@/***************************************************************************** @* @* 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_sao_band_offset_chroma.s @* @* ,:brief @* Contains function definitions for inter prediction interpolation. @* Functions are coded using NEON intrinsics and can be compiled using@ ARM @* RVCT @* @* ,:author @* Parthiban V @* @* ,:par List of Functions: @* @* @* ,:remarks @* None @* @******************************************************************************* @*/ @void ihevc_sao_band_offset_chroma(UWORD8 *pu1_src, @ WORD32 src_strd, @ UWORD8 *pu1_src_left, @ UWORD8 *pu1_src_top, @ UWORD8 *pu1_src_top_left, @ WORD32 sao_band_pos_u, @ WORD32 sao_band_pos_v, @ WORD8 *pi1_sao_offset_u, @ WORD8 *pi1_sao_offset_v, @ WORD32 wd, @ WORD32 ht) @ @**************Variables Vs Registers***************************************** @r0 => *pu1_src @r1 => src_strd @r2 => *pu1_src_left @r3 => *pu1_src_top @r4 => *pu1_src_top_left @r5 => sao_band_pos_u @r6 => sao_band_pos_v @r7 => *pi1_sao_offset_u @r8 => *pi1_sao_offset_v @r9 => wd @r10=> ht .text .p2align 2 .extern gu1_table_band_idx .globl ihevc_sao_band_offset_chroma_a9q gu1_table_band_idx_addr_1: .long gu1_table_band_idx - ulbl1 - 8 gu1_table_band_idx_addr_2: .long gu1_table_band_idx - ulbl2 - 8 ihevc_sao_band_offset_chroma_a9q: STMFD sp!, {r4-r12, r14} @stack stores the values of the arguments LDR r4,[sp,#40] @Loads pu1_src_top_left LDR r10,[sp,#64] @Loads ht LDR r9,[sp,#60] @Loads wd MOV r11,r10 @Move the ht to r9 for loop counter ADD r12,r0,r9 @pu1_src[row * src_strd + (wd)] LDR r14, gu1_table_band_idx_addr_1 ulbl1: add r14,r14,pc SUB r12,r12,#2 @wd-2 SRC_LEFT_LOOP: LDRH r5,[r12],r1 @Load the value SUBS r11,r11,#1 @Decrement the loop counter STRH r5,[r2],#2 @Store the value in pu1_src_left pointer BNE SRC_LEFT_LOOP LDR r5,[sp,#44] @Loads sao_band_pos_u VLD1.8 D1,[r14]! @band_table_u.val[0] ADD r12,r3,r9 @pu1_src_top[wd] LDRH r11,[r12,#-2] VLD1.8 D2,[r14]! @band_table_u.val[1] LSL r6,r5,#3 @sao_band_pos_u STRH r11,[r4] @store to pu1_src_top_left[0] VLD1.8 D3,[r14]! @band_table_u.val[2] LDR r7,[sp,#52] @Loads pi1_sao_offset_u SUB r4,r10,#1 @ht-1 VDUP.8 D31,r6 @band_pos_u MUL r4,r4,r1 @ht-1 * src_strd ADD r4,r4,r0 @pu1_src[(ht - 1) * src_strd] VLD1.8 D4,[r14]! @band_table_u.val[3] MOV r11,r9 @Move the wd to r9 for loop counter SRC_TOP_LOOP: @wd is always multiple of 8 VLD1.8 D0,[r4]! @Load pu1_src[(ht - 1) * src_strd + col] SUBS r11,r11,#8 @Decrement the loop counter by 8 VST1.8 D0,[r3]! @Store to pu1_src_top[col] BNE SRC_TOP_LOOP VLD1.8 D30,[r7] @pi1_sao_offset_u load VADD.I8 D5,D1,D31 @band_table_u.val[0] = vadd_u8(band_table_u.val[0], sao_band_pos_u) VDUP.8 D29,D30[1] @vdup_n_u8(pi1_sao_offset_u[1]) VADD.I8 D6,D2,D31 @band_table_u.val[1] = vadd_u8(band_table_u.val[1], sao_band_pos_u) VDUP.8 D28,D30[2] @vdup_n_u8(pi1_sao_offset_u[2]) VADD.I8 D7,D3,D31 @band_table_u.val[2] = vadd_u8(band_table_u.val[2], sao_band_pos_u) VDUP.8 D27,D30[3] @vdup_n_u8(pi1_sao_offset_u[3]) VADD.I8 D8,D4,D31 @band_table_u.val[3] = vadd_u8(band_table_u.val[3], sao_band_pos_u) CMP r5,#28 VDUP.8 D26,D30[4] @vdup_n_u8(pi1_sao_offset_u[4]) LDR r14, gu1_table_band_idx_addr_2 ulbl2: add r14,r14,pc VMOV.I8 D30,#16 @vdup_n_u8(16) VADD.I8 D1,D5,D29 @band_table_u.val[0] = vadd_u8(band_table_u.val[0], vdup_n_u8(pi1_sao_offset_u[1])) VLD1.8 D9,[r14]! @band_table_v.val[0] VADD.I8 D2,D6,D28 @band_table_u.val[1] = vadd_u8(band_table_u.val[1], vdup_n_u8(pi1_sao_offset_u[2])) VLD1.8 D10,[r14]! @band_table_v.val[1] VADD.I8 D3,D7,D27 @band_table_u.val[2] = vadd_u8(band_table_u.val[2], vdup_n_u8(pi1_sao_offset_u[3])) LDR r6,[sp,#48] @Loads sao_band_pos_v VADD.I8 D4,D8,D26 @band_table_u.val[3] = vadd_u8(band_table_u.val[3], vdup_n_u8(pi1_sao_offset_u[4])) LSL r11,r6,#3 @sao_band_pos_v BLT SAO_BAND_POS_U_0 SAO_BAND_POS_U_28: @case 28 VCLE.U8 D13,D4,D30 @vcle_u8(band_table.val[3], vdup_n_u8(16)) BNE SAO_BAND_POS_U_29 VORR.U8 D4,D4,D13 @band_table.val[3] = vorr_u8(band_table.val[3], au1_cmp) B SWITCH_BREAK_U SAO_BAND_POS_U_29: @case 29 CMP r5,#29 VCLE.U8 D14,D3,D30 @vcle_u8(band_table.val[2], vdup_n_u8(16)) BNE SAO_BAND_POS_U_30 VORR.U8 D3,D3,D14 @band_table.val[2] = vorr_u8(band_table.val[2], au1_cmp) VAND.U8 D4,D4,D13 @band_table.val[3] = vand_u8(band_table.val[3], au1_cmp) B SWITCH_BREAK_U SAO_BAND_POS_U_30: @case 30 CMP r5,#30 VCLE.U8 D15,D2,D30 @vcle_u8(band_table.val[1], vdup_n_u8(16)) BNE SAO_BAND_POS_U_31 VORR.U8 D2,D2,D15 @band_table.val[1] = vorr_u8(band_table.val[1], au1_cmp) VAND.U8 D3,D3,D14 @band_table.val[2] = vand_u8(band_table.val[2], au1_cmp) SAO_BAND_POS_U_31: @case 31 CMP r5,#31 BNE SWITCH_BREAK_U VCLE.U8 D16,D1,D30 @vcle_u8(band_table.val[0], vdup_n_u8(16)) VORR.U8 D1,D1,D16 @band_table.val[0] = vorr_u8(band_table.val[0], au1_cmp) VAND.U8 D2,D2,D15 @band_table.val[1] = vand_u8(band_table.val[1], au1_cmp) B SWITCH_BREAK_U SAO_BAND_POS_U_0: CMP r5,#0 @case 0 BNE SWITCH_BREAK_U VCLE.U8 D16,D1,D30 @vcle_u8(band_table.val[0], vdup_n_u8(16)) VAND.U8 D1,D1,D16 @band_table.val[0] = vand_u8(band_table.val[0], au1_cmp) SWITCH_BREAK_U: VDUP.8 D30,r11 @band_pos_v LDR r8,[sp,#56] @Loads pi1_sao_offset_v VLD1.8 D11,[r14]! @band_table_v.val[2] VADD.I8 D13,D9,D30 @band_table_v.val[0] = vadd_u8(band_table_v.val[0], band_pos_v) VLD1.8 D12,[r14]! @band_table_v.val[3] VADD.I8 D14,D10,D30 @band_table_v.val[1] = vadd_u8(band_table_v.val[1], band_pos_v) VLD1.8 D25,[r8] @pi1_sao_offset_v load VADD.I8 D15,D11,D30 @band_table_v.val[2] = vadd_u8(band_table_v.val[2], band_pos_v) VDUP.8 D29,D25[1] @vdup_n_u8(pi1_sao_offset_v[1]) VADD.I8 D16,D12,D30 @band_table_v.val[3] = vadd_u8(band_table_v.val[3], band_pos_v) VDUP.8 D28,D25[2] @vdup_n_u8(pi1_sao_offset_v[2]) VADD.I8 D9,D13,D29 @band_table_v.val[0] = vadd_u8(band_table_v.val[0], vdup_n_u8(pi1_sao_offset_v[1])) VDUP.8 D27,D25[3] @vdup_n_u8(pi1_sao_offset_v[3]) VADD.I8 D10,D14,D28 @band_table_v.val[1] = vadd_u8(band_table_v.val[1], vdup_n_u8(pi1_sao_offset_v[2])) VDUP.8 D26,D25[4] @vdup_n_u8(pi1_sao_offset_v[4]) VADD.I8 D11,D15,D27 @band_table_v.val[2] = vadd_u8(band_table_v.val[2], vdup_n_u8(pi1_sao_offset_v[3])) VMOV.I8 D29,#16 @vdup_n_u8(16) VADD.I8 D12,D16,D26 @band_table_v.val[3] = vadd_u8(band_table_v.val[3], vdup_n_u8(pi1_sao_offset_v[4])) AND r12,r9,#0xf CMP r6,#28 BLT SAO_BAND_POS_V_0 SAO_BAND_POS_V_28: @case 28 VCLE.U8 D17,D12,D29 @vcle_u8(band_table.val[3], vdup_n_u8(16)) BNE SAO_BAND_POS_V_29 VORR.U8 D12,D12,D17 @band_table.val[3] = vorr_u8(band_table.val[3], au1_cmp) B SWITCH_BREAK_V SAO_BAND_POS_V_29: @case 29 CMP r6,#29 VCLE.U8 D18,D11,D29 @vcle_u8(band_table.val[2], vdup_n_u8(16)) BNE SAO_BAND_POS_V_30 VORR.U8 D11,D11,D18 @band_table.val[2] = vorr_u8(band_table.val[2], au1_cmp) VAND.U8 D12,D12,D17 @band_table.val[3] = vand_u8(band_table.val[3], au1_cmp) B SWITCH_BREAK_V SAO_BAND_POS_V_30: @case 30 CMP r6,#30 VCLE.U8 D19,D10,D29 @vcle_u8(band_table.val[1], vdup_n_u8(16)) BNE SAO_BAND_POS_V_31 VORR.U8 D10,D10,D19 @band_table.val[1] = vorr_u8(band_table.val[1], au1_cmp) VAND.U8 D11,D11,D18 @band_table.val[2] = vand_u8(band_table.val[2], au1_cmp) B SWITCH_BREAK_V SAO_BAND_POS_V_31: @case 31 CMP r6,#31 BNE SWITCH_BREAK_V VCLE.U8 D20,D9,D29 @vcle_u8(band_table.val[0], vdup_n_u8(16)) VORR.U8 D9,D9,D20 @band_table.val[0] = vorr_u8(band_table.val[0], au1_cmp) VAND.U8 D10,D10,D19 @band_table.val[1] = vand_u8(band_table.val[1], au1_cmp) B SWITCH_BREAK_V SAO_BAND_POS_V_0: CMP r6,#0 @case 0 BNE SWITCH_BREAK_V VCLE.U8 D20,D9,D29 @vcle_u8(band_table.val[0], vdup_n_u8(16)) VAND.U8 D9,D9,D20 @band_table.val[0] = vand_u8(band_table.val[0], au1_cmp) SWITCH_BREAK_V: CMP r9,#16 MOV r4,r0 @pu1_src_cpy BLT WIDTH_RESIDUE WIDTH_LOOP: @Width is assigned to be multiple of 16 MOV r4,r0 @pu1_src_cpy MOV r11,r10 @move ht ADD r5,r4,r1 HEIGHT_LOOP: @unrolled for 4 rows ADD r6,r5,r1 VLD2.8 {D5,D6},[r4] @vld1q_u8(pu1_src_cpy) ADD r7,r6,r1 VLD2.8 {D13,D14},[r5] @vld1q_u8(pu1_src_cpy) VSUB.I8 D7,D5,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VLD2.8 {D17,D18},[r6] @vld1q_u8(pu1_src_cpy) VSUB.I8 D8,D6,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VLD2.8 {D21,D22},[r7] @vld1q_u8(pu1_src_cpy) VSUB.I8 D15,D13,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VTBX.8 D5,{D1-D4},D7 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VSUB.I8 D16,D14,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VTBX.8 D6,{D9-D12},D8 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VSUB.I8 D19,D17,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VTBX.8 D13,{D1-D4},D15 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VSUB.I8 D20,D18,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VTBX.8 D14,{D9-D12},D16 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VSUB.I8 D23,D21,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VST2.8 {D5,D6},[r4] @vst1q_u8(pu1_src_cpy, au1_cur_row) VSUB.I8 D24,D22,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) SUBS r11,r11,#4 @Decrement the ht loop count by 4 VTBX.8 D17,{D1-D4},D19 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VST2.8 {D13,D14},[r5] @vst1q_u8(pu1_src_cpy, au1_cur_row) VTBX.8 D18,{D9-D12},D20 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VTBX.8 D21,{D1-D4},D23 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VTBX.8 D22,{D9-D12},D24 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VST2.8 {D17,D18},[r6],r1 @vst1q_u8(pu1_src_cpy, au1_cur_row) ADD r4,r6,r1 VST2.8 {D21,D22},[r7] @vst1q_u8(pu1_src_cpy, au1_cur_row) ADD r5,r4,r1 BNE HEIGHT_LOOP SUB r9,r9,#16 @Decrement the width loop by 16 ADD r0,r0,#16 CMP r9,#8 BGT WIDTH_LOOP BLT END_LOOP MOV r4,r0 @pu1_src_cpy WIDTH_RESIDUE: @If width is not multiple of 16 ADD r5,r4,r1 VLD2.8 {D5,D6},[r4] @vld1q_u8(pu1_src_cpy) ADD r6,r5,r1 ADD r7,r6,r1 VLD2.8 {D13,D14},[r5] @vld1q_u8(pu1_src_cpy) VSUB.I8 D7,D5,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VLD2.8 {D17,D18},[r6] @vld1q_u8(pu1_src_cpy) VSUB.I8 D8,D6,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VTBX.8 D5,{D1-D4},D7 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VSUB.I8 D15,D13,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VTBX.8 D6,{D9-D12},D8 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VSUB.I8 D16,D14,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VLD2.8 {D21,D22},[r7] @vld1q_u8(pu1_src_cpy) VSUB.I8 D19,D17,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VTBX.8 D13,{D1-D4},D15 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VSUB.I8 D20,D18,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VTBX.8 D14,{D9-D12},D16 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VZIP.8 D5,D6 VTBX.8 D17,{D1-D4},D19 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VSUB.I8 D23,D21,D31 @vsub_u8(au1_cur_row_deint.val[0], band_pos_u) VST1.8 {D5},[r4] @vst1q_u8(pu1_src_cpy, au1_cur_row) VZIP.8 D13,D14 VTBX.8 D18,{D9-D12},D20 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VSUB.I8 D24,D22,D30 @vsub_u8(au1_cur_row_deint.val[1], band_pos_v) VST1.8 {D13},[r5] @vst1q_u8(pu1_src_cpy, au1_cur_row) SUBS r10,r10,#4 @Decrement the ht loop count by 4 VTBX.8 D21,{D1-D4},D23 @vtbx4_u8(au1_cur_row_deint.val[0], band_table_u, vsub_u8(au1_cur_row_deint.val[0], band_pos_u)) VZIP.8 D17,D18 VTBX.8 D22,{D9-D12},D24 @vtbx4_u8(au1_cur_row_deint.val[1], band_table_v, vsub_u8(au1_cur_row_deint.val[1], band_pos_v)) VST1.8 {D17},[r6],r1 @vst1q_u8(pu1_src_cpy, au1_cur_row) VZIP.8 D21,D22 ADD r4,r6,r1 VST1.8 {D21},[r7] @vst1q_u8(pu1_src_cpy, au1_cur_row) ADD r5,r4,r1 BNE WIDTH_RESIDUE END_LOOP: LDMFD sp!,{r4-r12,r15} @Reload the registers from SP