| 1 |
/************************************************************************** |
/***************************************************************************** |
| 2 |
* |
* |
| 3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
| 4 |
* native api |
* - Native API implementation - |
| 5 |
* |
* |
| 6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
|
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
| 7 |
* |
* |
| 8 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
| 9 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
| 17 |
* |
* |
| 18 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
| 19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
| 20 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
| 21 |
* |
* |
| 22 |
*************************************************************************/ |
* $Id: xvid.c,v 1.45.2.13 2003-08-22 15:52:35 Isibaar Exp $ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
| 23 |
* |
* |
| 24 |
* 17.03.2002 Added interpolate8x8_halfpel_hv_xmm |
****************************************************************************/ |
|
* 22.12.2001 API change: added xvid_init() - Isibaar |
|
|
* 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
*************************************************************************/ |
|
| 25 |
|
|
| 26 |
|
#include <stdio.h> |
| 27 |
|
#include <stdlib.h> |
| 28 |
|
#include <string.h> |
| 29 |
|
#include <time.h> |
| 30 |
|
|
| 31 |
#include "xvid.h" |
#include "xvid.h" |
| 32 |
#include "decoder.h" |
#include "decoder.h" |
| 36 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
| 37 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
| 38 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
| 39 |
|
#include "image/reduced.h" |
| 40 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
| 41 |
|
#include "utils/mbfunctions.h" |
| 42 |
#include "quant/quant_h263.h" |
#include "quant/quant_h263.h" |
| 43 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_mpeg4.h" |
| 44 |
|
#include "motion/motion.h" |
| 45 |
#include "motion/sad.h" |
#include "motion/sad.h" |
| 46 |
#include "utils/emms.h" |
#include "utils/emms.h" |
| 47 |
#include "utils/timer.h" |
#include "utils/timer.h" |
| 48 |
|
#include "bitstream/mbcoding.h" |
| 49 |
|
#include "image/qpel.h" |
| 50 |
|
|
| 51 |
|
#if defined(_DEBUG) |
| 52 |
|
unsigned int xvid_debug = 0; /* xvid debug mask */ |
| 53 |
|
#endif |
| 54 |
|
|
| 55 |
|
#if defined(ARCH_IS_IA32) |
| 56 |
|
|
| 57 |
|
#if defined(_MSC_VER) |
| 58 |
|
# include <windows.h> |
| 59 |
|
#else |
| 60 |
|
# include <signal.h> |
| 61 |
|
# include <setjmp.h> |
| 62 |
|
|
| 63 |
int xvid_init(void *handle, int opt, void *param1, void *param2) |
static jmp_buf mark; |
| 64 |
|
|
| 65 |
|
static void |
| 66 |
|
sigill_handler(int signal) |
| 67 |
{ |
{ |
| 68 |
int cpu_flags; |
longjmp(mark, 1); |
| 69 |
XVID_INIT_PARAM *init_param; |
} |
| 70 |
|
#endif |
| 71 |
|
|
|
init_param = (XVID_INIT_PARAM *) param1; |
|
| 72 |
|
|
| 73 |
// force specific cpu settings? |
/* |
| 74 |
if((init_param->cpu_flags & XVID_CPU_FORCE) > 0) |
* Calls the funcptr, and returns whether SIGILL (illegal instruction) was |
| 75 |
cpu_flags = init_param->cpu_flags; |
* signalled |
| 76 |
else { |
* |
| 77 |
|
* Return values: |
| 78 |
|
* -1 : could not determine |
| 79 |
|
* 0 : SIGILL was *not* signalled |
| 80 |
|
* 1 : SIGILL was signalled |
| 81 |
|
*/ |
| 82 |
|
|
| 83 |
|
int |
| 84 |
|
sigill_check(void (*func)()) |
| 85 |
|
{ |
| 86 |
|
#if defined(_MSC_VER) |
| 87 |
|
_try { |
| 88 |
|
func(); |
| 89 |
|
} |
| 90 |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
| 91 |
|
|
| 92 |
#ifdef ARCH_X86 |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
| 93 |
cpu_flags = check_cpu_features(); |
return 1; |
| 94 |
|
} |
| 95 |
|
return 0; |
| 96 |
#else |
#else |
| 97 |
cpu_flags = 0; |
void * old_handler; |
| 98 |
|
int jmpret; |
| 99 |
|
|
| 100 |
|
|
| 101 |
|
old_handler = signal(SIGILL, sigill_handler); |
| 102 |
|
if (old_handler == SIG_ERR) |
| 103 |
|
{ |
| 104 |
|
return -1; |
| 105 |
|
} |
| 106 |
|
|
| 107 |
|
jmpret = setjmp(mark); |
| 108 |
|
if (jmpret == 0) |
| 109 |
|
{ |
| 110 |
|
func(); |
| 111 |
|
} |
| 112 |
|
|
| 113 |
|
signal(SIGILL, old_handler); |
| 114 |
|
|
| 115 |
|
return jmpret; |
| 116 |
|
#endif |
| 117 |
|
} |
| 118 |
|
#endif |
| 119 |
|
|
| 120 |
|
|
| 121 |
|
/* detect cpu flags */ |
| 122 |
|
static unsigned int |
| 123 |
|
detect_cpu_flags() |
| 124 |
|
{ |
| 125 |
|
/* enable native assembly optimizations by default */ |
| 126 |
|
unsigned int cpu_flags = XVID_CPU_ASM; |
| 127 |
|
|
| 128 |
|
#if defined(ARCH_IS_IA32) |
| 129 |
|
cpu_flags |= check_cpu_features(); |
| 130 |
|
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
| 131 |
|
cpu_flags &= ~XVID_CPU_SSE; |
| 132 |
|
|
| 133 |
|
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
| 134 |
|
cpu_flags &= ~XVID_CPU_SSE2; |
| 135 |
|
#endif |
| 136 |
|
|
| 137 |
|
#if defined(ARCH_IS_PPC) |
| 138 |
|
#if defined(ARCH_IS_PPC_ALTIVEC) |
| 139 |
|
cpu_flags |= XVID_CPU_ALTIVEC; |
| 140 |
#endif |
#endif |
| 141 |
init_param->cpu_flags = cpu_flags; |
#endif |
| 142 |
|
|
| 143 |
|
return cpu_flags; |
| 144 |
} |
} |
| 145 |
|
|
| 146 |
// initialize the function pointers |
|
| 147 |
|
/***************************************************************************** |
| 148 |
|
* XviD Init Entry point |
| 149 |
|
* |
| 150 |
|
* Well this function initialize all internal function pointers according |
| 151 |
|
* to the CPU features forced by the library client or autodetected (depending |
| 152 |
|
* on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all |
| 153 |
|
* image colorspace transformation tables. |
| 154 |
|
* |
| 155 |
|
* Returned value : XVID_ERR_OK |
| 156 |
|
* + API_VERSION in the input XVID_INIT_PARAM structure |
| 157 |
|
* + core build " " " " " |
| 158 |
|
* |
| 159 |
|
****************************************************************************/ |
| 160 |
|
|
| 161 |
|
|
| 162 |
|
static |
| 163 |
|
int xvid_gbl_init(xvid_gbl_init_t * init) |
| 164 |
|
{ |
| 165 |
|
unsigned int cpu_flags; |
| 166 |
|
|
| 167 |
|
if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */ |
| 168 |
|
return XVID_ERR_VERSION; |
| 169 |
|
|
| 170 |
|
cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
| 171 |
|
|
| 172 |
|
/* Initialize the function pointers */ |
| 173 |
idct_int32_init(); |
idct_int32_init(); |
| 174 |
|
init_vlc_tables(); |
| 175 |
|
|
| 176 |
|
/* Fixed Point Forward/Inverse DCT transformations */ |
| 177 |
fdct = fdct_int32; |
fdct = fdct_int32; |
| 178 |
idct = idct_int32; |
idct = idct_int32; |
| 179 |
|
|
| 180 |
|
/* Only needed on PPC Altivec archs */ |
| 181 |
|
sadInit = 0; |
| 182 |
|
|
| 183 |
|
/* Restore FPU context : emms_c is a nop functions */ |
| 184 |
emms = emms_c; |
emms = emms_c; |
| 185 |
|
|
| 186 |
|
/* Qpel stuff */ |
| 187 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_C; |
| 188 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_C; |
| 189 |
|
xvid_Init_QP_mmx(); |
| 190 |
|
|
| 191 |
|
/* Quantization functions */ |
| 192 |
quant_intra = quant_intra_c; |
quant_intra = quant_intra_c; |
| 193 |
dequant_intra = dequant_intra_c; |
dequant_intra = dequant_intra_c; |
| 194 |
quant_inter = quant_inter_c; |
quant_inter = quant_inter_c; |
| 199 |
quant4_inter = quant4_inter_c; |
quant4_inter = quant4_inter_c; |
| 200 |
dequant4_inter = dequant4_inter_c; |
dequant4_inter = dequant4_inter_c; |
| 201 |
|
|
| 202 |
|
/* Block transfer related functions */ |
| 203 |
transfer_8to16copy = transfer_8to16copy_c; |
transfer_8to16copy = transfer_8to16copy_c; |
| 204 |
transfer_16to8copy = transfer_16to8copy_c; |
transfer_16to8copy = transfer_16to8copy_c; |
| 205 |
transfer_8to16sub = transfer_8to16sub_c; |
transfer_8to16sub = transfer_8to16sub_c; |
| 206 |
|
transfer_8to16subro = transfer_8to16subro_c; |
| 207 |
|
transfer_8to16sub2 = transfer_8to16sub2_c; |
| 208 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
| 209 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
| 210 |
|
|
| 211 |
|
/* Interlacing functions */ |
| 212 |
|
MBFieldTest = MBFieldTest_c; |
| 213 |
|
|
| 214 |
|
/* Image interpolation related functions */ |
| 215 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
| 216 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
| 217 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
| 218 |
|
|
| 219 |
|
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
| 220 |
|
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
| 221 |
|
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
| 222 |
|
|
| 223 |
|
interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; |
| 224 |
|
interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; |
| 225 |
|
interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; |
| 226 |
|
|
| 227 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; |
| 228 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; |
| 229 |
|
|
| 230 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
| 231 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
| 232 |
|
|
| 233 |
|
/* reduced resoltuion */ |
| 234 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
| 235 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
| 236 |
|
vfilter_31 = xvid_VFilter_31_C; |
| 237 |
|
hfilter_31 = xvid_HFilter_31_C; |
| 238 |
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
| 239 |
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
| 240 |
|
|
| 241 |
|
/* Initialize internal colorspace transformation tables */ |
| 242 |
colorspace_init(); |
colorspace_init(); |
| 243 |
|
|
| 244 |
|
/* All colorspace transformation functions User Format->YV12 */ |
| 245 |
|
yv12_to_yv12 = yv12_to_yv12_c; |
| 246 |
rgb555_to_yv12 = rgb555_to_yv12_c; |
rgb555_to_yv12 = rgb555_to_yv12_c; |
| 247 |
rgb565_to_yv12 = rgb565_to_yv12_c; |
rgb565_to_yv12 = rgb565_to_yv12_c; |
| 248 |
rgb24_to_yv12 = rgb24_to_yv12_c; |
bgr_to_yv12 = bgr_to_yv12_c; |
| 249 |
rgb32_to_yv12 = rgb32_to_yv12_c; |
bgra_to_yv12 = bgra_to_yv12_c; |
| 250 |
yuv_to_yv12 = yuv_to_yv12_c; |
abgr_to_yv12 = abgr_to_yv12_c; |
| 251 |
|
rgba_to_yv12 = rgba_to_yv12_c; |
| 252 |
yuyv_to_yv12 = yuyv_to_yv12_c; |
yuyv_to_yv12 = yuyv_to_yv12_c; |
| 253 |
uyvy_to_yv12 = uyvy_to_yv12_c; |
uyvy_to_yv12 = uyvy_to_yv12_c; |
| 254 |
|
|
| 255 |
|
rgb555i_to_yv12 = rgb555i_to_yv12_c; |
| 256 |
|
rgb565i_to_yv12 = rgb565i_to_yv12_c; |
| 257 |
|
bgri_to_yv12 = bgri_to_yv12_c; |
| 258 |
|
bgrai_to_yv12 = bgrai_to_yv12_c; |
| 259 |
|
abgri_to_yv12 = abgri_to_yv12_c; |
| 260 |
|
rgbai_to_yv12 = rgbai_to_yv12_c; |
| 261 |
|
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
| 262 |
|
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
| 263 |
|
|
| 264 |
|
|
| 265 |
|
/* All colorspace transformation functions YV12->User format */ |
| 266 |
yv12_to_rgb555 = yv12_to_rgb555_c; |
yv12_to_rgb555 = yv12_to_rgb555_c; |
| 267 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
| 268 |
yv12_to_rgb24 = yv12_to_rgb24_c; |
yv12_to_bgr = yv12_to_bgr_c; |
| 269 |
yv12_to_rgb32 = yv12_to_rgb32_c; |
yv12_to_bgra = yv12_to_bgra_c; |
| 270 |
yv12_to_yuv = yv12_to_yuv_c; |
yv12_to_abgr = yv12_to_abgr_c; |
| 271 |
|
yv12_to_rgba = yv12_to_rgba_c; |
| 272 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
| 273 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
| 274 |
|
|
| 275 |
|
yv12_to_rgb555i = yv12_to_rgb555i_c; |
| 276 |
|
yv12_to_rgb565i = yv12_to_rgb565i_c; |
| 277 |
|
yv12_to_bgri = yv12_to_bgri_c; |
| 278 |
|
yv12_to_bgrai = yv12_to_bgrai_c; |
| 279 |
|
yv12_to_abgri = yv12_to_abgri_c; |
| 280 |
|
yv12_to_rgbai = yv12_to_rgbai_c; |
| 281 |
|
yv12_to_yuyvi = yv12_to_yuyvi_c; |
| 282 |
|
yv12_to_uyvyi = yv12_to_uyvyi_c; |
| 283 |
|
|
| 284 |
|
/* Functions used in motion estimation algorithms */ |
| 285 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
| 286 |
sad16 = sad16_c; |
sad16 = sad16_c; |
| 287 |
sad8 = sad8_c; |
sad8 = sad8_c; |
| 288 |
|
sad16bi = sad16bi_c; |
| 289 |
|
sad8bi = sad8bi_c; |
| 290 |
dev16 = dev16_c; |
dev16 = dev16_c; |
| 291 |
|
sad16v = sad16v_c; |
| 292 |
|
|
| 293 |
|
/* Halfpel8_Refine = Halfpel8_Refine_c; */ |
| 294 |
|
|
| 295 |
|
#if defined(ARCH_IS_IA32) |
| 296 |
|
|
| 297 |
|
if ((cpu_flags & XVID_CPU_ASM)) { |
| 298 |
|
vfilter_31 = xvid_VFilter_31_x86; |
| 299 |
|
hfilter_31 = xvid_HFilter_31_x86; |
| 300 |
|
} |
| 301 |
|
|
| 302 |
|
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
| 303 |
|
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
| 304 |
|
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
| 305 |
|
{ |
| 306 |
|
/* Restore FPU context : emms_c is a nop functions */ |
| 307 |
|
emms = emms_mmx; |
| 308 |
|
} |
| 309 |
|
|
| 310 |
|
if ((cpu_flags & XVID_CPU_MMX)) { |
| 311 |
|
|
| 312 |
#ifdef ARCH_X86 |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
|
if((cpu_flags & XVID_CPU_MMX) > 0) { |
|
| 313 |
fdct = fdct_mmx; |
fdct = fdct_mmx; |
| 314 |
idct = idct_mmx; |
idct = idct_mmx; |
| 315 |
|
|
| 316 |
emms = emms_mmx; |
/* Qpel stuff */ |
| 317 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_mmx; |
| 318 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx; |
| 319 |
|
|
| 320 |
|
/* Quantization related functions */ |
| 321 |
quant_intra = quant_intra_mmx; |
quant_intra = quant_intra_mmx; |
| 322 |
dequant_intra = dequant_intra_mmx; |
dequant_intra = dequant_intra_mmx; |
| 323 |
quant_inter = quant_inter_mmx; |
quant_inter = quant_inter_mmx; |
| 328 |
quant4_inter = quant4_inter_mmx; |
quant4_inter = quant4_inter_mmx; |
| 329 |
dequant4_inter = dequant4_inter_mmx; |
dequant4_inter = dequant4_inter_mmx; |
| 330 |
|
|
| 331 |
|
/* Block related functions */ |
| 332 |
transfer_8to16copy = transfer_8to16copy_mmx; |
transfer_8to16copy = transfer_8to16copy_mmx; |
| 333 |
transfer_16to8copy = transfer_16to8copy_mmx; |
transfer_16to8copy = transfer_16to8copy_mmx; |
| 334 |
transfer_8to16sub = transfer_8to16sub_mmx; |
transfer_8to16sub = transfer_8to16sub_mmx; |
| 335 |
|
transfer_8to16subro = transfer_8to16subro_mmx; |
| 336 |
|
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
| 337 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
| 338 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
| 339 |
|
|
| 340 |
|
/* Interlacing Functions */ |
| 341 |
|
MBFieldTest = MBFieldTest_mmx; |
| 342 |
|
|
| 343 |
|
/* Image Interpolation related functions */ |
| 344 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
| 345 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
| 346 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
| 347 |
|
|
| 348 |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
| 349 |
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
| 350 |
yuv_to_yv12 = yuv_to_yv12_mmx; |
|
| 351 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
| 352 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
| 353 |
|
|
| 354 |
|
/* reduced resolution */ |
| 355 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; |
| 356 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; |
| 357 |
|
hfilter_31 = xvid_HFilter_31_mmx; |
| 358 |
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; |
| 359 |
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; |
| 360 |
|
|
| 361 |
|
/* image input xxx_to_yv12 related functions */ |
| 362 |
|
yv12_to_yv12 = yv12_to_yv12_mmx; |
| 363 |
|
bgr_to_yv12 = bgr_to_yv12_mmx; |
| 364 |
|
bgra_to_yv12 = bgra_to_yv12_mmx; |
| 365 |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
| 366 |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
| 367 |
|
|
| 368 |
yv12_to_rgb24 = yv12_to_rgb24_mmx; |
/* image output yv12_to_xxx related functions */ |
| 369 |
yv12_to_rgb32 = yv12_to_rgb32_mmx; |
yv12_to_bgr = yv12_to_bgr_mmx; |
| 370 |
|
yv12_to_bgra = yv12_to_bgra_mmx; |
| 371 |
yv12_to_yuyv = yv12_to_yuyv_mmx; |
yv12_to_yuyv = yv12_to_yuyv_mmx; |
| 372 |
yv12_to_uyvy = yv12_to_uyvy_mmx; |
yv12_to_uyvy = yv12_to_uyvy_mmx; |
| 373 |
|
|
| 374 |
|
yv12_to_yuyvi = yv12_to_yuyvi_mmx; |
| 375 |
|
yv12_to_uyvyi = yv12_to_uyvyi_mmx; |
| 376 |
|
|
| 377 |
|
/* Motion estimation related functions */ |
| 378 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
| 379 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
| 380 |
sad8 = sad8_mmx; |
sad8 = sad8_mmx; |
| 381 |
|
sad16bi = sad16bi_mmx; |
| 382 |
|
sad8bi = sad8bi_mmx; |
| 383 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
| 384 |
|
sad16v = sad16v_mmx; |
| 385 |
|
} |
| 386 |
|
|
| 387 |
|
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
| 388 |
|
if ((cpu_flags & XVID_CPU_3DNOW)) { |
| 389 |
|
|
| 390 |
|
emms = emms_3dn; |
| 391 |
|
|
| 392 |
|
/* ME functions */ |
| 393 |
|
sad16bi = sad16bi_3dn; |
| 394 |
|
sad8bi = sad8bi_3dn; |
| 395 |
|
|
| 396 |
|
yuyv_to_yv12 = yuyv_to_yv12_3dn; |
| 397 |
|
uyvy_to_yv12 = uyvy_to_yv12_3dn; |
| 398 |
} |
} |
| 399 |
|
|
| 400 |
if((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
|
| 401 |
|
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
| 402 |
|
|
| 403 |
|
/* Inverse DCT */ |
| 404 |
idct = idct_xmm; |
idct = idct_xmm; |
| 405 |
|
|
| 406 |
|
/* Interpolation */ |
| 407 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; |
| 408 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
| 409 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
|
yuv_to_yv12 = yuv_to_yv12_xmm; |
|
| 410 |
|
|
| 411 |
|
/* reduced resolution */ |
| 412 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
| 413 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
| 414 |
|
|
| 415 |
|
/* Quantization */ |
| 416 |
|
quant4_intra = quant4_intra_xmm; |
| 417 |
|
quant4_inter = quant4_inter_xmm; |
| 418 |
|
|
| 419 |
|
dequant_intra = dequant_intra_xmm; |
| 420 |
|
dequant_inter = dequant_inter_xmm; |
| 421 |
|
|
| 422 |
|
/* Buffer transfer */ |
| 423 |
|
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
| 424 |
|
|
| 425 |
|
/* Colorspace transformation */ |
| 426 |
|
yv12_to_yv12 = yv12_to_yv12_xmm; |
| 427 |
|
yuyv_to_yv12 = yuyv_to_yv12_xmm; |
| 428 |
|
uyvy_to_yv12 = uyvy_to_yv12_xmm; |
| 429 |
|
|
| 430 |
|
/* ME functions */ |
| 431 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
| 432 |
sad8 = sad8_xmm; |
sad8 = sad8_xmm; |
| 433 |
|
sad16bi = sad16bi_xmm; |
| 434 |
|
sad8bi = sad8bi_xmm; |
| 435 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
| 436 |
|
sad16v = sad16v_xmm; |
| 437 |
} |
} |
| 438 |
|
|
| 439 |
if((cpu_flags & XVID_CPU_3DNOW) > 0) { |
if ((cpu_flags & XVID_CPU_3DNOW)) { |
| 440 |
|
|
| 441 |
|
/* Interpolation */ |
| 442 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
| 443 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; |
| 444 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
| 445 |
|
} |
| 446 |
|
|
| 447 |
|
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
| 448 |
|
|
| 449 |
|
/* Inverse DCT */ |
| 450 |
|
idct = idct_3dne; |
| 451 |
|
|
| 452 |
|
/* Buffer transfer */ |
| 453 |
|
transfer_8to16copy = transfer_8to16copy_3dne; |
| 454 |
|
transfer_16to8copy = transfer_16to8copy_3dne; |
| 455 |
|
transfer_8to16sub = transfer_8to16sub_3dne; |
| 456 |
|
transfer_8to16subro = transfer_8to16subro_3dne; |
| 457 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
| 458 |
|
transfer_16to8add = transfer_16to8add_3dne; |
| 459 |
|
transfer8x8_copy = transfer8x8_copy_3dne; |
| 460 |
|
|
| 461 |
|
/* Quantization */ |
| 462 |
|
dequant4_intra = dequant4_intra_3dne; |
| 463 |
|
dequant4_inter = dequant4_inter_3dne; |
| 464 |
|
quant_intra = quant_intra_3dne; |
| 465 |
|
quant_inter = quant_inter_3dne; |
| 466 |
|
dequant_intra = dequant_intra_3dne; |
| 467 |
|
dequant_inter = dequant_inter_3dne; |
| 468 |
|
|
| 469 |
|
/* ME functions */ |
| 470 |
|
calc_cbp = calc_cbp_3dne; |
| 471 |
|
sad16 = sad16_3dne; |
| 472 |
|
sad8 = sad8_3dne; |
| 473 |
|
sad16bi = sad16bi_3dne; |
| 474 |
|
sad8bi = sad8bi_3dne; |
| 475 |
|
dev16 = dev16_3dne; |
| 476 |
|
|
| 477 |
|
/* Interpolation */ |
| 478 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
| 479 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
| 480 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
| 481 |
|
} |
| 482 |
|
|
| 483 |
|
#if defined(EXPERIMENTAL_SSE2_CODE) /* mark the whole SSE2 stuff as experimental. At least on |
| 484 |
|
my P4, it crashes... */ |
| 485 |
|
if ((cpu_flags & XVID_CPU_SSE2)) { |
| 486 |
|
|
| 487 |
|
calc_cbp = calc_cbp_sse2; |
| 488 |
|
|
| 489 |
|
/* Quantization */ |
| 490 |
|
quant_intra = quant_intra_sse2; |
| 491 |
|
dequant_intra = dequant_intra_sse2; |
| 492 |
|
quant_inter = quant_inter_sse2; |
| 493 |
|
dequant_inter = dequant_inter_sse2; |
| 494 |
|
|
| 495 |
|
/* ME; slower than xmm */ |
| 496 |
|
sad16 = sad16_sse2; |
| 497 |
|
dev16 = dev16_sse2; |
| 498 |
|
/* Forward and Inverse DCT */ |
| 499 |
|
#if 0 /* Both function are known to be unprecise, better keep them deactivated */ |
| 500 |
|
idct = idct_sse2; |
| 501 |
|
fdct = fdct_sse2; |
| 502 |
|
#endif |
| 503 |
|
} |
| 504 |
|
#endif |
| 505 |
|
#endif |
| 506 |
|
|
| 507 |
|
#if defined(ARCH_IS_IA64) |
| 508 |
|
if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
| 509 |
|
idct_ia64_init(); |
| 510 |
|
fdct = fdct_ia64; |
| 511 |
|
idct = idct_ia64; /*not yet working, crashes */ |
| 512 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
| 513 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
| 514 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
| 515 |
|
sad16 = sad16_ia64; |
| 516 |
|
sad16bi = sad16bi_ia64; |
| 517 |
|
sad8 = sad8_ia64; |
| 518 |
|
dev16 = dev16_ia64; |
| 519 |
|
/* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
| 520 |
|
quant_intra = quant_intra_ia64; |
| 521 |
|
dequant_intra = dequant_intra_ia64; |
| 522 |
|
quant_inter = quant_inter_ia64; |
| 523 |
|
dequant_inter = dequant_inter_ia64; |
| 524 |
|
transfer_8to16copy = transfer_8to16copy_ia64; |
| 525 |
|
transfer_16to8copy = transfer_16to8copy_ia64; |
| 526 |
|
transfer_8to16sub = transfer_8to16sub_ia64; |
| 527 |
|
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
| 528 |
|
transfer_16to8add = transfer_16to8add_ia64; |
| 529 |
|
transfer8x8_copy = transfer8x8_copy_ia64; |
| 530 |
|
} |
| 531 |
|
#endif |
| 532 |
|
|
| 533 |
|
#if defined(ARCH_IS_PPC) |
| 534 |
|
if ((cpu_flags & XVID_CPU_ASM)) |
| 535 |
|
{ |
| 536 |
|
calc_cbp = calc_cbp_ppc; |
| 537 |
|
} |
| 538 |
|
|
| 539 |
|
if ((cpu_flags & XVID_CPU_ALTIVEC)) |
| 540 |
|
{ |
| 541 |
|
calc_cbp = calc_cbp_altivec; |
| 542 |
|
fdct = fdct_altivec; |
| 543 |
|
idct = idct_altivec; |
| 544 |
|
sadInit = sadInit_altivec; |
| 545 |
|
sad16 = sad16_altivec; |
| 546 |
|
sad8 = sad8_altivec; |
| 547 |
|
dev16 = dev16_altivec; |
| 548 |
|
} |
| 549 |
|
#endif |
| 550 |
|
|
| 551 |
|
#if defined(_DEBUG) |
| 552 |
|
xvid_debug = init->debug; |
| 553 |
|
#endif |
| 554 |
|
|
| 555 |
|
return 0; |
| 556 |
|
} |
| 557 |
|
|
| 558 |
|
|
| 559 |
|
static int |
| 560 |
|
xvid_gbl_info(xvid_gbl_info_t * info) |
| 561 |
|
{ |
| 562 |
|
if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */ |
| 563 |
|
return XVID_ERR_VERSION; |
| 564 |
|
|
| 565 |
|
info->actual_version = XVID_VERSION; |
| 566 |
|
info->build = "dev-api-4"; |
| 567 |
|
info->cpu_flags = detect_cpu_flags(); |
| 568 |
|
|
| 569 |
|
#if defined(_SMP) && defined(WIN32) |
| 570 |
|
info->num_threads = pthread_num_processors_np();; |
| 571 |
|
#else |
| 572 |
|
info->num_threads = 0; |
| 573 |
|
#endif |
| 574 |
|
|
| 575 |
|
return 0; |
| 576 |
|
} |
| 577 |
|
|
| 578 |
|
|
| 579 |
|
static int |
| 580 |
|
xvid_gbl_convert(xvid_gbl_convert_t* convert) |
| 581 |
|
{ |
| 582 |
|
int width; |
| 583 |
|
int height; |
| 584 |
|
int width2; |
| 585 |
|
int height2; |
| 586 |
|
IMAGE img; |
| 587 |
|
|
| 588 |
|
if (XVID_VERSION_MAJOR(convert->version) != 1) /* v1.x.x */ |
| 589 |
|
return XVID_ERR_VERSION; |
| 590 |
|
|
| 591 |
|
#if 0 |
| 592 |
|
const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
| 593 |
|
#endif |
| 594 |
|
width = convert->width; |
| 595 |
|
height = convert->height; |
| 596 |
|
width2 = convert->width/2; |
| 597 |
|
height2 = convert->height/2; |
| 598 |
|
|
| 599 |
|
switch (convert->input.csp & ~XVID_CSP_VFLIP) |
| 600 |
|
{ |
| 601 |
|
case XVID_CSP_YV12 : |
| 602 |
|
img.y = convert->input.plane[0]; |
| 603 |
|
img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; |
| 604 |
|
img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; |
| 605 |
|
image_output(&img, width, height, width, |
| 606 |
|
(uint8_t**)convert->output.plane, convert->output.stride, |
| 607 |
|
convert->output.csp, convert->interlacing); |
| 608 |
|
break; |
| 609 |
|
|
| 610 |
|
default : |
| 611 |
|
return XVID_ERR_FORMAT; |
| 612 |
|
} |
| 613 |
|
|
| 614 |
|
|
| 615 |
|
emms(); |
| 616 |
|
return 0; |
| 617 |
|
} |
| 618 |
|
|
| 619 |
|
|
| 620 |
|
|
| 621 |
|
void fill8(uint8_t * block, int size, int value) |
| 622 |
|
{ |
| 623 |
|
int i; |
| 624 |
|
for (i = 0; i < size; i++) |
| 625 |
|
block[i] = value; |
| 626 |
|
} |
| 627 |
|
|
| 628 |
|
void fill16(int16_t * block, int size, int value) |
| 629 |
|
{ |
| 630 |
|
int i; |
| 631 |
|
for (i = 0; i < size; i++) |
| 632 |
|
block[i] = value; |
| 633 |
|
} |
| 634 |
|
|
| 635 |
|
#define RANDOM(min,max) min + (rand() % (max-min)) |
| 636 |
|
|
| 637 |
|
void random8(uint8_t * block, int size, int min, int max) |
| 638 |
|
{ |
| 639 |
|
int i; |
| 640 |
|
for (i = 0; i < size; i++) |
| 641 |
|
block[i] = RANDOM(min,max); |
| 642 |
|
} |
| 643 |
|
|
| 644 |
|
void random16(int16_t * block, int size, int min, int max) |
| 645 |
|
{ |
| 646 |
|
int i; |
| 647 |
|
for (i = 0; i < size; i++) |
| 648 |
|
block[i] = RANDOM(min,max); |
| 649 |
|
} |
| 650 |
|
|
| 651 |
|
int compare16(const int16_t * blockA, const int16_t * blockB, int size) |
| 652 |
|
{ |
| 653 |
|
int i; |
| 654 |
|
for (i = 0; i < size; i++) |
| 655 |
|
if (blockA[i] != blockB[i]) |
| 656 |
|
return 1; |
| 657 |
|
|
| 658 |
|
return 0; |
| 659 |
|
} |
| 660 |
|
|
| 661 |
|
int diff16(const int16_t * blockA, const int16_t * blockB, int size) |
| 662 |
|
{ |
| 663 |
|
int i, diff = 0; |
| 664 |
|
for (i = 0; i < size; i++) |
| 665 |
|
diff += abs(blockA[i]-blockB[i]); |
| 666 |
|
return diff; |
| 667 |
|
} |
| 668 |
|
|
| 669 |
|
|
| 670 |
|
#define XVID_TEST_RANDOM 0x00000001 /* random input data */ |
| 671 |
|
#define XVID_TEST_VERBOSE 0x00000002 /* verbose error output */ |
| 672 |
|
|
| 673 |
|
|
| 674 |
|
#define TEST_FORWARD 0x00000001 /* intra */ |
| 675 |
|
#define TEST_FDCT (TEST_FORWARD) |
| 676 |
|
#define TEST_IDCT (0) |
| 677 |
|
|
| 678 |
|
static int test_transform(void * funcA, void * funcB, const char * nameB, |
| 679 |
|
int test, int flags) |
| 680 |
|
{ |
| 681 |
|
int i; |
| 682 |
|
int64_t timeSTART; |
| 683 |
|
int64_t timeA = 0; |
| 684 |
|
int64_t timeB = 0; |
| 685 |
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
| 686 |
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
| 687 |
|
int min, max; |
| 688 |
|
int count = 0; |
| 689 |
|
|
| 690 |
|
int tmp; |
| 691 |
|
int min_error = 0x10000*64; |
| 692 |
|
int max_error = 0; |
| 693 |
|
|
| 694 |
|
|
| 695 |
|
if ((test & TEST_FORWARD)) /* forward */ |
| 696 |
|
{ |
| 697 |
|
min = -256; |
| 698 |
|
max = 255; |
| 699 |
|
}else{ /* inverse */ |
| 700 |
|
min = -2048; |
| 701 |
|
max = 2047; |
| 702 |
|
} |
| 703 |
|
|
| 704 |
|
for (i = 0; i < 64*64; i++) |
| 705 |
|
{ |
| 706 |
|
if ((flags & XVID_TEST_RANDOM)) |
| 707 |
|
{ |
| 708 |
|
random16(arrayA, 64, min, max); |
| 709 |
|
}else{ |
| 710 |
|
fill16(arrayA, 64, i); |
| 711 |
|
} |
| 712 |
|
memcpy(arrayB, arrayA, 64*sizeof(int16_t)); |
| 713 |
|
|
| 714 |
|
if ((test & TEST_FORWARD)) |
| 715 |
|
{ |
| 716 |
|
timeSTART = read_counter(); |
| 717 |
|
((fdctFunc*)funcA)(arrayA); |
| 718 |
|
timeA += read_counter() - timeSTART; |
| 719 |
|
|
| 720 |
|
timeSTART = read_counter(); |
| 721 |
|
((fdctFunc*)funcB)(arrayB); |
| 722 |
|
timeB += read_counter() - timeSTART; |
| 723 |
|
} |
| 724 |
|
else |
| 725 |
|
{ |
| 726 |
|
timeSTART = read_counter(); |
| 727 |
|
((idctFunc*)funcA)(arrayA); |
| 728 |
|
timeA += read_counter() - timeSTART; |
| 729 |
|
|
| 730 |
|
timeSTART = read_counter(); |
| 731 |
|
((idctFunc*)funcB)(arrayB); |
| 732 |
|
timeB += read_counter() - timeSTART; |
| 733 |
|
} |
| 734 |
|
|
| 735 |
|
tmp = diff16(arrayA, arrayB, 64) / 64; |
| 736 |
|
if (tmp > max_error) |
| 737 |
|
max_error = tmp; |
| 738 |
|
if (tmp < min_error) |
| 739 |
|
min_error = tmp; |
| 740 |
|
|
| 741 |
|
count++; |
| 742 |
|
} |
| 743 |
|
|
| 744 |
|
/* print the "average difference" of best/worst transforms */ |
| 745 |
|
printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error); |
| 746 |
|
|
| 747 |
|
return 0; |
| 748 |
|
} |
| 749 |
|
|
| 750 |
|
|
| 751 |
|
#define TEST_QUANT 0x00000001 /* forward quantization */ |
| 752 |
|
#define TEST_INTRA 0x00000002 /* intra */ |
| 753 |
|
#define TEST_QUANT_INTRA (TEST_QUANT|TEST_INTRA) |
| 754 |
|
#define TEST_QUANT_INTER (TEST_QUANT) |
| 755 |
|
#define TEST_DEQUANT_INTRA (TEST_INTRA) |
| 756 |
|
#define TEST_DEQUANT_INTER (0) |
| 757 |
|
|
| 758 |
|
static int test_quant(void * funcA, void * funcB, const char * nameB, |
| 759 |
|
int test, int flags) |
| 760 |
|
{ |
| 761 |
|
int q,i; |
| 762 |
|
int64_t timeSTART; |
| 763 |
|
int64_t timeA = 0; |
| 764 |
|
int64_t timeB = 0; |
| 765 |
|
int retA = 0, retB = 0; |
| 766 |
|
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
| 767 |
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
| 768 |
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
| 769 |
|
int min, max; |
| 770 |
|
int count = 0; |
| 771 |
|
int errors = 0; |
| 772 |
|
|
| 773 |
|
if ((test & TEST_QUANT)) /* quant */ |
| 774 |
|
{ |
| 775 |
|
min = -2048; |
| 776 |
|
max = 2047; |
| 777 |
|
}else{ /* dequant */ |
| 778 |
|
min = -256; |
| 779 |
|
max = 255; |
| 780 |
|
} |
| 781 |
|
|
| 782 |
|
for (q = 1; q <= 31; q++) /* quantizer */ |
| 783 |
|
{ |
| 784 |
|
for (i = min; i < max; i++) /* input coeff */ |
| 785 |
|
{ |
| 786 |
|
if ((flags & XVID_TEST_RANDOM)) |
| 787 |
|
{ |
| 788 |
|
random16(arrayX, 64, min, max); |
| 789 |
|
}else{ |
| 790 |
|
fill16(arrayX, 64, i); |
| 791 |
|
} |
| 792 |
|
|
| 793 |
|
if ((test & TEST_INTRA)) /* intra */ |
| 794 |
|
{ |
| 795 |
|
timeSTART = read_counter(); |
| 796 |
|
((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q); |
| 797 |
|
timeA += read_counter() - timeSTART; |
| 798 |
|
|
| 799 |
|
timeSTART = read_counter(); |
| 800 |
|
((quanth263_intraFunc*)funcB)(arrayB, arrayX, q, q); |
| 801 |
|
timeB += read_counter() - timeSTART; |
| 802 |
|
} |
| 803 |
|
else /* inter */ |
| 804 |
|
{ |
| 805 |
|
timeSTART = read_counter(); |
| 806 |
|
retA = ((quanth263_interFunc*)funcA)(arrayA, arrayX, q); |
| 807 |
|
timeA += read_counter() - timeSTART; |
| 808 |
|
|
| 809 |
|
timeSTART = read_counter(); |
| 810 |
|
retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q); |
| 811 |
|
timeB += read_counter() - timeSTART; |
| 812 |
|
} |
| 813 |
|
|
| 814 |
|
/* compare return value from quant_inter, and compare (de)quantiz'd arrays */ |
| 815 |
|
if ( ((test&TEST_QUANT) && !(test&TEST_INTRA) && retA != retB ) || |
| 816 |
|
compare16(arrayA, arrayB, 64)) |
| 817 |
|
{ |
| 818 |
|
errors++; |
| 819 |
|
if ((flags & XVID_TEST_VERBOSE)) |
| 820 |
|
printf("%s error: q=%i, i=%i\n", nameB, q, i); |
| 821 |
} |
} |
| 822 |
|
|
| 823 |
|
count++; |
| 824 |
|
} |
| 825 |
|
} |
| 826 |
|
|
| 827 |
|
printf("%s:\t%i", nameB, (int)(timeB / count)); |
| 828 |
|
if (errors>0) |
| 829 |
|
printf("\t(%i errors out of %i)", errors, count); |
| 830 |
|
printf("\n"); |
| 831 |
|
|
| 832 |
|
return 0; |
| 833 |
|
} |
| 834 |
|
|
| 835 |
|
|
| 836 |
|
|
| 837 |
|
int xvid_init_test(int flags) |
| 838 |
|
{ |
| 839 |
|
#if defined(ARCH_IS_IA32) |
| 840 |
|
int cpu_flags; |
| 841 |
|
#endif |
| 842 |
|
|
| 843 |
|
printf("XviD tests\n\n"); |
| 844 |
|
|
| 845 |
|
#if defined(ARCH_IS_IA32) |
| 846 |
|
cpu_flags = detect_cpu_flags(); |
| 847 |
|
#endif |
| 848 |
|
|
| 849 |
|
idct_int32_init(); |
| 850 |
|
emms(); |
| 851 |
|
|
| 852 |
|
srand(time(0)); |
| 853 |
|
|
| 854 |
|
/* fDCT test */ |
| 855 |
|
printf("--- fdct ---\n"); |
| 856 |
|
test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags); |
| 857 |
|
|
| 858 |
|
#if defined(ARCH_IS_IA32) |
| 859 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 860 |
|
test_transform(fdct_int32, fdct_mmx, "mmx", TEST_FDCT, flags); |
| 861 |
|
if (cpu_flags & XVID_CPU_SSE2) |
| 862 |
|
test_transform(fdct_int32, fdct_sse2, "sse2", TEST_FDCT, flags); |
| 863 |
|
#endif |
| 864 |
|
|
| 865 |
|
/* iDCT test */ |
| 866 |
|
printf("\n--- idct ---\n"); |
| 867 |
|
test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags); |
| 868 |
|
|
| 869 |
|
#if defined(ARCH_IS_IA32) |
| 870 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 871 |
|
test_transform(idct_int32, idct_mmx, "mmx", TEST_IDCT, flags); |
| 872 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
| 873 |
|
test_transform(idct_int32, idct_xmm, "xmm", TEST_IDCT, flags); |
| 874 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 875 |
|
test_transform(idct_int32, idct_3dne, "3dne", TEST_IDCT, flags); |
| 876 |
|
if (cpu_flags & XVID_CPU_SSE2) |
| 877 |
|
test_transform(idct_int32, idct_sse2, "sse2", TEST_IDCT, flags); |
| 878 |
|
#endif |
| 879 |
|
|
| 880 |
|
/* Intra quantization test */ |
| 881 |
|
printf("\n--- quant intra ---\n"); |
| 882 |
|
test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags); |
| 883 |
|
|
| 884 |
|
#if defined(ARCH_IS_IA32) |
| 885 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 886 |
|
test_quant(quant_intra_c, quant_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
| 887 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 888 |
|
test_quant(quant_intra_c, quant_intra_3dne, "3dne", TEST_QUANT_INTRA, flags); |
| 889 |
|
if (cpu_flags & XVID_CPU_SSE2) |
| 890 |
|
test_quant(quant_intra_c, quant_intra_sse2, "sse2", TEST_QUANT_INTRA, flags); |
| 891 |
|
#endif |
| 892 |
|
|
| 893 |
|
/* Inter quantization test */ |
| 894 |
|
printf("\n--- quant inter ---\n"); |
| 895 |
|
test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags); |
| 896 |
|
|
| 897 |
|
#if defined(ARCH_IS_IA32) |
| 898 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 899 |
|
test_quant(quant_inter_c, quant_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
| 900 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 901 |
|
test_quant(quant_inter_c, quant_inter_3dne, "3dne", TEST_QUANT_INTER, flags); |
| 902 |
|
if (cpu_flags & XVID_CPU_SSE2) |
| 903 |
|
test_quant(quant_inter_c, quant_inter_sse2, "sse2", TEST_QUANT_INTER, flags); |
| 904 |
|
#endif |
| 905 |
|
|
| 906 |
|
/* Intra dequantization test */ |
| 907 |
|
printf("\n--- dequant intra ---\n"); |
| 908 |
|
test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
| 909 |
|
|
| 910 |
|
#if defined(ARCH_IS_IA32) |
| 911 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 912 |
|
test_quant(dequant_intra_c, dequant_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
| 913 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
| 914 |
|
test_quant(dequant_intra_c, dequant_intra_xmm, "xmm", TEST_DEQUANT_INTRA, flags); |
| 915 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 916 |
|
test_quant(dequant_intra_c, dequant_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
| 917 |
|
if (cpu_flags & XVID_CPU_SSE2) |
| 918 |
|
test_quant(dequant_intra_c, dequant_intra_sse2, "sse2", TEST_DEQUANT_INTRA, flags); |
| 919 |
|
#endif |
| 920 |
|
|
| 921 |
|
/* Inter dequantization test */ |
| 922 |
|
printf("\n--- dequant inter ---\n"); |
| 923 |
|
test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags); |
| 924 |
|
|
| 925 |
|
#if defined(ARCH_IS_IA32) |
| 926 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 927 |
|
test_quant(dequant_inter_c, dequant_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
| 928 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
| 929 |
|
test_quant(dequant_inter_c, dequant_inter_xmm, "xmm", TEST_DEQUANT_INTER, flags); |
| 930 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 931 |
|
test_quant(dequant_inter_c, dequant_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
| 932 |
|
if (cpu_flags & XVID_CPU_SSE2) |
| 933 |
|
test_quant(dequant_inter_c, dequant_inter_sse2, "sse2", TEST_DEQUANT_INTER, flags); |
| 934 |
|
#endif |
| 935 |
|
|
| 936 |
|
/* Intra quantization test */ |
| 937 |
|
printf("\n--- quant4 intra ---\n"); |
| 938 |
|
test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags); |
| 939 |
|
|
| 940 |
|
#if defined(ARCH_IS_IA32) |
| 941 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 942 |
|
test_quant(quant4_intra_c, quant4_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
| 943 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
| 944 |
|
test_quant(quant4_intra_c, quant4_intra_xmm, "xmm", TEST_QUANT_INTRA, flags); |
| 945 |
|
#endif |
| 946 |
|
|
| 947 |
|
/* Inter quantization test */ |
| 948 |
|
printf("\n--- quant4 inter ---\n"); |
| 949 |
|
test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags); |
| 950 |
|
|
| 951 |
|
#if defined(ARCH_IS_IA32) |
| 952 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 953 |
|
test_quant(quant4_inter_c, quant4_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
| 954 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
| 955 |
|
test_quant(quant4_inter_c, quant4_inter_xmm, "xmm", TEST_QUANT_INTER, flags); |
| 956 |
#endif |
#endif |
| 957 |
|
|
| 958 |
// API version |
/* Intra dequantization test */ |
| 959 |
init_param->api_version = API_VERSION; |
printf("\n--- dequant4 intra ---\n"); |
| 960 |
|
test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
| 961 |
|
|
| 962 |
|
#if defined(ARCH_IS_IA32) |
| 963 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 964 |
|
test_quant(dequant4_intra_c, dequant4_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
| 965 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 966 |
|
test_quant(dequant4_intra_c, dequant4_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
| 967 |
|
#endif |
| 968 |
|
|
| 969 |
|
/* Inter dequantization test */ |
| 970 |
|
printf("\n--- dequant4 inter ---\n"); |
| 971 |
|
test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags); |
| 972 |
|
|
| 973 |
|
#if defined(ARCH_IS_IA32) |
| 974 |
|
if (cpu_flags & XVID_CPU_MMX) |
| 975 |
|
test_quant(dequant4_inter_c, dequant4_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
| 976 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 977 |
|
test_quant(dequant4_inter_c, dequant4_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
| 978 |
|
#endif |
| 979 |
|
|
| 980 |
// something clever has to be done for this |
emms(); |
|
init_param->core_build = 1000; |
|
| 981 |
|
|
| 982 |
return XVID_ERR_OK; |
return 0; |
| 983 |
} |
} |
| 984 |
|
|
| 985 |
int xvid_decore(void * handle, int opt, void * param1, void * param2) |
|
| 986 |
|
/***************************************************************************** |
| 987 |
|
* XviD Global Entry point |
| 988 |
|
* |
| 989 |
|
* Well this function initialize all internal function pointers according |
| 990 |
|
* to the CPU features forced by the library client or autodetected (depending |
| 991 |
|
* on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all |
| 992 |
|
* image colorspace transformation tables. |
| 993 |
|
* |
| 994 |
|
****************************************************************************/ |
| 995 |
|
|
| 996 |
|
|
| 997 |
|
int |
| 998 |
|
xvid_global(void *handle, |
| 999 |
|
int opt, |
| 1000 |
|
void *param1, |
| 1001 |
|
void *param2) |
| 1002 |
{ |
{ |
| 1003 |
switch (opt) |
switch (opt) |
| 1004 |
{ |
{ |
| 1005 |
case XVID_DEC_DECODE : |
case XVID_GBL_INIT : |
| 1006 |
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1); |
return xvid_gbl_init((xvid_gbl_init_t*)param1); |
| 1007 |
|
|
| 1008 |
|
case XVID_GBL_INFO : |
| 1009 |
|
return xvid_gbl_info((xvid_gbl_info_t*)param1); |
| 1010 |
|
|
| 1011 |
|
case XVID_GBL_CONVERT : |
| 1012 |
|
return xvid_gbl_convert((xvid_gbl_convert_t*)param1); |
| 1013 |
|
|
| 1014 |
|
case XVID_GBL_TEST : |
| 1015 |
|
{ |
| 1016 |
|
ptr_t flags = (ptr_t)param1; |
| 1017 |
|
return xvid_init_test((int)flags); |
| 1018 |
|
} |
| 1019 |
|
default : |
| 1020 |
|
return XVID_ERR_FAIL; |
| 1021 |
|
} |
| 1022 |
|
} |
| 1023 |
|
|
| 1024 |
|
/***************************************************************************** |
| 1025 |
|
* XviD Native decoder entry point |
| 1026 |
|
* |
| 1027 |
|
* This function is just a wrapper to all the option cases. |
| 1028 |
|
* |
| 1029 |
|
* Returned values : XVID_ERR_FAIL when opt is invalid |
| 1030 |
|
* else returns the wrapped function result |
| 1031 |
|
* |
| 1032 |
|
****************************************************************************/ |
| 1033 |
|
|
| 1034 |
|
int |
| 1035 |
|
xvid_decore(void *handle, |
| 1036 |
|
int opt, |
| 1037 |
|
void *param1, |
| 1038 |
|
void *param2) |
| 1039 |
|
{ |
| 1040 |
|
switch (opt) { |
| 1041 |
case XVID_DEC_CREATE : |
case XVID_DEC_CREATE : |
| 1042 |
return decoder_create((XVID_DEC_PARAM *) param1); |
return decoder_create((xvid_dec_create_t *) param1); |
| 1043 |
|
|
| 1044 |
case XVID_DEC_DESTROY : |
case XVID_DEC_DESTROY : |
| 1045 |
return decoder_destroy((DECODER *) handle); |
return decoder_destroy((DECODER *) handle); |
| 1046 |
|
|
| 1047 |
|
case XVID_DEC_DECODE: |
| 1048 |
|
return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); |
| 1049 |
|
|
| 1050 |
default: |
default: |
| 1051 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
| 1052 |
} |
} |
| 1053 |
} |
} |
| 1054 |
|
|
| 1055 |
|
|
| 1056 |
int xvid_encore(void * handle, int opt, void * param1, void * param2) |
/***************************************************************************** |
| 1057 |
{ |
* XviD Native encoder entry point |
| 1058 |
switch (opt) |
* |
| 1059 |
|
* This function is just a wrapper to all the option cases. |
| 1060 |
|
* |
| 1061 |
|
* Returned values : XVID_ERR_FAIL when opt is invalid |
| 1062 |
|
* else returns the wrapped function result |
| 1063 |
|
* |
| 1064 |
|
****************************************************************************/ |
| 1065 |
|
|
| 1066 |
|
int |
| 1067 |
|
xvid_encore(void *handle, |
| 1068 |
|
int opt, |
| 1069 |
|
void *param1, |
| 1070 |
|
void *param2) |
| 1071 |
{ |
{ |
| 1072 |
|
switch (opt) { |
| 1073 |
case XVID_ENC_ENCODE : |
case XVID_ENC_ENCODE : |
| 1074 |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, (XVID_ENC_STATS *) param2); |
|
| 1075 |
|
return enc_encode((Encoder *) handle, |
| 1076 |
|
(xvid_enc_frame_t *) param1, |
| 1077 |
|
(xvid_enc_stats_t *) param2); |
| 1078 |
|
|
| 1079 |
case XVID_ENC_CREATE : |
case XVID_ENC_CREATE : |
| 1080 |
return encoder_create((XVID_ENC_PARAM *) param1); |
return enc_create((xvid_enc_create_t *) param1); |
| 1081 |
|
|
| 1082 |
case XVID_ENC_DESTROY : |
case XVID_ENC_DESTROY : |
| 1083 |
return encoder_destroy((Encoder *) handle); |
return enc_destroy((Encoder *) handle); |
| 1084 |
|
|
| 1085 |
default: |
default: |
| 1086 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
Parent Directory
| 
Revision Log
| 
Patch