@/***************************************************************************** @* @* 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_filters_luma_vert_w16inp.s @* @* @brief @* contains function definitions for inter prediction interpolation. @* functions are coded using neon intrinsics and can be compiled using @* rvct @* @* @author @* yogeswaran rs @* @* @par list of functions: @* @* - ihevc_inter_pred_luma_vert() @* @* @remarks @* none @* @******************************************************************************* @*/ @/* all the functions here are replicated from ihevc_inter_pred_filters.c and modified to */ @/* include reconstruction */ @ @/** @******************************************************************************* @* @* @brief @* luma vertical filter for 16bit input. @* @* @par description: @* applies a vertical filter with coefficients pointed to by 'pi1_coeff' to @* the elements pointed by 'pu1_src' and writes to the location pointed by @* 'pu1_dst' input is 16 bits the filter output is downshifted by 12 and @* clipped to lie between 0 and 255 assumptions : the function is @* optimized considering the fact width is multiple of 4. and height as @* multiple of 2. @* @* @param[in] pi2_src @* word16 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_luma_vert_w16inp(word16 *pi2_src, @ uword8 *pu1_dst, @ word32 src_strd, @ word32 dst_strd, @ word8 *pi1_coeff, @ word32 ht, @ word32 wd ) @**************variables vs registers***************************************** @ r0 => *pu2_src @ r1 => *pu1_dst @ r2 => src_strd @ r3 => dst_strd @ r4 => *pi1_coeff @ r5 => ht @ r6 => wd .equ coeff_offset, 104 .equ ht_offset, 108 .equ wd_offset, 112 .text .align 4 .globl ihevc_inter_pred_luma_vert_w16inp_w16out_a9q .type ihevc_inter_pred_luma_vert_w16inp_w16out_a9q, %function ihevc_inter_pred_luma_vert_w16inp_w16out_a9q: stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments vpush {d8 - d15} ldr r12,[sp,#coeff_offset] @load pi1_coeff mov r6,r3,lsl #1 ldr r5,[sp,#wd_offset] @load wd vld1.8 {d0},[r12] @coeff = vld1_s8(pi1_coeff) mov r2, r2, lsl #1 sub r12,r2,r2,lsl #2 @src_ctrd & pi1_coeff @vabs.s8 d0,d0 @vabs_s8(coeff) add r0,r0,r12 @r0->pu1_src r12->pi1_coeff ldr r3,[sp,#ht_offset] @load ht subs r7,r3,#0 @r3->ht @ble end_loops @end loop jump vmovl.s8 q0,d0 vdup.16 d22,d0[0] @coeffabs_0 = vdup_lane_u8(coeffabs, 0)@ vdup.16 d23,d0[1] @coeffabs_1 = vdup_lane_u8(coeffabs, 1)@ vdup.16 d24,d0[2] @coeffabs_2 = vdup_lane_u8(coeffabs, 2)@ vdup.16 d25,d0[3] @coeffabs_3 = vdup_lane_u8(coeffabs, 3)@ vdup.16 d26,d1[0] @coeffabs_4 = vdup_lane_u8(coeffabs, 4)@ vdup.16 d27,d1[1] @coeffabs_5 = vdup_lane_u8(coeffabs, 5)@ vdup.16 d28,d1[2] @coeffabs_6 = vdup_lane_u8(coeffabs, 6)@ vdup.16 d29,d1[3] @coeffabs_7 = vdup_lane_u8(coeffabs, 7)@ vmov.i32 q15,#0x80000 rsb r9,r5,r6,lsl #2 @r6->dst_strd r5 ->wd rsb r8,r5,r2,lsl #2 @r2->src_strd sub r8,r8,r5 sub r9,r9,r5 mov r3, r5, lsr #2 @divide by 4 mul r7, r3 @multiply height by width sub r7, #4 @subtract by one for epilog mov r4,r5 @r5 ->wd @mov r2, r2, lsl #1 prolog: add r3,r0,r2 @pu1_src_tmp += src_strd@ vld1.16 {d1},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vld1.16 {d0},[r0]! @src_tmp1 = vld1_u8(pu1_src_tmp)@ subs r4,r4,#4 vld1.16 {d2},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmull.s16 q4,d1,d23 @mul_res1 = vmull_u8(src_tmp2, coeffabs_1)@ vld1.16 {d3},[r3],r2 @src_tmp4 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q4,d0,d22 @mul_res1 = vmlal_u8(mul_res1, src_tmp1, coeffabs_0)@ vld1.16 {d4},[r3],r2 @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q4,d2,d24 @mul_res1 = vmlal_u8(mul_res1, src_tmp3, coeffabs_2)@ vld1.16 {d5},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q4,d3,d25 @mul_res1 = vmlal_u8(mul_res1, src_tmp4, coeffabs_3)@ vld1.16 {d6},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q4,d4,d26 @mul_res1 = vmlal_u8(mul_res1, src_tmp1, coeffabs_4)@ vld1.16 {d7},[r3],r2 @src_tmp4 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q4,d5,d27 @mul_res1 = vmlal_u8(mul_res1, src_tmp2, coeffabs_5)@ vmlal.s16 q4,d6,d28 @mul_res1 = vmlal_u8(mul_res1, src_tmp3, coeffabs_6)@ vmlal.s16 q4,d7,d29 @mul_res1 = vmlal_u8(mul_res1, src_tmp4, coeffabs_7)@ vld1.16 {d16},[r3],r2 @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmull.s16 q5,d2,d23 @mul_res2 = vmull_u8(src_tmp3, coeffabs_1)@ addle r0,r0,r8,lsl #0 vmlal.s16 q5,d1,d22 @mul_res2 = vmlal_u8(mul_res2, src_tmp2, coeffabs_0)@ movle r4,r5 @r5 ->wd vmlal.s16 q5,d3,d24 @mul_res2 = vmlal_u8(mul_res2, src_tmp4, coeffabs_2)@ vld1.16 {d17},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q5,d4,d25 @mul_res2 = vmlal_u8(mul_res2, src_tmp1, coeffabs_3)@ vld1.16 {d18},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q5,d5,d26 @mul_res2 = vmlal_u8(mul_res2, src_tmp2, coeffabs_4)@ add r3,r0,r2 @pu1_src_tmp += src_strd@ vmlal.s16 q5,d6,d27 @mul_res2 = vmlal_u8(mul_res2, src_tmp3, coeffabs_5)@ vmlal.s16 q5,d7,d28 @mul_res2 = vmlal_u8(mul_res2, src_tmp4, coeffabs_6)@ vmlal.s16 q5,d16,d29 @mul_res2 = vmlal_u8(mul_res2, src_tmp1, coeffabs_7)@ vsub.s32 q4, q4, q15 vld1.16 {d1},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmull.s16 q6,d3,d23 vld1.16 {d0},[r0]! @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q6,d2,d22 vld1.16 {d2},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q6,d4,d24 vmlal.s16 q6,d5,d25 vmlal.s16 q6,d6,d26 vmlal.s16 q6,d7,d27 vmlal.s16 q6,d16,d28 vmlal.s16 q6,d17,d29 add r14,r1,r6 vsub.s32 q5, q5, q15 vshrn.s32 d8, q4, #6 @vqrshrun.s16 d8,q4,#6 @sto_res = vqmovun_s16(sto_res_tmp)@ vmull.s16 q7,d4,d23 vmlal.s16 q7,d3,d22 vmlal.s16 q7,d5,d24 vmlal.s16 q7,d6,d25 vld1.16 {d3},[r3],r2 @src_tmp4 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d7,d26 vld1.16 {d4},[r3],r2 @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d16,d27 vld1.16 {d5},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d17,d28 vld1.16 {d6},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d18,d29 vld1.16 {d7},[r3],r2 @src_tmp4 = vld1_u8(pu1_src_tmp)@ vst1.32 {d8},[r1]! @vst1_u8(pu1_dst,sto_res)@ vsub.s32 q6, q6, q15 vshrn.s32 d10, q5, #6 @vqrshrun.s16 d10,q5,#6 @sto_res = vqmovun_s16(sto_res_tmp)@ addle r1,r1,r9 subs r7,r7,#4 blt epilog_end @jumps to epilog_end beq epilog @jumps to epilog kernel_8: vmull.s16 q4,d1,d23 @mul_res1 = vmull_u8(src_tmp2, coeffabs_1)@ subs r4,r4,#4 vmlal.s16 q4,d0,d22 @mul_res1 = vmlal_u8(mul_res1, src_tmp1, coeffabs_0)@ addle r0,r0,r8,lsl #0 vmlal.s16 q4,d2,d24 @mul_res1 = vmlal_u8(mul_res1, src_tmp3, coeffabs_2)@ vmlal.s16 q4,d3,d25 @mul_res1 = vmlal_u8(mul_res1, src_tmp4, coeffabs_3)@ vmlal.s16 q4,d4,d26 @mul_res1 = vmlal_u8(mul_res1, src_tmp1, coeffabs_4)@ vmlal.s16 q4,d5,d27 @mul_res1 = vmlal_u8(mul_res1, src_tmp2, coeffabs_5)@ vmlal.s16 q4,d6,d28 @mul_res1 = vmlal_u8(mul_res1, src_tmp3, coeffabs_6)@ vmlal.s16 q4,d7,d29 @mul_res1 = vmlal_u8(mul_res1, src_tmp4, coeffabs_7)@ vst1.32 {d10},[r14],r6 @vst1_u8(pu1_dst_tmp,sto_res)@ vsub.s32 q7, q7, q15 vshrn.s32 d12, q6, #6 @vqrshrun.s16 d12,q6,#6 vld1.16 {d16},[r3],r2 @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmull.s16 q5,d2,d23 @mul_res2 = vmull_u8(src_tmp3, coeffabs_1)@ vmlal.s16 q5,d1,d22 @mul_res2 = vmlal_u8(mul_res2, src_tmp2, coeffabs_0)@ vmlal.s16 q5,d3,d24 @mul_res2 = vmlal_u8(mul_res2, src_tmp4, coeffabs_2)@ vmlal.s16 q5,d4,d25 @mul_res2 = vmlal_u8(mul_res2, src_tmp1, coeffabs_3)@ vmlal.s16 q5,d5,d26 @mul_res2 = vmlal_u8(mul_res2, src_tmp2, coeffabs_4)@ vmlal.s16 q5,d6,d27 @mul_res2 = vmlal_u8(mul_res2, src_tmp3, coeffabs_5)@ vst1.32 {d12},[r14],r6 vmlal.s16 q5,d7,d28 @mul_res2 = vmlal_u8(mul_res2, src_tmp4, coeffabs_6)@ vld1.16 {d17},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q5,d16,d29 @mul_res2 = vmlal_u8(mul_res2, src_tmp1, coeffabs_7)@ vsub.s32 q4, q4, q15 vshrn.s32 d14, q7, #6 @vqrshrun.s16 d14,q7,#6 vmull.s16 q6,d3,d23 movle r4,r5 @r5 ->wd vmlal.s16 q6,d2,d22 vld1.16 {d18},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q6,d4,d24 add r3,r0,r2 @pu1_src_tmp += src_strd@ vmlal.s16 q6,d5,d25 vmlal.s16 q6,d6,d26 vst1.32 {d14},[r14],r6 vmlal.s16 q6,d7,d27 vld1.16 {d1},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q6,d16,d28 add r14,r1,r6 vmlal.s16 q6,d17,d29 vld1.16 {d0},[r0]! @src_tmp1 = vld1_u8(pu1_src_tmp)@ vsub.s32 q5, q5, q15 vshrn.s32 d8, q4, #6 @vqrshrun.s16 d8,q4,#6 @sto_res = vqmovun_s16(sto_res_tmp)@ vld1.16 {d2},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmull.s16 q7,d4,d23 vmlal.s16 q7,d3,d22 vmlal.s16 q7,d5,d24 vld1.16 {d3},[r3],r2 @src_tmp4 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d6,d25 vld1.16 {d4},[r3],r2 @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d7,d26 vld1.16 {d5},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d16,d27 vld1.16 {d6},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d17,d28 vld1.16 {d7},[r3],r2 @src_tmp4 = vld1_u8(pu1_src_tmp)@ vmlal.s16 q7,d18,d29 vst1.32 {d8},[r1]! @vst1_u8(pu1_dst,sto_res)@ vsub.s32 q6, q6, q15 vshrn.s32 d10, q5, #6 addle r1,r1,r9 @vqrshrun.s16 d10,q5,#6 @sto_res = vqmovun_s16(sto_res_tmp)@ subs r7,r7,#4 bgt kernel_8 @jumps to kernel_8 epilog: vmull.s16 q4,d1,d23 @mul_res1 = vmull_u8(src_tmp2, coeffabs_1)@ vmlal.s16 q4,d0,d22 @mul_res1 = vmlal_u8(mul_res1, src_tmp1, coeffabs_0)@ vmlal.s16 q4,d2,d24 @mul_res1 = vmlal_u8(mul_res1, src_tmp3, coeffabs_2)@ vmlal.s16 q4,d3,d25 @mul_res1 = vmlal_u8(mul_res1, src_tmp4, coeffabs_3)@ vmlal.s16 q4,d4,d26 @mul_res1 = vmlal_u8(mul_res1, src_tmp1, coeffabs_4)@ vmlal.s16 q4,d5,d27 @mul_res1 = vmlal_u8(mul_res1, src_tmp2, coeffabs_5)@ vmlal.s16 q4,d6,d28 @mul_res1 = vmlal_u8(mul_res1, src_tmp3, coeffabs_6)@ vmlal.s16 q4,d7,d29 @mul_res1 = vmlal_u8(mul_res1, src_tmp4, coeffabs_7)@ vst1.32 {d10},[r14],r6 vsub.s32 q7, q7, q15 vshrn.s32 d12, q6, #6 @vqrshrun.s16 d12,q6,#6 vld1.16 {d16},[r3],r2 @src_tmp1 = vld1_u8(pu1_src_tmp)@ vmull.s16 q5,d2,d23 @mul_res2 = vmull_u8(src_tmp3, coeffabs_1)@ vmlal.s16 q5,d1,d22 @mul_res2 = vmlal_u8(mul_res2, src_tmp2, coeffabs_0)@ vmlal.s16 q5,d3,d24 @mul_res2 = vmlal_u8(mul_res2, src_tmp4, coeffabs_2)@ vmlal.s16 q5,d4,d25 @mul_res2 = vmlal_u8(mul_res2, src_tmp1, coeffabs_3)@ vmlal.s16 q5,d5,d26 @mul_res2 = vmlal_u8(mul_res2, src_tmp2, coeffabs_4)@ vmlal.s16 q5,d6,d27 @mul_res2 = vmlal_u8(mul_res2, src_tmp3, coeffabs_5)@ vmlal.s16 q5,d7,d28 @mul_res2 = vmlal_u8(mul_res2, src_tmp4, coeffabs_6)@ vmlal.s16 q5,d16,d29 @mul_res2 = vmlal_u8(mul_res2, src_tmp1, coeffabs_7)@ vst1.32 {d12},[r14],r6 vsub.s32 q4, q4, q15 vshrn.s32 d14, q7, #6 @vqrshrun.s16 d14,q7,#6 vld1.16 {d17},[r3],r2 @src_tmp2 = vld1_u8(pu1_src_tmp)@ vmull.s16 q6,d3,d23 vmlal.s16 q6,d2,d22 vmlal.s16 q6,d4,d24 vmlal.s16 q6,d5,d25 vmlal.s16 q6,d6,d26 vmlal.s16 q6,d7,d27 vmlal.s16 q6,d16,d28 vmlal.s16 q6,d17,d29 vst1.32 {d14},[r14],r6 vsub.s32 q5, q5, q15 vshrn.s32 d8, q4, #6 @vqrshrun.s16 d8,q4,#6 @sto_res = vqmovun_s16(sto_res_tmp)@ vld1.16 {d18},[r3],r2 @src_tmp3 = vld1_u8(pu1_src_tmp)@ vmull.s16 q7,d4,d23 vmlal.s16 q7,d3,d22 vmlal.s16 q7,d5,d24 vmlal.s16 q7,d6,d25 vmlal.s16 q7,d7,d26 vmlal.s16 q7,d16,d27 vmlal.s16 q7,d17,d28 vmlal.s16 q7,d18,d29 vsub.s32 q6, q6, q15 vshrn.s32 d10, q5, #6 @vqrshrun.s16 d10,q5,#6 @sto_res = vqmovun_s16(sto_res_tmp)@ add r14,r1,r6 vst1.32 {d8},[r1]! @vst1_u8(pu1_dst,sto_res)@ epilog_end: vst1.32 {d10},[r14],r6 @vst1_u8(pu1_dst_tmp,sto_res)@ vshrn.s32 d12, q6, #6 @vqrshrun.s16 d12,q6,#6 vst1.32 {d12},[r14],r6 vsub.s32 q7, q7, q15 vshrn.s32 d14, q7, #6 @vqrshrun.s16 d14,q7,#6 vst1.32 {d14},[r14],r6 end_loops: vpop {d8 - d15} ldmfd sp!,{r4-r12,r15} @reload the registers from sp