], v7.d[0] uabd v8.16b, v2.16b , v4.16b //|p0-q0| uabd v10.16b, v6.16b , v4.16b //|q1-q0| uabd v12.16b, v0.16b , v2.16b //|p1-p0| uaddl v14.8h, v2.8b, v6.8b uaddl v16.8h, v3.8b, v7.8b //(p0 + q1) cmhi v8.16b, v22.16b , v8.16b //|p0-q0| < alpha ? cmhi v10.16b, v24.16b , v10.16b //|q1-q0| < beta ? cmhi v12.16b, v24.16b , v12.16b //|p1-p0| < beta ? umlal v14.8h, v0.8b, v31.8b umlal v16.8h, v1.8b, v31.8b //2*p1 + (p0 + q1) uaddl v18.8h, v0.8b, v4.8b uaddl v20.8h, v1.8b, v5.8b //(p1 + q0) and v8.16b, v8.16b , v10.16b //|p0-q0| < alpha && |q1-q0| < beta umlal v18.8h, v6.8b, v31.8b umlal v20.8h, v7.8b, v31.8b //2*q1 + (p1 + q0) rshrn v14.8b, v14.8h, #2 rshrn v15.8b, v16.8h, #2 //(2*p1 + (p0 + q1) + 2) >> 2 mov v14.d[1], v15.d[0] and v8.16b, v8.16b , v12.16b //|p0-q0| < alpha && |q1-q0| < beta && |p1-p0| < beta rshrn v18.8b, v18.8h, #2 rshrn v19.8b, v20.8h, #2 //(2*q1 + (p1 + q0) + 2) >> 2 mov v18.d[1], v19.d[0] bit v2.16b, v14.16b , v8.16b bit v4.16b, v18.16b , v8.16b mov v1.d[0], v0.d[1] mov v3.d[0], v2.d[1] mov v5.d[0], v4.d[1] mov v7.d[0], v6.d[1] mov v10.16b, v1.16b mov v1.16b, v2.16b mov v2.16b, v4.16b mov v4.16b, v10.16b mov v10.16b, v3.16b mov v3.16b, v6.16b mov v6.16b, v5.16b mov v5.16b, v10.16b st4 {v0.h, v1.h, v2.h, v3.h}[0], [x12], x1 st4 {v0.h, v1.h, v2.h, v3.h}[1], [x12], x1 st4 {v0.h, v1.h, v2.h, v3.h}[2], [x12], x1 st4 {v0.h, v1.h, v2.h, v3.h}[3], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[0], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[1], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[2], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[3], [x12], x1 // LDMFD sp!,{x4,x5,x12,pc} ldp x19, x20, [sp], #16 pop_v_regs ret ///** //******************************************************************************* //* //* @brief //* Performs filtering of a chroma block horizontal edge for cases where the //* boundary strength is less than 4 in high profile //* //* @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] x0 - pu1_src //* Pointer to the src sample q0 //* //* @param[in] w1 - src_strd //* Source stride //* //* @param[in] w2 - alpha_cb //* Alpha Value for the boundary in U //* //* @param[in] w3 - beta_cb //* Beta Value for the boundary in U //* //* @param[in] w4 - alpha_cr //* Alpha Value for the boundary in V //* //* @param[in] w5 - beta_cr //* Beta Value for the boundary in V //* //* @param[in] w6 - u4_bs //* Packed Boundary strength array //* //* @param[in] x7 - pu1_cliptab_cb //* tc0_table for U //* //* @param[in] sp(0) - pu1_cliptab_cr //* tc0_table for V //* //* @returns //* None //* //* @remarks //* None //* //******************************************************************************* //*/ .global ih264_deblk_chroma_horz_bslt4_av8 ih264_deblk_chroma_horz_bslt4_av8: // STMFD sp!,{x4-x9,x14} // push_v_regs stp x19, x20, [sp, #-16]! sxtw x1, w1 ldr x8, [sp, #80] sub x0, x0, x1, lsl #1 //x0 = uc_edgePixelU pointing to p1 of chroma U rev w6, w6 // mov v12.s[0], w6 //D12[0] = ui_Bs ld1 {v16.s}[0], [x7] //D16[0] contains cliptab_cb ld1 {v17.s}[0], [x8] //D17[0] contains cliptab_cr ld2 {v6.8b, v7.8b}, [x0], x1 //Q3=p1 tbl v14.8b, {v16.16b}, v12.8b //Retreiving cliptab values for U tbl v28.8b, {v17.16b}, v12.8b //Retrieving cliptab values for V uxtl v12.8h, v12.8b //Q6 = uc_Bs in each 16 bit scalar mov x6, x0 //Keeping a backup of the pointer to chroma U P0 ld2 {v4.8b, v5.8b}, [x0], x1 //Q2=p0 movi v30.8b, #1 // dup v20.8b, w2 //D20 contains alpha_cb dup v21.8b, w4 //D21 contains alpha_cr mov v20.d[1], v21.d[0] ld2 {v0.8b, v1.8b}, [x0], x1 //Q0=q0 uxtl v14.8h, v14.8b // uxtl v28.8h, v28.8b // mov v15.d[0], v28.d[0] //D14 has cliptab values for U, D15 for V mov v14.d[1], v28.d[0] ld2 {v2.8b, v3.8b}, [x0] //Q1=q1 usubl v10.8h, v1.8b, v5.8b // usubl v8.8h, v0.8b, v4.8b //Q5,Q4 = (q0 - p0) mov v6.d[1], v7.d[0] mov v4.d[1], v5.d[0] uabd v26.16b, v6.16b , v4.16b //Q13 = ABS(p1 - p0) shl v10.8h, v10.8h, #2 //Q5 = (q0 - p0)<<2 mov v0.d[1], v1.d[0] uabd v22.16b, v4.16b , v0.16b //Q11 = ABS(p0 - q0) shl v8.8h, v8.8h, #2 //Q4 = (q0 - p0)<<2 mov v14.d[1], v15.d[0] sli v14.8h, v14.8h, #8 mov v15.d[0], v14.d[1] mov v2.d[1], v3.d[0] uabd v24.16b, v2.16b , v0.16b //Q12 = ABS(q1 - q0) cmhs v18.16b, v22.16b, v20.16b usubl v20.8h, v6.8b, v2.8b //Q10 = (p1 - q1)L usubl v6.8h, v7.8b, v3.8b //Q3 = (p1 - q1)H dup v16.8b, w3 //Q8 contains beta_cb dup v17.8b, w5 //Q8 contains beta_cr mov v16.d[1], v17.d[0] add v8.8h, v8.8h , v20.8h // add v10.8h, v10.8h , v6.8h //Q5,Q4 = [ (q0 - p0)<<2 ] + (p1 - q1) cmhs v24.16b, v24.16b, v16.16b cmgt v12.4h, v12.4h, #0 sqrshrn v8.8b, v8.8h, #3 // sqrshrn v9.8b, v10.8h, #3 //Q4 = i_macro = (((q0 - p0)<<2) + (p1 - q1) + 4)>>3 mov v8.d[1], v9.d[0] add v14.8b, v14.8b , v30.8b //D14 = C = C0+1 for U cmhs v26.16b, v26.16b, v16.16b orr v18.16b, v18.16b , v24.16b //Q9 = ( ABS(p0 - q0) >= Alpha ) | ( ABS(q1 - q0) >= Beta ) abs v6.16b, v8.16b //Q4 = ABS (i_macro) add v15.8b, v15.8b , v30.8b //D15 = C = C0+1 for V mov v14.d[1], v15.d[0] mov v13.8b, v12.8b mov v12.d[1], v13.d[0] // orr v18.16b, v18.16b , v26.16b //Q9 = ( ABS(p0 - q0) >= Alpha ) | ( ABS(q1 - q0) >= Beta ) | ( ABS(p1 - p0) >= Beta ) umin v14.16b, v6.16b , v14.16b //Q7 = delta = (ABS(i_macro) > C) ? C : ABS(i_macro) bic v12.16b, v12.16b , v18.16b //final condition cmge v8.16b, v8.16b, #0 and v14.16b, v14.16b , v12.16b //Making delta zero in places where values shouldn be filterd uqadd v16.16b, v4.16b , v14.16b //Q8 = p0 + delta uqsub v4.16b, v4.16b , v14.16b //Q2 = p0 - delta uqadd v18.16b, v0.16b , v14.16b //Q9 = q0 + delta uqsub v0.16b, v0.16b , v14.16b //Q0 = q0 - delta bif v16.16b, v4.16b , v8.16b //Q8 = (i_macro >= 0 ) ? (p0+delta) : (p0-delta) bif v0.16b, v18.16b , v8.16b //Q0 = (i_macro >= 0 ) ? (q0-delta) : (q0+delta) mov v17.d[0], v16.d[1] mov v1.d[0], v0.d[1] st2 {v16.8b, v17.8b}, [x6], x1 // st2 {v0.8b, v1.8b}, [x6] // ldp x19, x20, [sp], #16 pop_v_regs ret ///** //******************************************************************************* //* //* @brief //* Performs filtering of a chroma block vertical edge for cases where the //* boundary strength is less than 4 in high profile //* //* @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] x0 - pu1_src //* Pointer to the src sample q0 //* //* @param[in] w1 - src_strd //* Source stride //* //* @param[in] w2 - alpha_cb //* Alpha Value for the boundary in U //* //* @param[in] w3 - beta_cb //* Beta Value for the boundary in U //* //* @param[in] w4 - alpha_cr //* Alpha Value for the boundary in V //* //* @param[in] w5 - beta_cr //* Beta Value for the boundary in V //* //* @param[in] w6 - u4_bs //* Packed Boundary strength array //* //* @param[in] x7 - pu1_cliptab_cb //* tc0_table for U //* //* @param[in] sp(0) - pu1_cliptab_cr //* tc0_table for V //* //* @returns //* None //* //* @remarks //* None //* //******************************************************************************* //*/ .global ih264_deblk_chroma_vert_bslt4_av8 ih264_deblk_chroma_vert_bslt4_av8: // STMFD sp!,{x4-x7,x10-x12,x14} push_v_regs stp x19, x20, [sp, #-16]! sxtw x1, w1 mov x10, x7 ldr x11, [sp, #80] //x11 = u4_bs sub x0, x0, #4 //point x0 to p1u of row0. add w2, w2, w4, lsl #8 add w3, w3, w5, lsl #8 mov x12, x0 //keep a back up of x0 for buffer write ld4 {v0.h, v1.h, v2.h, v3.h}[0], [x0], x1 ld4 {v0.h, v1.h, v2.h, v3.h}[1], [x0], x1 ld4 {v0.h, v1.h, v2.h, v3.h}[2], [x0], x1 ld4 {v0.h, v1.h, v2.h, v3.h}[3], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[0], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[1], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[2], [x0], x1 ld4 {v4.h, v5.h, v6.h, v7.h}[3], [x0], x1 mov v10.16b, v2.16b mov v2.16b, v1.16b mov v1.16b, v4.16b mov v4.16b, v10.16b mov v10.16b, v6.16b mov v6.16b, v3.16b mov v3.16b, v5.16b mov v5.16b, v10.16b dup v22.8h, w2 //Q11 = alpha mov v2.d[1], v3.d[0] mov v4.d[1], v5.d[0] uabd v8.16b, v2.16b , v4.16b //|p0-q0| dup v24.8h, w3 //Q12 = beta mov v25.d[0], v24.d[1] mov v6.d[1], v7.d[0] mov v0.d[1], v1.d[0] uabd v10.16b, v6.16b , v4.16b //|q1-q0| uabd v12.16b, v0.16b , v2.16b //|p1-p0| cmhi v8.16b, v22.16b , v8.16b //|p0-q0| < alpha ? usubl v14.8h, v0.8b, v6.8b cmhi v10.16b, v24.16b , v10.16b //|q1-q0| < beta ? usubl v16.8h, v1.8b, v7.8b //(p1 - q1) cmhi v12.16b, v24.16b , v12.16b //|p1-p0| < beta ? usubl v18.8h, v4.8b, v2.8b and v8.16b, v8.16b , v10.16b //|p0-q0| < alpha && |q1-q0| < beta usubl v20.8h, v5.8b, v3.8b //(q0 - p0) movi v28.8h, #4 ld1 {v24.s}[0], [x10] //Load ClipTable for U ld1 {v25.s}[0], [x11] //Load ClipTable for V rev w6, w6 //Blocking strengths and v8.16b, v8.16b , v12.16b //|p0-q0| < alpha && |q1-q0| < beta && |p1-p0| < beta mov v10.s[0], w6 mla v14.8h, v18.8h , v28.8h mla v16.8h, v20.8h , v28.8h //4*(q0 - p0) + (p1 - q1) uxtl v10.8h, v10.8b sli v10.4h, v10.4h, #8 tbl v12.8b, {v24.16b}, v10.8b //tC0 for U tbl v13.8b, {v25.16b}, v10.8b //tC0 for V zip1 v31.8b, v12.8b, v13.8b zip2 v13.8b, v12.8b, v13.8b mov v12.8b, v31.8b mov v12.d[1], v13.d[0] uxtl v10.4s, v10.4h sli v10.4s, v10.4s, #16 movi v24.16b, #1 add v12.16b, v12.16b , v24.16b //tC0 + 1 cmhs v10.16b, v10.16b , v24.16b and v8.16b, v8.16b , v10.16b //|p0-q0| < alpha && |q1-q0| < beta && |p1-p0| < beta && u4_bs != 0 // Q0 - Q3(inputs), // Q4 (|p0-q0| < alpha && |q1-q0| < beta && |p1-p0| < beta && u4_bs != 0), // Q6 (tC) srshr v14.8h, v14.8h, #3 srshr v16.8h, v16.8h, #3 //(((q0 - p0) << 2) + (p1 - q1) + 4) >> 3) cmgt v18.8h, v14.8h , #0 cmgt v20.8h, v16.8h , #0 xtn v18.8b, v18.8h xtn v19.8b, v20.8h //Q9 = sign(delta) mov v18.d[1], v19.d[0] abs v14.8h, v14.8h abs v16.8h, v16.8h xtn v14.8b, v14.8h xtn v15.8b, v16.8h mov v14.d[1], v15.d[0] umin v14.16b, v14.16b , v12.16b //Q7 = |delta| uqadd v20.16b, v2.16b , v14.16b //p0+|delta| uqadd v22.16b, v4.16b , v14.16b //q0+|delta| uqsub v24.16b, v2.16b , v14.16b //p0-|delta| uqsub v26.16b, v4.16b , v14.16b //q0-|delta| bit v24.16b, v20.16b , v18.16b //p0 + delta bit v22.16b, v26.16b , v18.16b //q0 - delta bit v2.16b, v24.16b , v8.16b bit v4.16b, v22.16b , v8.16b mov v1.d[0], v0.d[1] mov v3.d[0], v2.d[1] mov v5.d[0], v4.d[1] mov v7.d[0], v6.d[1] mov v10.16b, v1.16b mov v1.16b, v2.16b mov v2.16b, v4.16b mov v4.16b, v10.16b mov v10.16b, v3.16b mov v3.16b, v6.16b mov v6.16b, v5.16b mov v5.16b, v10.16b st4 {v0.h, v1.h, v2.h, v3.h}[0], [x12], x1 st4 {v0.h, v1.h, v2.h, v3.h}[1], [x12], x1 st4 {v0.h, v1.h, v2.h, v3.h}[2], [x12], x1 st4 {v0.h, v1.h, v2.h, v3.h}[3], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[0], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[1], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[2], [x12], x1 st4 {v4.h, v5.h, v6.h, v7.h}[3], [x12], x1 ldp x19, x20, [sp], #16 pop_v_regs ret