1/****************************************************************************** 2* 3* Copyright (C) 2012 Ittiam Systems Pvt Ltd, Bangalore 4* 5* Licensed under the Apache License, Version 2.0 (the "License"); 6* you may not use this file except in compliance with the License. 7* You may obtain a copy of the License at: 8* 9* http://www.apache.org/licenses/LICENSE-2.0 10* 11* Unless required by applicable law or agreed to in writing, software 12* distributed under the License is distributed on an "AS IS" BASIS, 13* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14* See the License for the specific language governing permissions and 15* limitations under the License. 16* 17******************************************************************************/ 18/** 19 ******************************************************************************* 20 * @file 21 * ihevc_chroma_iquant_itrans_recon.c 22 * 23 * @brief 24 * Contains function definitions for inverse quantization, inverse 25 * transform and reconstruction of chroma interleaved data. 26 * 27 * @author 28 * 100470 29 * 30 * @par List of Functions: 31 * - ihevc_chroma_iquant_itrans_recon_4x4() 32 * 33 * @remarks 34 * None 35 * 36 ******************************************************************************* 37 */ 38#include <stdio.h> 39#include <string.h> 40#include "ihevc_typedefs.h" 41#include "ihevc_macros.h" 42#include "ihevc_platform_macros.h" 43#include "ihevc_defs.h" 44#include "ihevc_trans_tables.h" 45#include "ihevc_chroma_iquant_itrans_recon.h" 46#include "ihevc_func_selector.h" 47#include "ihevc_trans_macros.h" 48 49/* All the functions work one component(U or V) of interleaved data depending upon pointers passed to it */ 50/* Data visualization */ 51/* U V U V U V U V */ 52/* U V U V U V U V */ 53/* U V U V U V U V */ 54/* U V U V U V U V */ 55/* If the pointer points to first byte of above stream (U) , functions will operate on U component */ 56/* If the pointer points to second byte of above stream (V) , functions will operate on V component */ 57 58 59/** 60 ******************************************************************************* 61 * 62 * @brief 63 * This function performs inverse quantization, inverse transform and 64 * reconstruction for 4x4 input block 65 * 66 * @par Description: 67 * Performs inverse quantization , inverse transform and adds the 68 * prediction data and clips output to 8 bit 69 * 70 * @param[in] pi2_src 71 * Input 4x4 coefficients 72 * 73 * @param[in] pi2_tmp 74 * Temporary 4x4 buffer for storing inverse transform 75 * 1st stage output 76 * 77 * @param[in] pu1_pred 78 * Prediction 4x4 block 79 * 80 * @param[in] pi2_dequant_coeff 81 * Dequant Coeffs 82 * 83 * @param[out] pu1_dst 84 * Output 4x4 block 85 * 86 * @param[in] qp_div 87 * Quantization parameter / 6 88 * 89 * @param[in] qp_rem 90 * Quantization parameter % 6 91 * 92 * @param[in] src_strd 93 * Input stride 94 * 95 * @param[in] pred_strd 96 * Prediction stride 97 * 98 * @param[in] dst_strd 99 * Output Stride 100 * 101 * @param[in] zero_cols 102 * Zero columns in pi2_src 103 * 104 * @param[in] zero_rows 105 * Zero Rows in pi2_src 106 * 107 * @returns Void 108 * 109 * @remarks 110 * None 111 * 112 ******************************************************************************* 113 */ 114 115 116void ihevc_chroma_iquant_itrans_recon_4x4(WORD16 *pi2_src, 117 WORD16 *pi2_tmp, 118 UWORD8 *pu1_pred, 119 WORD16 *pi2_dequant_coeff, 120 UWORD8 *pu1_dst, 121 WORD32 qp_div, /* qpscaled / 6 */ 122 WORD32 qp_rem, /* qpscaled % 6 */ 123 WORD32 src_strd, 124 WORD32 pred_strd, 125 WORD32 dst_strd, 126 WORD32 zero_cols, 127 WORD32 zero_rows) 128{ 129 UNUSED(zero_rows); 130 131 /* Inverse Transform */ 132 { 133 WORD32 j; 134 WORD32 e[2], o[2]; 135 WORD32 add; 136 WORD32 shift; 137 WORD16 *pi2_tmp_orig; 138 WORD32 shift_iq; 139 WORD32 trans_size; 140 /* Inverse Quantization constants */ 141 { 142 WORD32 log2_trans_size, bit_depth; 143 144 log2_trans_size = 2; 145 bit_depth = 8 + 0; 146 shift_iq = bit_depth + log2_trans_size - 5; 147 } 148 149 trans_size = TRANS_SIZE_4; 150 pi2_tmp_orig = pi2_tmp; 151 152 /* Inverse Transform 1st stage */ 153 shift = IT_SHIFT_STAGE_1; 154 add = 1 << (shift - 1); 155 156 for(j = 0; j < trans_size; j++) 157 { 158 /* Checking for Zero Cols */ 159 if((zero_cols & 1) == 1) 160 { 161 memset(pi2_tmp, 0, trans_size * sizeof(WORD16)); 162 } 163 else 164 { 165 WORD32 iq_tmp_1, iq_tmp_2; 166 /* Utilizing symmetry properties to the maximum to minimize the number of multiplications */ 167 IQUANT_4x4(iq_tmp_1, 168 pi2_src[1 * src_strd], 169 pi2_dequant_coeff[1 * trans_size] * g_ihevc_iquant_scales[qp_rem], 170 shift_iq, qp_div); 171 IQUANT_4x4(iq_tmp_2, 172 pi2_src[3 * src_strd], 173 pi2_dequant_coeff[3 * trans_size] * g_ihevc_iquant_scales[qp_rem], 174 shift_iq, qp_div); 175 176 o[0] = g_ai2_ihevc_trans_4[1][0] * iq_tmp_1 177 + g_ai2_ihevc_trans_4[3][0] * iq_tmp_2; 178 o[1] = g_ai2_ihevc_trans_4[1][1] * iq_tmp_1 179 + g_ai2_ihevc_trans_4[3][1] * iq_tmp_2; 180 181 IQUANT_4x4(iq_tmp_1, 182 pi2_src[0 * src_strd], 183 pi2_dequant_coeff[0 * trans_size] * g_ihevc_iquant_scales[qp_rem], 184 shift_iq, qp_div); 185 IQUANT_4x4(iq_tmp_2, 186 pi2_src[2 * src_strd], 187 pi2_dequant_coeff[2 * trans_size] * g_ihevc_iquant_scales[qp_rem], 188 shift_iq, qp_div); 189 190 e[0] = g_ai2_ihevc_trans_4[0][0] * iq_tmp_1 191 + g_ai2_ihevc_trans_4[2][0] * iq_tmp_2; 192 e[1] = g_ai2_ihevc_trans_4[0][1] * iq_tmp_1 193 + g_ai2_ihevc_trans_4[2][1] * iq_tmp_2; 194 195 pi2_tmp[0] = 196 CLIP_S16(((e[0] + o[0] + add) >> shift)); 197 pi2_tmp[1] = 198 CLIP_S16(((e[1] + o[1] + add) >> shift)); 199 pi2_tmp[2] = 200 CLIP_S16(((e[1] - o[1] + add) >> shift)); 201 pi2_tmp[3] = 202 CLIP_S16(((e[0] - o[0] + add) >> shift)); 203 } 204 pi2_src++; 205 pi2_dequant_coeff++; 206 pi2_tmp += trans_size; 207 zero_cols = zero_cols >> 1; 208 } 209 210 pi2_tmp = pi2_tmp_orig; 211 212 /* Inverse Transform 2nd stage */ 213 shift = IT_SHIFT_STAGE_2; 214 add = 1 << (shift - 1); 215 216 for(j = 0; j < trans_size; j++) 217 { 218 WORD32 itrans_out; 219 220 /* Utilizing symmetry properties to the maximum to minimize the number of multiplications */ 221 o[0] = g_ai2_ihevc_trans_4[1][0] * pi2_tmp[trans_size] 222 + g_ai2_ihevc_trans_4[3][0] 223 * pi2_tmp[3 * trans_size]; 224 o[1] = g_ai2_ihevc_trans_4[1][1] * pi2_tmp[trans_size] 225 + g_ai2_ihevc_trans_4[3][1] 226 * pi2_tmp[3 * trans_size]; 227 e[0] = g_ai2_ihevc_trans_4[0][0] * pi2_tmp[0] 228 + g_ai2_ihevc_trans_4[2][0] 229 * pi2_tmp[2 * trans_size]; 230 e[1] = g_ai2_ihevc_trans_4[0][1] * pi2_tmp[0] 231 + g_ai2_ihevc_trans_4[2][1] 232 * pi2_tmp[2 * trans_size]; 233 234 itrans_out = 235 CLIP_S16(((e[0] + o[0] + add) >> shift)); 236 pu1_dst[0 * 2] = CLIP_U8((itrans_out + pu1_pred[0 * 2])); 237 238 itrans_out = 239 CLIP_S16(((e[1] + o[1] + add) >> shift)); 240 pu1_dst[1 * 2] = CLIP_U8((itrans_out + pu1_pred[1 * 2])); 241 242 itrans_out = 243 CLIP_S16(((e[1] - o[1] + add) >> shift)); 244 pu1_dst[2 * 2] = CLIP_U8((itrans_out + pu1_pred[2 * 2])); 245 246 itrans_out = 247 CLIP_S16(((e[0] - o[0] + add) >> shift)); 248 pu1_dst[3 * 2] = CLIP_U8((itrans_out + pu1_pred[3 * 2])); 249 250 pi2_tmp++; 251 pu1_pred += pred_strd; 252 pu1_dst += dst_strd; 253 254 } 255 } 256} 257