Parent Directory
| 
Revision Log
Revision 1104 - (view) (download)
| 1 : | chl | 460 | ;/***************************************************************************** |
| 2 : | ; * | ||
| 3 : | ; * XVID MPEG-4 VIDEO CODEC | ||
| 4 : | ; * mmx version - inverse discrete cosine transformation | ||
| 5 : | ; * | ||
| 6 : | ; * Initial version provided by Intel at AppNote AP-922 | ||
| 7 : | ; * Copyright (C) 1999 Intel Corporation, | ||
| 8 : | ; * | ||
| 9 : | ; * Modifications | ||
| 10 : | ; * Copyright (c) 2000-2001 Peter Gubanov <peter@elecard.net.ru> | ||
| 11 : | ; * Copyright (c) 2000 Michel Lespinasse <walken@zoy.org> | ||
| 12 : | ; * | ||
| 13 : | ; * ported to NASM and some minor changes | ||
| 14 : | ; * Copyright (C) 2001 Peter Ross <pross@xvid.org> | ||
| 15 : | ; * | ||
| 16 : | edgomez | 649 | ; * This file is part of XviD, a free MPEG-4 video encoder/decoder |
| 17 : | chl | 460 | ; * |
| 18 : | edgomez | 649 | ; * XviD is free software; you can redistribute it and/or modify it |
| 19 : | ; * under the terms of the GNU General Public License as published by | ||
| 20 : | chl | 460 | ; * the Free Software Foundation; either version 2 of the License, or |
| 21 : | ; * (at your option) any later version. | ||
| 22 : | ; * | ||
| 23 : | ; * This program is distributed in the hope that it will be useful, | ||
| 24 : | ; * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 25 : | ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 26 : | ; * GNU General Public License for more details. | ||
| 27 : | ; * | ||
| 28 : | ; * You should have received a copy of the GNU General Public License | ||
| 29 : | ; * along with this program; if not, write to the Free Software | ||
| 30 : | ; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
| 31 : | ; * | ||
| 32 : | edgomez | 649 | ; * Under section 8 of the GNU General Public License, the copyright |
| 33 : | ; * holders of XVID explicitly forbid distribution in the following | ||
| 34 : | ; * countries: | ||
| 35 : | ; * | ||
| 36 : | ; * - Japan | ||
| 37 : | ; * - United States of America | ||
| 38 : | ; * | ||
| 39 : | ; * Linking XviD statically or dynamically with other modules is making a | ||
| 40 : | ; * combined work based on XviD. Thus, the terms and conditions of the | ||
| 41 : | ; * GNU General Public License cover the whole combination. | ||
| 42 : | ; * | ||
| 43 : | ; * As a special exception, the copyright holders of XviD give you | ||
| 44 : | ; * permission to link XviD with independent modules that communicate with | ||
| 45 : | ; * XviD solely through the VFW1.1 and DShow interfaces, regardless of the | ||
| 46 : | ; * license terms of these independent modules, and to copy and distribute | ||
| 47 : | ; * the resulting combined work under terms of your choice, provided that | ||
| 48 : | ; * every copy of the combined work is accompanied by a complete copy of | ||
| 49 : | ; * the source code of XviD (the version of XviD used to produce the | ||
| 50 : | ; * combined work), being distributed under the terms of the GNU General | ||
| 51 : | ; * Public License plus this exception. An independent module is a module | ||
| 52 : | ; * which is not derived from or based on XviD. | ||
| 53 : | ; * | ||
| 54 : | ; * Note that people who make modified versions of XviD are not obligated | ||
| 55 : | ; * to grant this special exception for their modified versions; it is | ||
| 56 : | ; * their choice whether to do so. The GNU General Public License gives | ||
| 57 : | ; * permission to release a modified version without this exception; this | ||
| 58 : | ; * exception also makes it possible to release a modified version which | ||
| 59 : | ; * carries forward this exception. | ||
| 60 : | ; * | ||
| 61 : | ; * $Id: idct_mmx.asm,v 1.5 2002-11-16 23:51:58 edgomez Exp $ | ||
| 62 : | ; * | ||
| 63 : | chl | 460 | ; *************************************************************************/ |
| 64 : | |||
| 65 : | Isibaar | 3 | ;============================================================================= |
| 66 : | ; | ||
| 67 : | ; These examples contain code fragments for first stage iDCT 8x8 | ||
| 68 : | ; (for rows) and first stage DCT 8x8 (for columns) | ||
| 69 : | ; | ||
| 70 : | ;============================================================================= | ||
| 71 : | |||
| 72 : | bits 32 | ||
| 73 : | |||
| 74 : | %macro cglobal 1 | ||
| 75 : | %ifdef PREFIX | ||
| 76 : | global _%1 | ||
| 77 : | %define %1 _%1 | ||
| 78 : | %else | ||
| 79 : | global %1 | ||
| 80 : | %endif | ||
| 81 : | %endmacro | ||
| 82 : | |||
| 83 : | %define BITS_INV_ACC 5 ; 4 or 5 for IEEE | ||
| 84 : | %define SHIFT_INV_ROW 16 - BITS_INV_ACC | ||
| 85 : | %define SHIFT_INV_COL 1 + BITS_INV_ACC | ||
| 86 : | %define RND_INV_ROW 1024 * (6 - BITS_INV_ACC) ; 1 << (SHIFT_INV_ROW-1) | ||
| 87 : | %define RND_INV_COL 16 * (BITS_INV_ACC - 3) ; 1 << (SHIFT_INV_COL-1) | ||
| 88 : | %define RND_INV_CORR RND_INV_COL - 1 ; correction -1.0 and round | ||
| 89 : | |||
| 90 : | %define BITS_FRW_ACC 3 ; 2 or 3 for accuracy | ||
| 91 : | %define SHIFT_FRW_COL BITS_FRW_ACC | ||
| 92 : | %define SHIFT_FRW_ROW BITS_FRW_ACC + 17 | ||
| 93 : | %define RND_FRW_ROW 262144 * (BITS_FRW_ACC - 1) ; 1 << (SHIFT_FRW_ROW-1) | ||
| 94 : | |||
| 95 : | suxen_drol | 207 | %ifdef FORMAT_COFF |
| 96 : | section .data data | ||
| 97 : | %else | ||
| 98 : | Isibaar | 154 | section .data data align=16 |
| 99 : | suxen_drol | 207 | %endif |
| 100 : | Isibaar | 3 | |
| 101 : | align 16 | ||
| 102 : | |||
| 103 : | one_corr dw 1, 1, 1, 1 | ||
| 104 : | round_inv_row dd RND_INV_ROW, RND_INV_ROW | ||
| 105 : | round_inv_col dw RND_INV_COL, RND_INV_COL, RND_INV_COL, RND_INV_COL | ||
| 106 : | round_inv_corr dw RND_INV_CORR, RND_INV_CORR, RND_INV_CORR, RND_INV_CORR | ||
| 107 : | round_frw_row dd RND_FRW_ROW, RND_FRW_ROW | ||
| 108 : | tg_1_16 dw 13036, 13036, 13036, 13036 ; tg * (2<<16) + 0.5 | ||
| 109 : | tg_2_16 dw 27146, 27146, 27146, 27146 ; tg * (2<<16) + 0.5 | ||
| 110 : | tg_3_16 dw -21746, -21746, -21746, -21746 ; tg * (2<<16) + 0.5 | ||
| 111 : | cos_4_16 dw -19195, -19195, -19195, -19195 ; cos * (2<<16) + 0.5 | ||
| 112 : | ocos_4_16 dw 23170, 23170, 23170, 23170 ; cos * (2<<15) + 0.5 | ||
| 113 : | |||
| 114 : | otg_3_16 dw 21895, 21895, 21895, 21895 ; tg * (2<<16) + 0.5 | ||
| 115 : | |||
| 116 : | %if SHIFT_INV_ROW == 12 ; assume SHIFT_INV_ROW == 12 | ||
| 117 : | rounder_0 dd 65536, 65536 | ||
| 118 : | rounder_4 dd 0, 0 | ||
| 119 : | rounder_1 dd 7195, 7195 | ||
| 120 : | rounder_7 dd 1024, 1024 | ||
| 121 : | rounder_2 dd 4520, 4520 | ||
| 122 : | rounder_6 dd 1024, 1024 | ||
| 123 : | rounder_3 dd 2407, 2407 | ||
| 124 : | rounder_5 dd 240, 240 | ||
| 125 : | |||
| 126 : | %elif SHIFT_INV_ROW == 11 ; assume SHIFT_INV_ROW == 11 | ||
| 127 : | rounder_0 dd 65536, 65536 | ||
| 128 : | rounder_4 dd 0, 0 | ||
| 129 : | rounder_1 dd 3597, 3597 | ||
| 130 : | rounder_7 dd 512, 512 | ||
| 131 : | rounder_2 dd 2260, 2260 | ||
| 132 : | rounder_6 dd 512, 512 | ||
| 133 : | rounder_3 dd 1203, 1203 | ||
| 134 : | rounder_5 dd 120, 120 | ||
| 135 : | %else | ||
| 136 : | |||
| 137 : | %error invalid _SHIFT_INV_ROW_ | ||
| 138 : | |||
| 139 : | %endif | ||
| 140 : | |||
| 141 : | ;============================================================================= | ||
| 142 : | ; | ||
| 143 : | ; The first stage iDCT 8x8 - inverse DCTs of rows | ||
| 144 : | ; | ||
| 145 : | ;----------------------------------------------------------------------------- | ||
| 146 : | ; The 8-point inverse DCT direct algorithm | ||
| 147 : | ;----------------------------------------------------------------------------- | ||
| 148 : | ; | ||
| 149 : | ; static const short w[32] = { | ||
| 150 : | ; FIX(cos_4_16), FIX(cos_2_16), FIX(cos_4_16), FIX(cos_6_16), | ||
| 151 : | ; FIX(cos_4_16), FIX(cos_6_16), -FIX(cos_4_16), -FIX(cos_2_16), | ||
| 152 : | ; FIX(cos_4_16), -FIX(cos_6_16), -FIX(cos_4_16), FIX(cos_2_16), | ||
| 153 : | ; FIX(cos_4_16), -FIX(cos_2_16), FIX(cos_4_16), -FIX(cos_6_16), | ||
| 154 : | ; FIX(cos_1_16), FIX(cos_3_16), FIX(cos_5_16), FIX(cos_7_16), | ||
| 155 : | ; FIX(cos_3_16), -FIX(cos_7_16), -FIX(cos_1_16), -FIX(cos_5_16), | ||
| 156 : | ; FIX(cos_5_16), -FIX(cos_1_16), FIX(cos_7_16), FIX(cos_3_16), | ||
| 157 : | ; FIX(cos_7_16), -FIX(cos_5_16), FIX(cos_3_16), -FIX(cos_1_16) }; | ||
| 158 : | ; | ||
| 159 : | ; #define DCT_8_INV_ROW(x, y) | ||
| 160 : | ; { | ||
| 161 : | ; int a0, a1, a2, a3, b0, b1, b2, b3; | ||
| 162 : | ; | ||
| 163 : | ; a0 =x[0]*w[0]+x[2]*w[1]+x[4]*w[2]+x[6]*w[3]; | ||
| 164 : | ; a1 =x[0]*w[4]+x[2]*w[5]+x[4]*w[6]+x[6]*w[7]; | ||
| 165 : | ; a2 = x[0] * w[ 8] + x[2] * w[ 9] + x[4] * w[10] + x[6] * w[11]; | ||
| 166 : | ; a3 = x[0] * w[12] + x[2] * w[13] + x[4] * w[14] + x[6] * w[15]; | ||
| 167 : | ; b0 = x[1] * w[16] + x[3] * w[17] + x[5] * w[18] + x[7] * w[19]; | ||
| 168 : | ; b1 = x[1] * w[20] + x[3] * w[21] + x[5] * w[22] + x[7] * w[23]; | ||
| 169 : | ; b2 = x[1] * w[24] + x[3] * w[25] + x[5] * w[26] + x[7] * w[27]; | ||
| 170 : | ; b3 = x[1] * w[28] + x[3] * w[29] + x[5] * w[30] + x[7] * w[31]; | ||
| 171 : | ; | ||
| 172 : | ; y[0] = SHIFT_ROUND ( a0 + b0 ); | ||
| 173 : | ; y[1] = SHIFT_ROUND ( a1 + b1 ); | ||
| 174 : | ; y[2] = SHIFT_ROUND ( a2 + b2 ); | ||
| 175 : | ; y[3] = SHIFT_ROUND ( a3 + b3 ); | ||
| 176 : | ; y[4] = SHIFT_ROUND ( a3 - b3 ); | ||
| 177 : | ; y[5] = SHIFT_ROUND ( a2 - b2 ); | ||
| 178 : | ; y[6] = SHIFT_ROUND ( a1 - b1 ); | ||
| 179 : | ; y[7] = SHIFT_ROUND ( a0 - b0 ); | ||
| 180 : | ; } | ||
| 181 : | ; | ||
| 182 : | ;----------------------------------------------------------------------------- | ||
| 183 : | ; | ||
| 184 : | ; In this implementation the outputs of the iDCT-1D are multiplied | ||
| 185 : | ; for rows 0,4 - by cos_4_16, | ||
| 186 : | ; for rows 1,7 - by cos_1_16, | ||
| 187 : | ; for rows 2,6 - by cos_2_16, | ||
| 188 : | ; for rows 3,5 - by cos_3_16 | ||
| 189 : | ; and are shifted to the left for better accuracy | ||
| 190 : | ; | ||
| 191 : | ; For the constants used, | ||
| 192 : | ; FIX(float_const) = (short) (float_const * (1<<15) + 0.5) | ||
| 193 : | ; | ||
| 194 : | ;============================================================================= | ||
| 195 : | |||
| 196 : | ;============================================================================= | ||
| 197 : | ; MMX code | ||
| 198 : | ;============================================================================= | ||
| 199 : | |||
| 200 : | ; Table for rows 0,4 - constants are multiplied by cos_4_16 | ||
| 201 : | |||
| 202 : | tab_i_04 dw 16384, 16384, 16384, -16384 ; movq-> w06 w04 w02 w00 | ||
| 203 : | dw 21407, 8867, 8867, -21407 ; w07 w05 w03 w01 | ||
| 204 : | dw 16384, -16384, 16384, 16384 ; w14 w12 w10 w08 | ||
| 205 : | dw -8867, 21407, -21407, -8867 ; w15 w13 w11 w09 | ||
| 206 : | dw 22725, 12873, 19266, -22725 ; w22 w20 w18 w16 | ||
| 207 : | dw 19266, 4520, -4520, -12873 ; w23 w21 w19 w17 | ||
| 208 : | dw 12873, 4520, 4520, 19266 ; w30 w28 w26 w24 | ||
| 209 : | dw -22725, 19266, -12873, -22725 ; w31 w29 w27 w25 | ||
| 210 : | |||
| 211 : | ; Table for rows 1,7 - constants are multiplied by cos_1_16 | ||
| 212 : | |||
| 213 : | tab_i_17 dw 22725, 22725, 22725, -22725 ; movq-> w06 w04 w02 w00 | ||
| 214 : | dw 29692, 12299, 12299, -29692 ; w07 w05 w03 w01 | ||
| 215 : | dw 22725, -22725, 22725, 22725 ; w14 w12 w10 w08 | ||
| 216 : | dw -12299, 29692, -29692, -12299 ; w15 w13 w11 w09 | ||
| 217 : | dw 31521, 17855, 26722, -31521 ; w22 w20 w18 w16 | ||
| 218 : | dw 26722, 6270, -6270, -17855 ; w23 w21 w19 w17 | ||
| 219 : | dw 17855, 6270, 6270, 26722 ; w30 w28 w26 w24 | ||
| 220 : | dw -31521, 26722, -17855, -31521 ; w31 w29 w27 w25 | ||
| 221 : | |||
| 222 : | ; Table for rows 2,6 - constants are multiplied by cos_2_16 | ||
| 223 : | |||
| 224 : | tab_i_26 dw 21407, 21407, 21407, -21407 ; movq-> w06 w04 w02 w00 | ||
| 225 : | dw 27969, 11585, 11585, -27969 ; w07 w05 w03 w01 | ||
| 226 : | dw 21407, -21407, 21407, 21407 ; w14 w12 w10 w08 | ||
| 227 : | dw -11585, 27969, -27969, -11585 ; w15 w13 w11 w09 | ||
| 228 : | dw 29692, 16819, 25172, -29692 ; w22 w20 w18 w16 | ||
| 229 : | dw 25172, 5906, -5906, -16819 ; w23 w21 w19 w17 | ||
| 230 : | dw 16819, 5906, 5906, 25172 ; w30 w28 w26 w24 | ||
| 231 : | dw -29692, 25172, -16819, -29692 ; w31 w29 w27 w25 | ||
| 232 : | |||
| 233 : | ; Table for rows 3,5 - constants are multiplied by cos_3_16 | ||
| 234 : | |||
| 235 : | tab_i_35 dw 19266, 19266, 19266, -19266 ; movq-> w06 w04 w02 w00 | ||
| 236 : | dw 25172, 10426, 10426, -25172 ; w07 w05 w03 w01 | ||
| 237 : | dw 19266, -19266, 19266, 19266 ; w14 w12 w10 w08 | ||
| 238 : | dw -10426, 25172, -25172, -10426 ; w15 w13 w11 w09 | ||
| 239 : | dw 26722, 15137, 22654, -26722 ; w22 w20 w18 w16 | ||
| 240 : | dw 22654, 5315, -5315, -15137 ; w23 w21 w19 w17 | ||
| 241 : | dw 15137, 5315, 5315, 22654 ; w30 w28 w26 w24 | ||
| 242 : | dw -26722, 22654, -15137, -26722 ; w31 w29 w27 w25 | ||
| 243 : | |||
| 244 : | ;----------------------------------------------------------------------------- | ||
| 245 : | |||
| 246 : | ; | ||
| 247 : | ; DCT_8_INV_ROW_1 INP, OUT, TABLE, ROUNDER | ||
| 248 : | ; | ||
| 249 : | |||
| 250 : | %macro DCT_8_INV_ROW_1 4 | ||
| 251 : | |||
| 252 : | movq mm0, [%1] ; 0 ; x3 x2 x1 x0 | ||
| 253 : | |||
| 254 : | movq mm1, [%1+8] ; 1 ; x7 x6 x5 x4 | ||
| 255 : | movq mm2, mm0 ; 2 ; x3 x2 x1 x0 | ||
| 256 : | |||
| 257 : | movq mm3, [%3] ; 3 ; w06 w04 w02 w00 | ||
| 258 : | punpcklwd mm0, mm1 ; x5 x1 x4 x0 | ||
| 259 : | |||
| 260 : | movq mm5, mm0 ; 5 ; x5 x1 x4 x0 | ||
| 261 : | punpckldq mm0, mm0 ; x4 x0 x4 x0 | ||
| 262 : | |||
| 263 : | movq mm4, [%3+8] ; 4 ; w07 w05 w03 w01 | ||
| 264 : | punpckhwd mm2, mm1 ; 1 ; x7 x3 x6 x2 | ||
| 265 : | |||
| 266 : | pmaddwd mm3, mm0 ; x4*w06+x0*w04 x4*w02+x0*w00 | ||
| 267 : | movq mm6, mm2 ; 6 ; x7 x3 x6 x2 | ||
| 268 : | |||
| 269 : | movq mm1, [%3+32] ; 1 ; w22 w20 w18 w16 | ||
| 270 : | punpckldq mm2, mm2 ; x6 x2 x6 x2 | ||
| 271 : | |||
| 272 : | pmaddwd mm4, mm2 ; x6*w07+x2*w05 x6*w03+x2*w01 | ||
| 273 : | punpckhdq mm5, mm5 ; x5 x1 x5 x1 | ||
| 274 : | |||
| 275 : | pmaddwd mm0, [%3+16] ; x4*w14+x0*w12 x4*w10+x0*w08 | ||
| 276 : | punpckhdq mm6, mm6 ; x7 x3 x7 x3 | ||
| 277 : | |||
| 278 : | movq mm7, [%3+40] ; 7 ; w23 w21 w19 w17 | ||
| 279 : | pmaddwd mm1, mm5 ; x5*w22+x1*w20 x5*w18+x1*w16 | ||
| 280 : | |||
| 281 : | paddd mm3, [%4] ; +%4 | ||
| 282 : | pmaddwd mm7, mm6 ; x7*w23+x3*w21 x7*w19+x3*w17 | ||
| 283 : | |||
| 284 : | pmaddwd mm2, [%3+24] ; x6*w15+x2*w13 x6*w11+x2*w09 | ||
| 285 : | paddd mm3, mm4 ; 4 ; a1=sum(even1) a0=sum(even0) | ||
| 286 : | |||
| 287 : | pmaddwd mm5, [%3+48] ; x5*w30+x1*w28 x5*w26+x1*w24 | ||
| 288 : | movq mm4, mm3 ; 4 ; a1 a0 | ||
| 289 : | |||
| 290 : | pmaddwd mm6, [%3+56] ; x7*w31+x3*w29 x7*w27+x3*w25 | ||
| 291 : | paddd mm1, mm7 ; 7 ; b1=sum(odd1) b0=sum(odd0) | ||
| 292 : | |||
| 293 : | paddd mm0, [%4] ; +%4 | ||
| 294 : | psubd mm3, mm1 ; a1-b1 a0-b0 | ||
| 295 : | |||
| 296 : | psrad mm3, SHIFT_INV_ROW ; y6=a1-b1 y7=a0-b0 | ||
| 297 : | paddd mm1, mm4 ; 4 ; a1+b1 a0+b0 | ||
| 298 : | |||
| 299 : | paddd mm0, mm2 ; 2 ; a3=sum(even3) a2=sum(even2) | ||
| 300 : | psrad mm1, SHIFT_INV_ROW ; y1=a1+b1 y0=a0+b0 | ||
| 301 : | |||
| 302 : | paddd mm5, mm6 ; 6 ; b3=sum(odd3) b2=sum(odd2) | ||
| 303 : | movq mm4, mm0 ; 4 ; a3 a2 | ||
| 304 : | |||
| 305 : | paddd mm0, mm5 ; a3+b3 a2+b2 | ||
| 306 : | psubd mm4, mm5 ; 5 ; a3-b3 a2-b2 | ||
| 307 : | |||
| 308 : | psrad mm0, SHIFT_INV_ROW ; y3=a3+b3 y2=a2+b2 | ||
| 309 : | psrad mm4, SHIFT_INV_ROW ; y4=a3-b3 y5=a2-b2 | ||
| 310 : | |||
| 311 : | packssdw mm1, mm0 ; 0 ; y3 y2 y1 y0 | ||
| 312 : | packssdw mm4, mm3 ; 3 ; y6 y7 y4 y5 | ||
| 313 : | |||
| 314 : | movq mm7, mm4 ; 7 ; y6 y7 y4 y5 | ||
| 315 : | psrld mm4, 16 ; 0 y6 0 y4 | ||
| 316 : | |||
| 317 : | pslld mm7, 16 ; y7 0 y5 0 | ||
| 318 : | movq [%2], mm1 ; 1 ; save y3 y2 y1 y0 | ||
| 319 : | |||
| 320 : | por mm7, mm4 ; 4 ; y7 y6 y5 y4 | ||
| 321 : | movq [%2+8], mm7 ; 7 ; save y7 y6 y5 y4 | ||
| 322 : | %endmacro | ||
| 323 : | |||
| 324 : | |||
| 325 : | |||
| 326 : | |||
| 327 : | ;============================================================================= | ||
| 328 : | ; code for Pentium III | ||
| 329 : | ;============================================================================= | ||
| 330 : | |||
| 331 : | ; %3 for rows 0,4 - constants are multiplied by cos_4_16 | ||
| 332 : | |||
| 333 : | tab_i_04_sse dw 16384, 21407, 16384, 8867 ; movq-> w05 w04 w01 w00 | ||
| 334 : | dw 16384, 8867, -16384, -21407 ; w07 w06 w03 w02 | ||
| 335 : | dw 16384, -8867, 16384, -21407 ; w13 w12 w09 w08 | ||
| 336 : | dw -16384, 21407, 16384, -8867 ; w15 w14 w11 w10 | ||
| 337 : | dw 22725, 19266, 19266, -4520 ; w21 w20 w17 w16 | ||
| 338 : | dw 12873, 4520, -22725, -12873 ; w23 w22 w19 w18 | ||
| 339 : | dw 12873, -22725, 4520, -12873 ; w29 w28 w25 w24 | ||
| 340 : | dw 4520, 19266, 19266, -22725 ; w31 w30 w27 w26 | ||
| 341 : | |||
| 342 : | ; %3 for rows 1,7 - constants are multiplied by cos_1_16 | ||
| 343 : | |||
| 344 : | tab_i_17_sse dw 22725, 29692, 22725, 12299 ; movq-> w05 w04 w01 w00 | ||
| 345 : | dw 22725, 12299, -22725, -29692 ; w07 w06 w03 w02 | ||
| 346 : | dw 22725, -12299, 22725, -29692 ; w13 w12 w09 w08 | ||
| 347 : | dw -22725, 29692, 22725, -12299 ; w15 w14 w11 w10 | ||
| 348 : | dw 31521, 26722, 26722, -6270 ; w21 w20 w17 w16 | ||
| 349 : | dw 17855, 6270, -31521, -17855 ; w23 w22 w19 w18 | ||
| 350 : | dw 17855, -31521, 6270, -17855 ; w29 w28 w25 w24 | ||
| 351 : | dw 6270, 26722, 26722, -31521 ; w31 w30 w27 w26 | ||
| 352 : | |||
| 353 : | ; %3 for rows 2,6 - constants are multiplied by cos_2_16 | ||
| 354 : | |||
| 355 : | tab_i_26_sse dw 21407, 27969, 21407, 11585 ; movq-> w05 w04 w01 w00 | ||
| 356 : | dw 21407, 11585, -21407, -27969 ; w07 w06 w03 w02 | ||
| 357 : | dw 21407, -11585, 21407, -27969 ; w13 w12 w09 w08 | ||
| 358 : | dw -21407, 27969, 21407, -11585 ; w15 w14 w11 w10 | ||
| 359 : | dw 29692, 25172, 25172, -5906 ; w21 w20 w17 w16 | ||
| 360 : | dw 16819, 5906, -29692, -16819 ; w23 w22 w19 w18 | ||
| 361 : | dw 16819, -29692, 5906, -16819 ; w29 w28 w25 w24 | ||
| 362 : | dw 5906, 25172, 25172, -29692 ; w31 w30 w27 w26 | ||
| 363 : | |||
| 364 : | ; %3 for rows 3,5 - constants are multiplied by cos_3_16 | ||
| 365 : | |||
| 366 : | tab_i_35_sse dw 19266, 25172, 19266, 10426 ; movq-> w05 w04 w01 w00 | ||
| 367 : | dw 19266, 10426, -19266, -25172 ; w07 w06 w03 w02 | ||
| 368 : | dw 19266, -10426, 19266, -25172 ; w13 w12 w09 w08 | ||
| 369 : | dw -19266, 25172, 19266, -10426 ; w15 w14 w11 w10 | ||
| 370 : | dw 26722, 22654, 22654, -5315 ; w21 w20 w17 w16 | ||
| 371 : | dw 15137, 5315, -26722, -15137 ; w23 w22 w19 w18 | ||
| 372 : | dw 15137, -26722, 5315, -15137 ; w29 w28 w25 w24 | ||
| 373 : | dw 5315, 22654, 22654, -26722 ; w31 w30 w27 w26 | ||
| 374 : | |||
| 375 : | ;----------------------------------------------------------------------------- | ||
| 376 : | |||
| 377 : | ; | ||
| 378 : | ; DCT_8_INV_ROW_1_sse INP, OUT, TABLE, ROUNDER | ||
| 379 : | ; | ||
| 380 : | |||
| 381 : | %macro DCT_8_INV_ROW_1_sse 4 | ||
| 382 : | |||
| 383 : | movq mm0, [%1] ; 0 ; x3 x2 x1 x0 | ||
| 384 : | |||
| 385 : | movq mm1, [%1+8] ; 1 ; x7 x6 x5 x4 | ||
| 386 : | movq mm2, mm0 ; 2 ; x3 x2 x1 x0 | ||
| 387 : | |||
| 388 : | movq mm3, [%3] ; 3 ; w05 w04 w01 w00 | ||
| 389 : | pshufw mm0, mm0, 10001000b ; x2 x0 x2 x0 | ||
| 390 : | |||
| 391 : | movq mm4, [%3+8] ; 4 ; w07 w06 w03 w02 | ||
| 392 : | movq mm5, mm1 ; 5 ; x7 x6 x5 x4 | ||
| 393 : | pmaddwd mm3, mm0 ; x2*w05+x0*w04 x2*w01+x0*w00 | ||
| 394 : | |||
| 395 : | movq mm6, [%3+32] ; 6 ; w21 w20 w17 w16 | ||
| 396 : | pshufw mm1, mm1, 10001000b ; x6 x4 x6 x4 | ||
| 397 : | pmaddwd mm4, mm1 ; x6*w07+x4*w06 x6*w03+x4*w02 | ||
| 398 : | |||
| 399 : | movq mm7, [%3+40] ; 7 ; w23 w22 w19 w18 | ||
| 400 : | pshufw mm2, mm2, 11011101b ; x3 x1 x3 x1 | ||
| 401 : | pmaddwd mm6, mm2 ; x3*w21+x1*w20 x3*w17+x1*w16 | ||
| 402 : | |||
| 403 : | pshufw mm5, mm5, 11011101b ; x7 x5 x7 x5 | ||
| 404 : | pmaddwd mm7, mm5 ; x7*w23+x5*w22 x7*w19+x5*w18 | ||
| 405 : | |||
| 406 : | paddd mm3, [%4] ; +%4 | ||
| 407 : | |||
| 408 : | pmaddwd mm0, [%3+16] ; x2*w13+x0*w12 x2*w09+x0*w08 | ||
| 409 : | paddd mm3, mm4 ; 4 ; a1=sum(even1) a0=sum(even0) | ||
| 410 : | |||
| 411 : | pmaddwd mm1, [%3+24] ; x6*w15+x4*w14 x6*w11+x4*w10 | ||
| 412 : | movq mm4, mm3 ; 4 ; a1 a0 | ||
| 413 : | |||
| 414 : | pmaddwd mm2, [%3+48] ; x3*w29+x1*w28 x3*w25+x1*w24 | ||
| 415 : | paddd mm6, mm7 ; 7 ; b1=sum(odd1) b0=sum(odd0) | ||
| 416 : | |||
| 417 : | pmaddwd mm5, [%3+56] ; x7*w31+x5*w30 x7*w27+x5*w26 | ||
| 418 : | paddd mm3, mm6 ; a1+b1 a0+b0 | ||
| 419 : | |||
| 420 : | paddd mm0, [%4] ; +%4 | ||
| 421 : | psrad mm3, SHIFT_INV_ROW ; y1=a1+b1 y0=a0+b0 | ||
| 422 : | |||
| 423 : | paddd mm0, mm1 ; 1 ; a3=sum(even3) a2=sum(even2) | ||
| 424 : | psubd mm4, mm6 ; 6 ; a1-b1 a0-b0 | ||
| 425 : | |||
| 426 : | movq mm7, mm0 ; 7 ; a3 a2 | ||
| 427 : | paddd mm2, mm5 ; 5 ; b3=sum(odd3) b2=sum(odd2) | ||
| 428 : | |||
| 429 : | paddd mm0, mm2 ; a3+b3 a2+b2 | ||
| 430 : | psrad mm4, SHIFT_INV_ROW ; y6=a1-b1 y7=a0-b0 | ||
| 431 : | |||
| 432 : | psubd mm7, mm2 ; 2 ; a3-b3 a2-b2 | ||
| 433 : | psrad mm0, SHIFT_INV_ROW ; y3=a3+b3 y2=a2+b2 | ||
| 434 : | |||
| 435 : | psrad mm7, SHIFT_INV_ROW ; y4=a3-b3 y5=a2-b2 | ||
| 436 : | |||
| 437 : | packssdw mm3, mm0 ; 0 ; y3 y2 y1 y0 | ||
| 438 : | |||
| 439 : | packssdw mm7, mm4 ; 4 ; y6 y7 y4 y5 | ||
| 440 : | |||
| 441 : | movq [%2], mm3 ; 3 ; save y3 y2 y1 y0 | ||
| 442 : | pshufw mm7, mm7, 10110001b ; y7 y6 y5 y4 | ||
| 443 : | |||
| 444 : | movq [%2+8], mm7 ; 7 ; save y7 y6 y5 y4 | ||
| 445 : | |||
| 446 : | %endmacro | ||
| 447 : | |||
| 448 : | |||
| 449 : | ;============================================================================= | ||
| 450 : | ; | ||
| 451 : | ;============================================================================= | ||
| 452 : | |||
| 453 : | ;============================================================================= | ||
| 454 : | ; | ||
| 455 : | ; The first stage DCT 8x8 - forward DCTs of columns | ||
| 456 : | ; | ||
| 457 : | ; The %2puts are multiplied | ||
| 458 : | ; for rows 0,4 - on cos_4_16, | ||
| 459 : | ; for rows 1,7 - on cos_1_16, | ||
| 460 : | ; for rows 2,6 - on cos_2_16, | ||
| 461 : | ; for rows 3,5 - on cos_3_16 | ||
| 462 : | ; and are shifted to the left for rise of accuracy | ||
| 463 : | ; | ||
| 464 : | ;----------------------------------------------------------------------------- | ||
| 465 : | ; | ||
| 466 : | ; The 8-point scaled forward DCT algorithm (26a8m) | ||
| 467 : | ; | ||
| 468 : | ;----------------------------------------------------------------------------- | ||
| 469 : | ; | ||
| 470 : | ; #define DCT_8_FRW_COL(x, y) | ||
| 471 : | ;{ | ||
| 472 : | ; short t0, t1, t2, t3, t4, t5, t6, t7; | ||
| 473 : | ; short tp03, tm03, tp12, tm12, tp65, tm65; | ||
| 474 : | ; short tp465, tm465, tp765, tm765; | ||
| 475 : | ; | ||
| 476 : | ; t0 = LEFT_SHIFT ( x[0] + x[7] ); | ||
| 477 : | ; t1 = LEFT_SHIFT ( x[1] + x[6] ); | ||
| 478 : | ; t2 = LEFT_SHIFT ( x[2] + x[5] ); | ||
| 479 : | ; t3 = LEFT_SHIFT ( x[3] + x[4] ); | ||
| 480 : | ; t4 = LEFT_SHIFT ( x[3] - x[4] ); | ||
| 481 : | ; t5 = LEFT_SHIFT ( x[2] - x[5] ); | ||
| 482 : | ; t6 = LEFT_SHIFT ( x[1] - x[6] ); | ||
| 483 : | ; t7 = LEFT_SHIFT ( x[0] - x[7] ); | ||
| 484 : | ; | ||
| 485 : | ; tp03 = t0 + t3; | ||
| 486 : | ; tm03 = t0 - t3; | ||
| 487 : | ; tp12 = t1 + t2; | ||
| 488 : | ; tm12 = t1 - t2; | ||
| 489 : | ; | ||
| 490 : | ; y[0] = tp03 + tp12; | ||
| 491 : | ; y[4] = tp03 - tp12; | ||
| 492 : | ; | ||
| 493 : | ; y[2] = tm03 + tm12 * tg_2_16; | ||
| 494 : | ; y[6] = tm03 * tg_2_16 - tm12; | ||
| 495 : | ; | ||
| 496 : | ; tp65 =(t6 +t5 )*cos_4_16; | ||
| 497 : | ; tm65 =(t6 -t5 )*cos_4_16; | ||
| 498 : | ; | ||
| 499 : | ; tp765 = t7 + tp65; | ||
| 500 : | ; tm765 = t7 - tp65; | ||
| 501 : | ; tp465 = t4 + tm65; | ||
| 502 : | ; tm465 = t4 - tm65; | ||
| 503 : | ; | ||
| 504 : | ; y[1] = tp765 + tp465 * tg_1_16; | ||
| 505 : | ; y[7] = tp765 * tg_1_16 - tp465; | ||
| 506 : | ; y[5] = tm765 * tg_3_16 + tm465; | ||
| 507 : | ; y[3] = tm765 - tm465 * tg_3_16; | ||
| 508 : | ;} | ||
| 509 : | ; | ||
| 510 : | ;============================================================================= | ||
| 511 : | |||
| 512 : | |||
| 513 : | ; | ||
| 514 : | ; DCT_8_FRW_COL_4 INP, OUT | ||
| 515 : | ; | ||
| 516 : | |||
| 517 : | %macro DCT_8_FRW_COL_4 2 | ||
| 518 : | |||
| 519 : | LOCAL x0, x1, x2, x3, x4, x5, x6, x7 | ||
| 520 : | LOCAL y0, y1, y2, y3, y4, y5, y6, y7 | ||
| 521 : | x0 equ [%1 + 0*16] | ||
| 522 : | x1 equ [%1 + 1*16] | ||
| 523 : | x2 equ [%1 + 2*16] | ||
| 524 : | x3 equ [%1 + 3*16] | ||
| 525 : | x4 equ [%1 + 4*16] | ||
| 526 : | x5 equ [%1 + 5*16] | ||
| 527 : | x6 equ [%1 + 6*16] | ||
| 528 : | x7 equ [%1 + 7*16] | ||
| 529 : | y0 equ [%2 + 0*16] | ||
| 530 : | y1 equ [%2 + 1*16] | ||
| 531 : | y2 equ [%2 + 2*16] | ||
| 532 : | y3 equ [%2 + 3*16] | ||
| 533 : | y4 equ [%2 + 4*16] | ||
| 534 : | y5 equ [%2 + 5*16] | ||
| 535 : | y6 equ [%2 + 6*16] | ||
| 536 : | y7 equ [%2 + 7*16] | ||
| 537 : | movq mm0, x1 ; 0 ; x1 | ||
| 538 : | movq mm1, x6 ; 1 ; x6 | ||
| 539 : | movq mm2, mm0 ; 2 ; x1 | ||
| 540 : | movq mm3, x2 ; 3 ; x2 | ||
| 541 : | paddsw mm0, mm1 ; t1 = x[1] + x[6] | ||
| 542 : | movq mm4, x5 ; 4 ; x5 | ||
| 543 : | psllw mm0, SHIFT_FRW_COL ; t1 | ||
| 544 : | movq mm5, x0 ; 5 ; x0 | ||
| 545 : | paddsw mm4, mm3 ; t2 = x[2] + x[5] | ||
| 546 : | paddsw mm5, x7 ; t0 = x[0] + x[7] | ||
| 547 : | psllw mm4, SHIFT_FRW_COL ; t2 | ||
| 548 : | movq mm6, mm0 ; 6 ; t1 | ||
| 549 : | psubsw mm2, mm1 ; 1 ; t6 = x[1] - x[6] | ||
| 550 : | movq mm1, [tg_2_16] ; 1 ; tg_2_16 | ||
| 551 : | psubsw mm0, mm4 ; tm12 = t1 - t2 | ||
| 552 : | movq mm7, x3 ; 7 ; x3 | ||
| 553 : | pmulhw mm1, mm0 ; tm12*tg_2_16 | ||
| 554 : | paddsw mm7, x4 ; t3 = x[3] + x[4] | ||
| 555 : | psllw mm5, SHIFT_FRW_COL ; t0 | ||
| 556 : | paddsw mm6, mm4 ; 4 ; tp12 = t1 + t2 | ||
| 557 : | psllw mm7, SHIFT_FRW_COL ; t3 | ||
| 558 : | movq mm4, mm5 ; 4 ; t0 | ||
| 559 : | psubsw mm5, mm7 ; tm03 = t0 - t3 | ||
| 560 : | paddsw mm1, mm5 ; y2 = tm03 + tm12*tg_2_16 | ||
| 561 : | paddsw mm4, mm7 ; 7 ; tp03 = t0 + t3 | ||
| 562 : | por mm1, [one_corr] ; correction y2 +0.5 | ||
| 563 : | psllw mm2, SHIFT_FRW_COL+1 ; t6 | ||
| 564 : | pmulhw mm5, [tg_2_16] ; tm03*tg_2_16 | ||
| 565 : | movq mm7, mm4 ; 7 ; tp03 | ||
| 566 : | psubsw mm3, x5 ; t5 = x[2] - x[5] | ||
| 567 : | psubsw mm4, mm6 ; y4 = tp03 - tp12 | ||
| 568 : | movq y2, mm1 ; 1 ; save y2 | ||
| 569 : | paddsw mm7, mm6 ; 6 ; y0 = tp03 + tp12 | ||
| 570 : | movq mm1, x3 ; 1 ; x3 | ||
| 571 : | psllw mm3, SHIFT_FRW_COL+1 ; t5 | ||
| 572 : | psubsw mm1, x4 ; t4 = x[3] - x[4] | ||
| 573 : | movq mm6, mm2 ; 6 ; t6 | ||
| 574 : | movq y4, mm4 ; 4 ; save y4 | ||
| 575 : | paddsw mm2, mm3 ; t6 + t5 | ||
| 576 : | pmulhw mm2, [ocos_4_16] ; tp65 = (t6 + t5)*cos_4_16 | ||
| 577 : | psubsw mm6, mm3 ; 3 ; t6 - t5 | ||
| 578 : | pmulhw mm6, [ocos_4_16] ; tm65 = (t6 - t5)*cos_4_16 | ||
| 579 : | psubsw mm5, mm0 ; 0 ; y6 = tm03*tg_2_16 - tm12 | ||
| 580 : | por mm5, [one_corr] ; correction y6 +0.5 | ||
| 581 : | psllw mm1, SHIFT_FRW_COL ; t4 | ||
| 582 : | por mm2, [one_corr] ; correction tp65 +0.5 | ||
| 583 : | movq mm4, mm1 ; 4 ; t4 | ||
| 584 : | movq mm3, x0 ; 3 ; x0 | ||
| 585 : | paddsw mm1, mm6 ; tp465 = t4 + tm65 | ||
| 586 : | psubsw mm3, x7 ; t7 = x[0] - x[7] | ||
| 587 : | psubsw mm4, mm6 ; 6 ; tm465 = t4 - tm65 | ||
| 588 : | movq mm0, [tg_1_16] ; 0 ; tg_1_16 | ||
| 589 : | psllw mm3, SHIFT_FRW_COL ; t7 | ||
| 590 : | movq mm6, [tg_3_16] ; 6 ; tg_3_16 | ||
| 591 : | pmulhw mm0, mm1 ; tp465*tg_1_16 | ||
| 592 : | movq y0, mm7 ; 7 ; save y0 | ||
| 593 : | pmulhw mm6, mm4 ; tm465*tg_3_16 | ||
| 594 : | movq y6, mm5 ; 5 ; save y6 | ||
| 595 : | movq mm7, mm3 ; 7 ; t7 | ||
| 596 : | movq mm5, [tg_3_16] ; 5 ; tg_3_16 | ||
| 597 : | psubsw mm7, mm2 ; tm765 = t7 - tp65 | ||
| 598 : | paddsw mm3, mm2 ; 2 ; tp765 = t7 + tp65 | ||
| 599 : | pmulhw mm5, mm7 ; tm765*tg_3_16 | ||
| 600 : | paddsw mm0, mm3 ; y1 = tp765 + tp465*tg_1_16 | ||
| 601 : | paddsw mm6, mm4 ; tm465*tg_3_16 | ||
| 602 : | pmulhw mm3, [tg_1_16] ; tp765*tg_1_16 | ||
| 603 : | por mm0, [one_corr] ; correction y1 +0.5 | ||
| 604 : | paddsw mm5, mm7 ; tm765*tg_3_16 | ||
| 605 : | psubsw mm7, mm6 ; 6 ; y3 = tm765 - tm465*tg_3_16 | ||
| 606 : | movq y1, mm0 ; 0 ; save y1 | ||
| 607 : | paddsw mm5, mm4 ; 4 ; y5 = tm765*tg_3_16 + tm465 | ||
| 608 : | movq y3, mm7 ; 7 ; save y3 | ||
| 609 : | psubsw mm3, mm1 ; 1 ; y7 = tp765*tg_1_16 - tp465 | ||
| 610 : | movq y5, mm5 ; 5 ; save y5 | ||
| 611 : | movq y7, mm3 ; 3 ; save y7 | ||
| 612 : | %endmacro | ||
| 613 : | |||
| 614 : | |||
| 615 : | ; | ||
| 616 : | ; DCT_8_INV_COL_4 INP,OUT | ||
| 617 : | ; | ||
| 618 : | |||
| 619 : | %macro DCT_8_INV_COL_4 2 | ||
| 620 : | movq mm0, [tg_3_16] | ||
| 621 : | |||
| 622 : | movq mm3, [%1+16*3] | ||
| 623 : | movq mm1, mm0 ; tg_3_16 | ||
| 624 : | |||
| 625 : | movq mm5, [%1+16*5] | ||
| 626 : | pmulhw mm0, mm3 ; x3*(tg_3_16-1) | ||
| 627 : | |||
| 628 : | movq mm4, [tg_1_16] | ||
| 629 : | pmulhw mm1, mm5 ; x5*(tg_3_16-1) | ||
| 630 : | |||
| 631 : | movq mm7, [%1+16*7] | ||
| 632 : | movq mm2, mm4 ; tg_1_16 | ||
| 633 : | |||
| 634 : | movq mm6, [%1+16*1] | ||
| 635 : | pmulhw mm4, mm7 ; x7*tg_1_16 | ||
| 636 : | |||
| 637 : | paddsw mm0, mm3 ; x3*tg_3_16 | ||
| 638 : | pmulhw mm2, mm6 ; x1*tg_1_16 | ||
| 639 : | |||
| 640 : | paddsw mm1, mm3 ; x3+x5*(tg_3_16-1) | ||
| 641 : | psubsw mm0, mm5 ; x3*tg_3_16-x5 = tm35 | ||
| 642 : | |||
| 643 : | movq mm3, [ocos_4_16] | ||
| 644 : | paddsw mm1, mm5 ; x3+x5*tg_3_16 = tp35 | ||
| 645 : | |||
| 646 : | paddsw mm4, mm6 ; x1+tg_1_16*x7 = tp17 | ||
| 647 : | psubsw mm2, mm7 ; x1*tg_1_16-x7 = tm17 | ||
| 648 : | |||
| 649 : | movq mm5, mm4 ; tp17 | ||
| 650 : | movq mm6, mm2 ; tm17 | ||
| 651 : | |||
| 652 : | paddsw mm5, mm1 ; tp17+tp35 = b0 | ||
| 653 : | psubsw mm6, mm0 ; tm17-tm35 = b3 | ||
| 654 : | |||
| 655 : | psubsw mm4, mm1 ; tp17-tp35 = t1 | ||
| 656 : | paddsw mm2, mm0 ; tm17+tm35 = t2 | ||
| 657 : | |||
| 658 : | movq mm7, [tg_2_16] | ||
| 659 : | movq mm1, mm4 ; t1 | ||
| 660 : | |||
| 661 : | ; movq [SCRATCH+0], mm5 ; save b0 | ||
| 662 : | movq [%2+3*16], mm5 ; save b0 | ||
| 663 : | paddsw mm1, mm2 ; t1+t2 | ||
| 664 : | |||
| 665 : | ; movq [SCRATCH+8], mm6 ; save b3 | ||
| 666 : | movq [%2+5*16], mm6 ; save b3 | ||
| 667 : | psubsw mm4, mm2 ; t1-t2 | ||
| 668 : | |||
| 669 : | movq mm5, [%1+2*16] | ||
| 670 : | movq mm0, mm7 ; tg_2_16 | ||
| 671 : | |||
| 672 : | movq mm6, [%1+6*16] | ||
| 673 : | pmulhw mm0, mm5 ; x2*tg_2_16 | ||
| 674 : | |||
| 675 : | pmulhw mm7, mm6 ; x6*tg_2_16 | ||
| 676 : | ; slot | ||
| 677 : | pmulhw mm1, mm3 ; ocos_4_16*(t1+t2) = b1/2 | ||
| 678 : | ; slot | ||
| 679 : | movq mm2, [%1+0*16] | ||
| 680 : | pmulhw mm4, mm3 ; ocos_4_16*(t1-t2) = b2/2 | ||
| 681 : | |||
| 682 : | psubsw mm0, mm6 ; t2*tg_2_16-x6 = tm26 | ||
| 683 : | movq mm3, mm2 ; x0 | ||
| 684 : | |||
| 685 : | movq mm6, [%1+4*16] | ||
| 686 : | paddsw mm7, mm5 ; x2+x6*tg_2_16 = tp26 | ||
| 687 : | |||
| 688 : | paddsw mm2, mm6 ; x0+x4 = tp04 | ||
| 689 : | psubsw mm3, mm6 ; x0-x4 = tm04 | ||
| 690 : | |||
| 691 : | movq mm5, mm2 ; tp04 | ||
| 692 : | movq mm6, mm3 ; tm04 | ||
| 693 : | |||
| 694 : | psubsw mm2, mm7 ; tp04-tp26 = a3 | ||
| 695 : | paddsw mm3, mm0 ; tm04+tm26 = a1 | ||
| 696 : | |||
| 697 : | paddsw mm1, mm1 ; b1 | ||
| 698 : | paddsw mm4, mm4 ; b2 | ||
| 699 : | |||
| 700 : | paddsw mm5, mm7 ; tp04+tp26 = a0 | ||
| 701 : | psubsw mm6, mm0 ; tm04-tm26 = a2 | ||
| 702 : | |||
| 703 : | movq mm7, mm3 ; a1 | ||
| 704 : | movq mm0, mm6 ; a2 | ||
| 705 : | |||
| 706 : | paddsw mm3, mm1 ; a1+b1 | ||
| 707 : | paddsw mm6, mm4 ; a2+b2 | ||
| 708 : | |||
| 709 : | psraw mm3, SHIFT_INV_COL ; dst1 | ||
| 710 : | psubsw mm7, mm1 ; a1-b1 | ||
| 711 : | |||
| 712 : | psraw mm6, SHIFT_INV_COL ; dst2 | ||
| 713 : | psubsw mm0, mm4 ; a2-b2 | ||
| 714 : | |||
| 715 : | ; movq mm1, [SCRATCH+0] ; load b0 | ||
| 716 : | movq mm1, [%2+3*16] ; load b0 | ||
| 717 : | psraw mm7, SHIFT_INV_COL ; dst6 | ||
| 718 : | |||
| 719 : | movq mm4, mm5 ; a0 | ||
| 720 : | psraw mm0, SHIFT_INV_COL ; dst5 | ||
| 721 : | |||
| 722 : | movq [%2+1*16], mm3 | ||
| 723 : | paddsw mm5, mm1 ; a0+b0 | ||
| 724 : | |||
| 725 : | movq [%2+2*16], mm6 | ||
| 726 : | psubsw mm4, mm1 ; a0-b0 | ||
| 727 : | |||
| 728 : | ; movq mm3, [SCRATCH+8] ; load b3 | ||
| 729 : | movq mm3, [%2+5*16] ; load b3 | ||
| 730 : | psraw mm5, SHIFT_INV_COL ; dst0 | ||
| 731 : | |||
| 732 : | movq mm6, mm2 ; a3 | ||
| 733 : | psraw mm4, SHIFT_INV_COL ; dst7 | ||
| 734 : | |||
| 735 : | movq [%2+5*16], mm0 | ||
| 736 : | paddsw mm2, mm3 ; a3+b3 | ||
| 737 : | |||
| 738 : | movq [%2+6*16], mm7 | ||
| 739 : | psubsw mm6, mm3 ; a3-b3 | ||
| 740 : | |||
| 741 : | movq [%2+0*16], mm5 | ||
| 742 : | psraw mm2, SHIFT_INV_COL ; dst3 | ||
| 743 : | |||
| 744 : | movq [%2+7*16], mm4 | ||
| 745 : | psraw mm6, SHIFT_INV_COL ; dst4 | ||
| 746 : | |||
| 747 : | movq [%2+3*16], mm2 | ||
| 748 : | |||
| 749 : | movq [%2+4*16], mm6 | ||
| 750 : | %endmacro | ||
| 751 : | |||
| 752 : | |||
| 753 : | |||
| 754 : | section .text | ||
| 755 : | |||
| 756 : | ;============================================================================= | ||
| 757 : | ; | ||
| 758 : | ; void idct_mmx (short * const src_result); | ||
| 759 : | ; | ||
| 760 : | ;============================================================================= | ||
| 761 : | |||
| 762 : | align 16 | ||
| 763 : | cglobal idct_mmx | ||
| 764 : | idct_mmx | ||
| 765 : | mov eax, dword [esp + 4] | ||
| 766 : | |||
| 767 : | DCT_8_INV_ROW_1 eax+0, eax+0, tab_i_04, rounder_0 | ||
| 768 : | DCT_8_INV_ROW_1 eax+16, eax+16, tab_i_17, rounder_1 | ||
| 769 : | DCT_8_INV_ROW_1 eax+32, eax+32, tab_i_26, rounder_2 | ||
| 770 : | DCT_8_INV_ROW_1 eax+48, eax+48, tab_i_35, rounder_3 | ||
| 771 : | DCT_8_INV_ROW_1 eax+64, eax+64, tab_i_04, rounder_4 | ||
| 772 : | DCT_8_INV_ROW_1 eax+80, eax+80, tab_i_35, rounder_5 | ||
| 773 : | DCT_8_INV_ROW_1 eax+96, eax+96, tab_i_26, rounder_6 | ||
| 774 : | DCT_8_INV_ROW_1 eax+112, eax+112, tab_i_17, rounder_7 | ||
| 775 : | |||
| 776 : | DCT_8_INV_COL_4 eax+0,eax+0 | ||
| 777 : | DCT_8_INV_COL_4 eax+8,eax+8 | ||
| 778 : | |||
| 779 : | ret | ||
| 780 : | |||
| 781 : | |||
| 782 : | |||
| 783 : | ;============================================================================= | ||
| 784 : | ; | ||
| 785 : | ; void idct_sse (short * const src_result); | ||
| 786 : | ; | ||
| 787 : | ;============================================================================= | ||
| 788 : | |||
| 789 : | align 16 | ||
| 790 : | cglobal idct_xmm | ||
| 791 : | idct_xmm | ||
| 792 : | mov eax, dword [esp + 4] | ||
| 793 : | |||
| 794 : | DCT_8_INV_ROW_1_sse eax+0, eax+0, tab_i_04_sse, rounder_0 | ||
| 795 : | DCT_8_INV_ROW_1_sse eax+16, eax+16, tab_i_17_sse, rounder_1 | ||
| 796 : | DCT_8_INV_ROW_1_sse eax+32, eax+32, tab_i_26_sse, rounder_2 | ||
| 797 : | DCT_8_INV_ROW_1_sse eax+48, eax+48, tab_i_35_sse, rounder_3 | ||
| 798 : | DCT_8_INV_ROW_1_sse eax+64, eax+64, tab_i_04_sse, rounder_4 | ||
| 799 : | DCT_8_INV_ROW_1_sse eax+80, eax+80, tab_i_35_sse, rounder_5 | ||
| 800 : | DCT_8_INV_ROW_1_sse eax+96, eax+96, tab_i_26_sse, rounder_6 | ||
| 801 : | DCT_8_INV_ROW_1_sse eax+112, eax+112, tab_i_17_sse, rounder_7 | ||
| 802 : | |||
| 803 : | DCT_8_INV_COL_4 eax+0, eax+0 | ||
| 804 : | DCT_8_INV_COL_4 eax+8, eax+8 | ||
| 805 : | |||
| 806 : | Isibaar | 154 | ret |
| 807 : | |||
| 808 : | ;============================================================================= | ||
| 809 : | ; The code below this line is for SSE2-equipped processors | ||
| 810 : | ; By Dmitry Rozhdestvensky | ||
| 811 : | ;============================================================================= | ||
| 812 : | |||
| 813 : | section .data | ||
| 814 : | |||
| 815 : | align 16 | ||
| 816 : | |||
| 817 : | tab_i_04_s2 dw 16384, 21407, 16384, 8867 ; movq-> w05 w04 w01 w00 | ||
| 818 : | dw 16384, -8867, 16384, -21407 ; w13 w12 w09 w08 | ||
| 819 : | dw 16384, 8867, -16384, -21407 ; w07 w06 w03 w02 | ||
| 820 : | dw -16384, 21407, 16384, -8867 ; w15 w14 w11 w10 | ||
| 821 : | dw 22725, 19266, 19266, -4520 ; w21 w20 w17 w16 | ||
| 822 : | dw 12873, -22725, 4520, -12873 ; w29 w28 w25 w24 | ||
| 823 : | dw 12873, 4520, -22725, -12873 ; w23 w22 w19 w18 | ||
| 824 : | dw 4520, 19266, 19266, -22725 ; w31 w30 w27 w26 | ||
| 825 : | |||
| 826 : | ; Table for rows 1,7 - constants are multiplied by cos_1_16 | ||
| 827 : | |||
| 828 : | tab_i_17_s2 dw 22725, 29692, 22725, 12299 ; movq-> w05 w04 w01 w00 | ||
| 829 : | dw 22725, -12299, 22725, -29692 ; w13 w12 w09 w08 | ||
| 830 : | dw 22725, 12299, -22725, -29692 ; w07 w06 w03 w02 | ||
| 831 : | dw -22725, 29692, 22725, -12299 ; w15 w14 w11 w10 | ||
| 832 : | dw 31521, 26722, 26722, -6270 ; w21 w20 w17 w16 | ||
| 833 : | dw 17855, -31521, 6270, -17855 ; w29 w28 w25 w24 | ||
| 834 : | dw 17855, 6270, -31521, -17855 ; w23 w22 w19 w18 | ||
| 835 : | dw 6270, 26722, 26722, -31521 ; w31 w30 w27 w26 | ||
| 836 : | |||
| 837 : | ; Table for rows 2,6 - constants are multiplied by cos_2_16 | ||
| 838 : | |||
| 839 : | tab_i_26_s2 dw 21407, 27969, 21407, 11585 ; movq-> w05 w04 w01 w00 | ||
| 840 : | dw 21407, -11585, 21407, -27969 ; w13 w12 w09 w08 | ||
| 841 : | dw 21407, 11585, -21407, -27969 ; w07 w06 w03 w02 | ||
| 842 : | dw -21407, 27969, 21407, -11585 ; w15 w14 w11 w10 | ||
| 843 : | dw 29692, 25172, 25172, -5906 ; w21 w20 w17 w16 | ||
| 844 : | dw 16819, -29692, 5906, -16819 ; w29 w28 w25 w24 | ||
| 845 : | dw 16819, 5906, -29692, -16819 ; w23 w22 w19 w18 | ||
| 846 : | dw 5906, 25172, 25172, -29692 ; w31 w30 w27 w26 | ||
| 847 : | |||
| 848 : | ; Table for rows 3,5 - constants are multiplied by cos_3_16 | ||
| 849 : | |||
| 850 : | tab_i_35_s2 dw 19266, 25172, 19266, 10426 ; movq-> w05 w04 w01 w00 | ||
| 851 : | dw 19266, -10426, 19266, -25172 ; w13 w12 w09 w08 | ||
| 852 : | dw 19266, 10426, -19266, -25172 ; w07 w06 w03 w02 | ||
| 853 : | dw -19266, 25172, 19266, -10426 ; w15 w14 w11 w10 | ||
| 854 : | dw 26722, 22654, 22654, -5315 ; w21 w20 w17 w16 | ||
| 855 : | dw 15137, -26722, 5315, -15137 ; w29 w28 w25 w24 | ||
| 856 : | dw 15137, 5315, -26722, -15137 ; w23 w22 w19 w18 | ||
| 857 : | dw 5315, 22654, 22654, -26722 ; w31 w30 w27 w26 | ||
| 858 : | |||
| 859 : | %if SHIFT_INV_ROW == 12 ; assume SHIFT_INV_ROW == 12 | ||
| 860 : | rounder_2_0 dd 65536, 65536 | ||
| 861 : | dd 65536, 65536 | ||
| 862 : | rounder_2_4 dd 0, 0 | ||
| 863 : | dd 0, 0 | ||
| 864 : | rounder_2_1 dd 7195, 7195 | ||
| 865 : | dd 7195, 7195 | ||
| 866 : | rounder_2_7 dd 1024, 1024 | ||
| 867 : | dd 1024, 1024 | ||
| 868 : | rounder_2_2 dd 4520, 4520 | ||
| 869 : | dd 4520, 4520 | ||
| 870 : | rounder_2_6 dd 1024, 1024 | ||
| 871 : | dd 1024, 1024 | ||
| 872 : | rounder_2_3 dd 2407, 2407 | ||
| 873 : | dd 2407, 2407 | ||
| 874 : | rounder_2_5 dd 240, 240 | ||
| 875 : | dd 240, 240 | ||
| 876 : | |||
| 877 : | %elif SHIFT_INV_ROW == 11 ; assume SHIFT_INV_ROW == 11 | ||
| 878 : | rounder_2_0 dd 65536, 65536 | ||
| 879 : | dd 65536, 65536 | ||
| 880 : | rounder_2_4 dd 0, 0 | ||
| 881 : | dd 0, 0 | ||
| 882 : | rounder_2_1 dd 3597, 3597 | ||
| 883 : | dd 3597, 3597 | ||
| 884 : | rounder_2_7 dd 512, 512 | ||
| 885 : | dd 512, 512 | ||
| 886 : | rounder_2_2 dd 2260, 2260 | ||
| 887 : | dd 2260, 2260 | ||
| 888 : | rounder_2_6 dd 512, 512 | ||
| 889 : | dd 512, 512 | ||
| 890 : | rounder_2_3 dd 1203, 1203 | ||
| 891 : | dd 1203, 1203 | ||
| 892 : | rounder_2_5 dd 120, 120 | ||
| 893 : | dd 120, 120 | ||
| 894 : | %else | ||
| 895 : | |||
| 896 : | %error invalid _SHIFT_INV_ROW_ | ||
| 897 : | |||
| 898 : | %endif | ||
| 899 : | |||
| 900 : | tg_1_16_2 dw 13036, 13036, 13036, 13036 ; tg * (2<<16) + 0.5 | ||
| 901 : | dw 13036, 13036, 13036, 13036 | ||
| 902 : | tg_2_16_2 dw 27146, 27146, 27146, 27146 ; tg * (2<<16) + 0.5 | ||
| 903 : | dw 27146, 27146, 27146, 27146 | ||
| 904 : | tg_3_16_2 dw -21746, -21746, -21746, -21746 ; tg * (2<<16) + 0.5 | ||
| 905 : | dw -21746, -21746, -21746, -21746 | ||
| 906 : | ocos_4_16_2 dw 23170, 23170, 23170, 23170 ; cos * (2<<15) + 0.5 | ||
| 907 : | dw 23170, 23170, 23170, 23170 | ||
| 908 : | |||
| 909 : | %macro DCT_8_INV_ROW_1_sse2 4 | ||
| 910 : | |||
| 911 : | pshufhw xmm1,[%1],11011000b ;x 75643210 | ||
| 912 : | pshuflw xmm1,xmm1,11011000b ;x 75643120 | ||
| 913 : | pshufd xmm0,xmm1,00000000b ;x 20202020 | ||
| 914 : | pmaddwd xmm0,[%3] ;w 13 12 9 8 5410 | ||
| 915 : | ;a 3210 first part | ||
| 916 : | |||
| 917 : | pshufd xmm2,xmm1,10101010b ;x 64646464 | ||
| 918 : | pmaddwd xmm2,[%3+16] ;w 15 14 11 10 7632 | ||
| 919 : | ;a 3210 second part | ||
| 920 : | |||
| 921 : | paddd xmm2,xmm0 ;a 3210 ready | ||
| 922 : | paddd xmm2,[%4] ;must be 4 dwords long, not 2 as for sse1 | ||
| 923 : | movdqa xmm5,xmm2 | ||
| 924 : | |||
| 925 : | pshufd xmm3,xmm1,01010101b ;x 31313131 | ||
| 926 : | pmaddwd xmm3,[%3+32] ;w 29 28 25 24 21 20 17 16 | ||
| 927 : | ;b 3210 first part | ||
| 928 : | |||
| 929 : | pshufd xmm4,xmm1,11111111b ;x 75757575 | ||
| 930 : | pmaddwd xmm4,[%3+48] ;w 31 30 27 26 23 22 19 18 | ||
| 931 : | ;b 3210 second part | ||
| 932 : | paddd xmm3,xmm4 ;b 3210 ready | ||
| 933 : | |||
| 934 : | paddd xmm2,xmm3 ;will be y 3210 | ||
| 935 : | psubd xmm5,xmm3 ;will be y 4567 | ||
| 936 : | psrad xmm2,SHIFT_INV_ROW | ||
| 937 : | psrad xmm5,SHIFT_INV_ROW | ||
| 938 : | packssdw xmm2,xmm5 ;y 45673210 | ||
| 939 : | pshufhw xmm6,xmm2,00011011b ;y 76543210 | ||
| 940 : | movdqa [%2],xmm6 | ||
| 941 : | |||
| 942 : | %endmacro | ||
| 943 : | |||
| 944 : | %macro DCT_8_INV_COL_4_sse2 2 | ||
| 945 : | |||
| 946 : | movdqa xmm0,[%1+16*0] ;x0 (all columns) | ||
| 947 : | movdqa xmm2,[%1+16*4] ;x4 | ||
| 948 : | movdqa xmm1,xmm0 | ||
| 949 : | |||
| 950 : | movdqa xmm4,[%1+16*2] ;x2 | ||
| 951 : | movdqa xmm5,[%1+16*6] ;x6 | ||
| 952 : | movdqa xmm6,[tg_2_16_2] | ||
| 953 : | movdqa xmm7,xmm6 | ||
| 954 : | |||
| 955 : | paddsw xmm0,xmm2 ;u04=x0+x4 | ||
| 956 : | psubsw xmm1,xmm2 ;v04=x0-x4 | ||
| 957 : | movdqa xmm3,xmm0 | ||
| 958 : | movdqa xmm2,xmm1 | ||
| 959 : | |||
| 960 : | pmulhw xmm6,xmm4 | ||
| 961 : | pmulhw xmm7,xmm5 | ||
| 962 : | psubsw xmm6,xmm5 ;v26=x2*T2-x6 | ||
| 963 : | paddsw xmm7,xmm4 ;u26=x6*T2+x2 | ||
| 964 : | |||
| 965 : | paddsw xmm1,xmm6 ;a1=v04+v26 | ||
| 966 : | paddsw xmm0,xmm7 ;a0=u04+u26 | ||
| 967 : | psubsw xmm2,xmm6 ;a2=v04-v26 | ||
| 968 : | psubsw xmm3,xmm7 ;a3=u04-u26 | ||
| 969 : | |||
| 970 : | movdqa [%2+16*0],xmm0 ;store a3-a0 to | ||
| 971 : | movdqa [%2+16*6],xmm1 ;free registers | ||
| 972 : | movdqa [%2+16*2],xmm2 | ||
| 973 : | movdqa [%2+16*4],xmm3 | ||
| 974 : | |||
| 975 : | movdqa xmm0,[%1+16*1] ;x1 | ||
| 976 : | movdqa xmm1,[%1+16*7] ;x7 | ||
| 977 : | movdqa xmm2,[tg_1_16_2] | ||
| 978 : | movdqa xmm3,xmm2 | ||
| 979 : | |||
| 980 : | movdqa xmm4,[%1+16*3] ;x3 | ||
| 981 : | movdqa xmm5,[%1+16*5] ;x5 | ||
| 982 : | movdqa xmm6,[tg_3_16_2] | ||
| 983 : | movdqa xmm7,xmm6 | ||
| 984 : | |||
| 985 : | pmulhw xmm2,xmm0 | ||
| 986 : | pmulhw xmm3,xmm1 | ||
| 987 : | psubsw xmm2,xmm1 ;v17=x1*T1-x7 | ||
| 988 : | paddsw xmm3,xmm0 ;u17=x7*T1+x1 | ||
| 989 : | movdqa xmm0,xmm3 ;u17 | ||
| 990 : | movdqa xmm1,xmm2 ;v17 | ||
| 991 : | |||
| 992 : | pmulhw xmm6,xmm4 ;x3*(t3-1) | ||
| 993 : | pmulhw xmm7,xmm5 ;x5*(t3-1) | ||
| 994 : | paddsw xmm6,xmm4 | ||
| 995 : | paddsw xmm7,xmm5 | ||
| 996 : | psubsw xmm6,xmm5 ;v35=x3*T3-x5 | ||
| 997 : | paddsw xmm7,xmm4 ;u35=x5*T3+x3 | ||
| 998 : | |||
| 999 : | movdqa xmm4,[ocos_4_16_2] | ||
| 1000 : | |||
| 1001 : | paddsw xmm0,xmm7 ;b0=u17+u35 | ||
| 1002 : | psubsw xmm1,xmm6 ;b3=v17-v35 | ||
| 1003 : | psubsw xmm3,xmm7 ;u12=u17-v35 | ||
| 1004 : | paddsw xmm2,xmm6 ;v12=v17+v35 | ||
| 1005 : | |||
| 1006 : | movdqa xmm5,xmm3 | ||
| 1007 : | paddsw xmm3,xmm2 ;tb1 | ||
| 1008 : | psubsw xmm5,xmm2 ;tb2 | ||
| 1009 : | pmulhw xmm5,xmm4 | ||
| 1010 : | pmulhw xmm4,xmm3 | ||
| 1011 : | paddsw xmm5,xmm5 | ||
| 1012 : | paddsw xmm4,xmm4 | ||
| 1013 : | |||
| 1014 : | movdqa xmm6,[%2+16*0] ;a0 | ||
| 1015 : | movdqa xmm7,xmm6 | ||
| 1016 : | movdqa xmm2,[%2+16*4] ;a3 | ||
| 1017 : | movdqa xmm3,xmm2 | ||
| 1018 : | |||
| 1019 : | paddsw xmm6,xmm0 | ||
| 1020 : | psubsw xmm7,xmm0 | ||
| 1021 : | psraw xmm6,SHIFT_INV_COL ;y0=a0+b0 | ||
| 1022 : | psraw xmm7,SHIFT_INV_COL ;y7=a0-b0 | ||
| 1023 : | movdqa [%2+16*0],xmm6 | ||
| 1024 : | movdqa [%2+16*7],xmm7 | ||
| 1025 : | |||
| 1026 : | paddsw xmm2,xmm1 | ||
| 1027 : | psubsw xmm3,xmm1 | ||
| 1028 : | psraw xmm2,SHIFT_INV_COL ;y3=a3+b3 | ||
| 1029 : | psraw xmm3,SHIFT_INV_COL ;y4=a3-b3 | ||
| 1030 : | movdqa [%2+16*3],xmm2 | ||
| 1031 : | movdqa [%2+16*4],xmm3 | ||
| 1032 : | |||
| 1033 : | movdqa xmm0,[%2+16*6] ;a1 | ||
| 1034 : | movdqa xmm1,xmm0 | ||
| 1035 : | movdqa xmm6,[%2+16*2] ;a2 | ||
| 1036 : | movdqa xmm7,xmm6 | ||
| 1037 : | |||
| 1038 : | |||
| 1039 : | paddsw xmm0,xmm4 | ||
| 1040 : | psubsw xmm1,xmm4 | ||
| 1041 : | psraw xmm0,SHIFT_INV_COL ;y1=a1+b1 | ||
| 1042 : | psraw xmm1,SHIFT_INV_COL ;y6=a1-b1 | ||
| 1043 : | movdqa [%2+16*1],xmm0 | ||
| 1044 : | movdqa [%2+16*6],xmm1 | ||
| 1045 : | |||
| 1046 : | paddsw xmm6,xmm5 | ||
| 1047 : | psubsw xmm7,xmm5 | ||
| 1048 : | psraw xmm6,SHIFT_INV_COL ;y2=a2+b2 | ||
| 1049 : | psraw xmm7,SHIFT_INV_COL ;y5=a2-b2 | ||
| 1050 : | movdqa [%2+16*2],xmm6 | ||
| 1051 : | movdqa [%2+16*5],xmm7 | ||
| 1052 : | |||
| 1053 : | %endmacro | ||
| 1054 : | |||
| 1055 : | section .text | ||
| 1056 : | |||
| 1057 : | align 16 | ||
| 1058 : | cglobal idct_sse2 | ||
| 1059 : | idct_sse2 | ||
| 1060 : | |||
| 1061 : | mov eax, dword [esp + 4] | ||
| 1062 : | |||
| 1063 : | DCT_8_INV_ROW_1_sse2 eax+0, eax+0, tab_i_04_s2, rounder_2_0 | ||
| 1064 : | DCT_8_INV_ROW_1_sse2 eax+16, eax+16, tab_i_17_s2, rounder_2_1 | ||
| 1065 : | DCT_8_INV_ROW_1_sse2 eax+32, eax+32, tab_i_26_s2, rounder_2_2 | ||
| 1066 : | DCT_8_INV_ROW_1_sse2 eax+48, eax+48, tab_i_35_s2, rounder_2_3 | ||
| 1067 : | DCT_8_INV_ROW_1_sse2 eax+64, eax+64, tab_i_04_s2, rounder_2_4 | ||
| 1068 : | DCT_8_INV_ROW_1_sse2 eax+80, eax+80, tab_i_35_s2, rounder_2_5 | ||
| 1069 : | DCT_8_INV_ROW_1_sse2 eax+96, eax+96, tab_i_26_s2, rounder_2_6 | ||
| 1070 : | DCT_8_INV_ROW_1_sse2 eax+112, eax+112, tab_i_17_s2, rounder_2_7 | ||
| 1071 : | |||
| 1072 : | DCT_8_INV_COL_4_sse2 eax, eax | ||
| 1073 : | ;DCT_8_INV_COL_4 eax+8, eax+8 | ||
| 1074 : | |||
| 1075 : | ret |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |