Libav
|
00001 /* 00002 * Simple IDCT (Alpha optimized) 00003 * 00004 * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at> 00005 * 00006 * based upon some outcommented C code from mpeg2dec (idct_mmx.c 00007 * written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>) 00008 * 00009 * Alpha optimizations by Måns Rullgård <mans@mansr.com> 00010 * and Falk Hueffner <falk@debian.org> 00011 * 00012 * This file is part of FFmpeg. 00013 * 00014 * FFmpeg is free software; you can redistribute it and/or 00015 * modify it under the terms of the GNU Lesser General Public 00016 * License as published by the Free Software Foundation; either 00017 * version 2.1 of the License, or (at your option) any later version. 00018 * 00019 * FFmpeg is distributed in the hope that it will be useful, 00020 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00021 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00022 * Lesser General Public License for more details. 00023 * 00024 * You should have received a copy of the GNU Lesser General Public 00025 * License along with FFmpeg; if not, write to the Free Software 00026 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00027 */ 00028 00029 #include "libavcodec/dsputil.h" 00030 #include "dsputil_alpha.h" 00031 #include "asm.h" 00032 00033 // cos(i * M_PI / 16) * sqrt(2) * (1 << 14) 00034 // W4 is actually exactly 16384, but using 16383 works around 00035 // accumulating rounding errors for some encoders 00036 #define W1 ((int_fast32_t) 22725) 00037 #define W2 ((int_fast32_t) 21407) 00038 #define W3 ((int_fast32_t) 19266) 00039 #define W4 ((int_fast32_t) 16383) 00040 #define W5 ((int_fast32_t) 12873) 00041 #define W6 ((int_fast32_t) 8867) 00042 #define W7 ((int_fast32_t) 4520) 00043 #define ROW_SHIFT 11 00044 #define COL_SHIFT 20 00045 00046 /* 0: all entries 0, 1: only first entry nonzero, 2: otherwise */ 00047 static inline int idct_row(DCTELEM *row) 00048 { 00049 int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3, t; 00050 uint64_t l, r, t2; 00051 l = ldq(row); 00052 r = ldq(row + 4); 00053 00054 if (l == 0 && r == 0) 00055 return 0; 00056 00057 a0 = W4 * sextw(l) + (1 << (ROW_SHIFT - 1)); 00058 00059 if (((l & ~0xffffUL) | r) == 0) { 00060 a0 >>= ROW_SHIFT; 00061 t2 = (uint16_t) a0; 00062 t2 |= t2 << 16; 00063 t2 |= t2 << 32; 00064 00065 stq(t2, row); 00066 stq(t2, row + 4); 00067 return 1; 00068 } 00069 00070 a1 = a0; 00071 a2 = a0; 00072 a3 = a0; 00073 00074 t = extwl(l, 4); /* row[2] */ 00075 if (t != 0) { 00076 t = sextw(t); 00077 a0 += W2 * t; 00078 a1 += W6 * t; 00079 a2 -= W6 * t; 00080 a3 -= W2 * t; 00081 } 00082 00083 t = extwl(r, 0); /* row[4] */ 00084 if (t != 0) { 00085 t = sextw(t); 00086 a0 += W4 * t; 00087 a1 -= W4 * t; 00088 a2 -= W4 * t; 00089 a3 += W4 * t; 00090 } 00091 00092 t = extwl(r, 4); /* row[6] */ 00093 if (t != 0) { 00094 t = sextw(t); 00095 a0 += W6 * t; 00096 a1 -= W2 * t; 00097 a2 += W2 * t; 00098 a3 -= W6 * t; 00099 } 00100 00101 t = extwl(l, 2); /* row[1] */ 00102 if (t != 0) { 00103 t = sextw(t); 00104 b0 = W1 * t; 00105 b1 = W3 * t; 00106 b2 = W5 * t; 00107 b3 = W7 * t; 00108 } else { 00109 b0 = 0; 00110 b1 = 0; 00111 b2 = 0; 00112 b3 = 0; 00113 } 00114 00115 t = extwl(l, 6); /* row[3] */ 00116 if (t) { 00117 t = sextw(t); 00118 b0 += W3 * t; 00119 b1 -= W7 * t; 00120 b2 -= W1 * t; 00121 b3 -= W5 * t; 00122 } 00123 00124 00125 t = extwl(r, 2); /* row[5] */ 00126 if (t) { 00127 t = sextw(t); 00128 b0 += W5 * t; 00129 b1 -= W1 * t; 00130 b2 += W7 * t; 00131 b3 += W3 * t; 00132 } 00133 00134 t = extwl(r, 6); /* row[7] */ 00135 if (t) { 00136 t = sextw(t); 00137 b0 += W7 * t; 00138 b1 -= W5 * t; 00139 b2 += W3 * t; 00140 b3 -= W1 * t; 00141 } 00142 00143 row[0] = (a0 + b0) >> ROW_SHIFT; 00144 row[1] = (a1 + b1) >> ROW_SHIFT; 00145 row[2] = (a2 + b2) >> ROW_SHIFT; 00146 row[3] = (a3 + b3) >> ROW_SHIFT; 00147 row[4] = (a3 - b3) >> ROW_SHIFT; 00148 row[5] = (a2 - b2) >> ROW_SHIFT; 00149 row[6] = (a1 - b1) >> ROW_SHIFT; 00150 row[7] = (a0 - b0) >> ROW_SHIFT; 00151 00152 return 2; 00153 } 00154 00155 static inline void idct_col(DCTELEM *col) 00156 { 00157 int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3; 00158 00159 col[0] += (1 << (COL_SHIFT - 1)) / W4; 00160 00161 a0 = W4 * col[8 * 0]; 00162 a1 = W4 * col[8 * 0]; 00163 a2 = W4 * col[8 * 0]; 00164 a3 = W4 * col[8 * 0]; 00165 00166 if (col[8 * 2]) { 00167 a0 += W2 * col[8 * 2]; 00168 a1 += W6 * col[8 * 2]; 00169 a2 -= W6 * col[8 * 2]; 00170 a3 -= W2 * col[8 * 2]; 00171 } 00172 00173 if (col[8 * 4]) { 00174 a0 += W4 * col[8 * 4]; 00175 a1 -= W4 * col[8 * 4]; 00176 a2 -= W4 * col[8 * 4]; 00177 a3 += W4 * col[8 * 4]; 00178 } 00179 00180 if (col[8 * 6]) { 00181 a0 += W6 * col[8 * 6]; 00182 a1 -= W2 * col[8 * 6]; 00183 a2 += W2 * col[8 * 6]; 00184 a3 -= W6 * col[8 * 6]; 00185 } 00186 00187 if (col[8 * 1]) { 00188 b0 = W1 * col[8 * 1]; 00189 b1 = W3 * col[8 * 1]; 00190 b2 = W5 * col[8 * 1]; 00191 b3 = W7 * col[8 * 1]; 00192 } else { 00193 b0 = 0; 00194 b1 = 0; 00195 b2 = 0; 00196 b3 = 0; 00197 } 00198 00199 if (col[8 * 3]) { 00200 b0 += W3 * col[8 * 3]; 00201 b1 -= W7 * col[8 * 3]; 00202 b2 -= W1 * col[8 * 3]; 00203 b3 -= W5 * col[8 * 3]; 00204 } 00205 00206 if (col[8 * 5]) { 00207 b0 += W5 * col[8 * 5]; 00208 b1 -= W1 * col[8 * 5]; 00209 b2 += W7 * col[8 * 5]; 00210 b3 += W3 * col[8 * 5]; 00211 } 00212 00213 if (col[8 * 7]) { 00214 b0 += W7 * col[8 * 7]; 00215 b1 -= W5 * col[8 * 7]; 00216 b2 += W3 * col[8 * 7]; 00217 b3 -= W1 * col[8 * 7]; 00218 } 00219 00220 col[8 * 0] = (a0 + b0) >> COL_SHIFT; 00221 col[8 * 7] = (a0 - b0) >> COL_SHIFT; 00222 col[8 * 1] = (a1 + b1) >> COL_SHIFT; 00223 col[8 * 6] = (a1 - b1) >> COL_SHIFT; 00224 col[8 * 2] = (a2 + b2) >> COL_SHIFT; 00225 col[8 * 5] = (a2 - b2) >> COL_SHIFT; 00226 col[8 * 3] = (a3 + b3) >> COL_SHIFT; 00227 col[8 * 4] = (a3 - b3) >> COL_SHIFT; 00228 } 00229 00230 /* If all rows but the first one are zero after row transformation, 00231 all rows will be identical after column transformation. */ 00232 static inline void idct_col2(DCTELEM *col) 00233 { 00234 int i; 00235 uint64_t l, r; 00236 00237 for (i = 0; i < 8; ++i) { 00238 int_fast32_t a0 = col[i] + (1 << (COL_SHIFT - 1)) / W4; 00239 00240 a0 *= W4; 00241 col[i] = a0 >> COL_SHIFT; 00242 } 00243 00244 l = ldq(col + 0 * 4); r = ldq(col + 1 * 4); 00245 stq(l, col + 2 * 4); stq(r, col + 3 * 4); 00246 stq(l, col + 4 * 4); stq(r, col + 5 * 4); 00247 stq(l, col + 6 * 4); stq(r, col + 7 * 4); 00248 stq(l, col + 8 * 4); stq(r, col + 9 * 4); 00249 stq(l, col + 10 * 4); stq(r, col + 11 * 4); 00250 stq(l, col + 12 * 4); stq(r, col + 13 * 4); 00251 stq(l, col + 14 * 4); stq(r, col + 15 * 4); 00252 } 00253 00254 void ff_simple_idct_axp(DCTELEM *block) 00255 { 00256 00257 int i; 00258 int rowsZero = 1; /* all rows except row 0 zero */ 00259 int rowsConstant = 1; /* all rows consist of a constant value */ 00260 00261 for (i = 0; i < 8; i++) { 00262 int sparseness = idct_row(block + 8 * i); 00263 00264 if (i > 0 && sparseness > 0) 00265 rowsZero = 0; 00266 if (sparseness == 2) 00267 rowsConstant = 0; 00268 } 00269 00270 if (rowsZero) { 00271 idct_col2(block); 00272 } else if (rowsConstant) { 00273 idct_col(block); 00274 for (i = 0; i < 8; i += 2) { 00275 uint64_t v = (uint16_t) block[0]; 00276 uint64_t w = (uint16_t) block[8]; 00277 00278 v |= v << 16; 00279 w |= w << 16; 00280 v |= v << 32; 00281 w |= w << 32; 00282 stq(v, block + 0 * 4); 00283 stq(v, block + 1 * 4); 00284 stq(w, block + 2 * 4); 00285 stq(w, block + 3 * 4); 00286 block += 4 * 4; 00287 } 00288 } else { 00289 for (i = 0; i < 8; i++) 00290 idct_col(block + i); 00291 } 00292 } 00293 00294 void ff_simple_idct_put_axp(uint8_t *dest, int line_size, DCTELEM *block) 00295 { 00296 ff_simple_idct_axp(block); 00297 put_pixels_clamped_axp_p(block, dest, line_size); 00298 } 00299 00300 void ff_simple_idct_add_axp(uint8_t *dest, int line_size, DCTELEM *block) 00301 { 00302 ff_simple_idct_axp(block); 00303 add_pixels_clamped_axp_p(block, dest, line_size); 00304 }