3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Native API implementation - |
* - Native API implementation - |
5 |
* |
* |
6 |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
* Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
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 |
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
* $Id: xvid.c,v 1.48 2004-03-22 22:36:23 edgomez Exp $ |
* $Id: xvid.c,v 1.73 2006-12-06 19:55:07 Isibaar Exp $ |
23 |
* |
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
25 |
|
|
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" |
|
#include "image/reduced.h" |
|
39 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
40 |
#include "utils/mbfunctions.h" |
#include "utils/mbfunctions.h" |
41 |
#include "quant/quant.h" |
#include "quant/quant.h" |
42 |
#include "motion/motion.h" |
#include "motion/motion.h" |
43 |
|
#include "motion/gmc.h" |
44 |
#include "motion/sad.h" |
#include "motion/sad.h" |
45 |
#include "utils/emms.h" |
#include "utils/emms.h" |
46 |
#include "utils/timer.h" |
#include "utils/timer.h" |
52 |
unsigned int xvid_debug = 0; /* xvid debug mask */ |
unsigned int xvid_debug = 0; /* xvid debug mask */ |
53 |
#endif |
#endif |
54 |
|
|
55 |
#if defined(ARCH_IS_IA32) |
#if (defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)) && defined(_MSC_VER) |
|
#if defined(_MSC_VER) |
|
56 |
# include <windows.h> |
# include <windows.h> |
57 |
#else |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || defined(ARCH_IS_PPC) |
58 |
# include <signal.h> |
# include <signal.h> |
59 |
# include <setjmp.h> |
# include <setjmp.h> |
60 |
|
|
77 |
* 0 : SIGILL was *not* signalled |
* 0 : SIGILL was *not* signalled |
78 |
* 1 : SIGILL was signalled |
* 1 : SIGILL was signalled |
79 |
*/ |
*/ |
80 |
|
#if (defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)) && defined(_MSC_VER) |
81 |
int |
static int |
82 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
83 |
{ |
{ |
|
#if defined(_MSC_VER) |
|
84 |
_try { |
_try { |
85 |
func(); |
func(); |
86 |
} |
} _except(EXCEPTION_EXECUTE_HANDLER) { |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
|
87 |
|
|
88 |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
89 |
return 1; |
return(1); |
90 |
} |
} |
91 |
return 0; |
return(0); |
92 |
#else |
} |
93 |
|
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || defined(ARCH_IS_PPC) |
94 |
|
static int |
95 |
|
sigill_check(void (*func)()) |
96 |
|
{ |
97 |
void * old_handler; |
void * old_handler; |
98 |
int jmpret; |
int jmpret; |
99 |
|
|
100 |
|
/* Set our SIGILL handler */ |
101 |
old_handler = signal(SIGILL, sigill_handler); |
old_handler = signal(SIGILL, sigill_handler); |
102 |
if (old_handler == SIG_ERR) |
|
103 |
{ |
/* Check for error */ |
104 |
return -1; |
if (old_handler == SIG_ERR) { |
105 |
|
return(-1); |
106 |
} |
} |
107 |
|
|
108 |
|
/* Save stack context, so if func triggers a SIGILL, we can still roll |
109 |
|
* back to a valid CPU state */ |
110 |
jmpret = setjmp(mark); |
jmpret = setjmp(mark); |
111 |
if (jmpret == 0) |
|
112 |
{ |
/* If setjmp returned directly, then its returned value is 0, and we still |
113 |
|
* have to test the passed func. Otherwise it means the stack context has |
114 |
|
* been restored by a longjmp() call, which in our case happens only in the |
115 |
|
* signal handler */ |
116 |
|
if (jmpret == 0) { |
117 |
func(); |
func(); |
118 |
} |
} |
119 |
|
|
120 |
|
/* Restore old signal handler */ |
121 |
signal(SIGILL, old_handler); |
signal(SIGILL, old_handler); |
122 |
|
|
123 |
return jmpret; |
return(jmpret); |
|
#endif |
|
124 |
} |
} |
125 |
#endif |
#endif |
126 |
|
|
127 |
|
|
128 |
/* detect cpu flags */ |
/* detect cpu flags */ |
129 |
static unsigned int |
static unsigned int |
130 |
detect_cpu_flags() |
detect_cpu_flags(void) |
131 |
{ |
{ |
132 |
/* enable native assembly optimizations by default */ |
/* enable native assembly optimizations by default */ |
133 |
unsigned int cpu_flags = XVID_CPU_ASM; |
unsigned int cpu_flags = XVID_CPU_ASM; |
134 |
|
|
135 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
136 |
cpu_flags |= check_cpu_features(); |
cpu_flags |= check_cpu_features(); |
137 |
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
138 |
cpu_flags &= ~XVID_CPU_SSE; |
cpu_flags &= ~XVID_CPU_SSE; |
139 |
|
|
140 |
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
if ((cpu_flags & (XVID_CPU_SSE2|XVID_CPU_SSE3)) && sigill_check(sse2_os_trigger)) |
141 |
cpu_flags &= ~XVID_CPU_SSE2; |
cpu_flags &= ~(XVID_CPU_SSE2|XVID_CPU_SSE3); |
142 |
#endif |
#endif |
143 |
|
|
144 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
145 |
#if defined(ARCH_IS_PPC_ALTIVEC) |
if (!sigill_check(altivec_trigger)) |
146 |
cpu_flags |= XVID_CPU_ALTIVEC; |
cpu_flags |= XVID_CPU_ALTIVEC; |
147 |
#endif |
#endif |
|
#endif |
|
148 |
|
|
149 |
return cpu_flags; |
return cpu_flags; |
150 |
} |
} |
184 |
idct = idct_int32; |
idct = idct_int32; |
185 |
|
|
186 |
/* Only needed on PPC Altivec archs */ |
/* Only needed on PPC Altivec archs */ |
187 |
sadInit = 0; |
sadInit = NULL; |
188 |
|
|
189 |
/* Restore FPU context : emms_c is a nop functions */ |
/* Restore FPU context : emms_c is a nop functions */ |
190 |
emms = emms_c; |
emms = emms_c; |
211 |
transfer_8to16sub = transfer_8to16sub_c; |
transfer_8to16sub = transfer_8to16sub_c; |
212 |
transfer_8to16subro = transfer_8to16subro_c; |
transfer_8to16subro = transfer_8to16subro_c; |
213 |
transfer_8to16sub2 = transfer_8to16sub2_c; |
transfer_8to16sub2 = transfer_8to16sub2_c; |
214 |
|
transfer_8to16sub2ro = transfer_8to16sub2ro_c; |
215 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
216 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
217 |
|
transfer8x4_copy = transfer8x4_copy_c; |
218 |
|
|
219 |
/* Interlacing functions */ |
/* Interlacing functions */ |
220 |
MBFieldTest = MBFieldTest_c; |
MBFieldTest = MBFieldTest_c; |
224 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
225 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
226 |
|
|
227 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_c; |
228 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_c; |
229 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_c; |
230 |
|
|
231 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_c; |
232 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_c; |
233 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_c; |
234 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_c; |
235 |
|
|
236 |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
237 |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
238 |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
247 |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
248 |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
249 |
|
|
250 |
/* reduced resolution */ |
/* postprocessing */ |
251 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
image_brightness = image_brightness_c; |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
|
|
vfilter_31 = xvid_VFilter_31_C; |
|
|
hfilter_31 = xvid_HFilter_31_C; |
|
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
|
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
|
252 |
|
|
253 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
254 |
colorspace_init(); |
colorspace_init(); |
257 |
yv12_to_yv12 = yv12_to_yv12_c; |
yv12_to_yv12 = yv12_to_yv12_c; |
258 |
rgb555_to_yv12 = rgb555_to_yv12_c; |
rgb555_to_yv12 = rgb555_to_yv12_c; |
259 |
rgb565_to_yv12 = rgb565_to_yv12_c; |
rgb565_to_yv12 = rgb565_to_yv12_c; |
260 |
|
rgb_to_yv12 = rgb_to_yv12_c; |
261 |
bgr_to_yv12 = bgr_to_yv12_c; |
bgr_to_yv12 = bgr_to_yv12_c; |
262 |
bgra_to_yv12 = bgra_to_yv12_c; |
bgra_to_yv12 = bgra_to_yv12_c; |
263 |
abgr_to_yv12 = abgr_to_yv12_c; |
abgr_to_yv12 = abgr_to_yv12_c; |
279 |
/* All colorspace transformation functions YV12->User format */ |
/* All colorspace transformation functions YV12->User format */ |
280 |
yv12_to_rgb555 = yv12_to_rgb555_c; |
yv12_to_rgb555 = yv12_to_rgb555_c; |
281 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
282 |
|
yv12_to_rgb = yv12_to_rgb_c; |
283 |
yv12_to_bgr = yv12_to_bgr_c; |
yv12_to_bgr = yv12_to_bgr_c; |
284 |
yv12_to_bgra = yv12_to_bgra_c; |
yv12_to_bgra = yv12_to_bgra_c; |
285 |
yv12_to_abgr = yv12_to_abgr_c; |
yv12_to_abgr = yv12_to_abgr_c; |
307 |
dev16 = dev16_c; |
dev16 = dev16_c; |
308 |
sad16v = sad16v_c; |
sad16v = sad16v_c; |
309 |
sse8_16bit = sse8_16bit_c; |
sse8_16bit = sse8_16bit_c; |
310 |
|
sse8_8bit = sse8_8bit_c; |
311 |
|
|
312 |
#if defined(ARCH_IS_IA32) |
init_GMC(cpu_flags); |
313 |
|
|
314 |
if ((cpu_flags & XVID_CPU_ASM)) { |
#if defined(ARCH_IS_IA32) |
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
} |
|
315 |
|
|
316 |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
317 |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
350 |
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
351 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
352 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
353 |
|
transfer8x4_copy = transfer8x4_copy_mmx; |
354 |
|
|
355 |
/* Interlacing Functions */ |
/* Interlacing Functions */ |
356 |
MBFieldTest = MBFieldTest_mmx; |
MBFieldTest = MBFieldTest_mmx; |
360 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
361 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
362 |
|
|
363 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_mmx; |
364 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_mmx; |
365 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_mmx; |
366 |
|
|
367 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_mmx; |
368 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_mmx; |
369 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_mmx; |
370 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_mmx; |
371 |
|
|
372 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
373 |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
374 |
|
|
375 |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
376 |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
377 |
|
|
378 |
/* reduced resolution */ |
/* postprocessing */ |
379 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; |
image_brightness = image_brightness_mmx; |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; |
|
|
hfilter_31 = xvid_HFilter_31_mmx; |
|
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; |
|
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; |
|
380 |
|
|
381 |
/* image input xxx_to_yv12 related functions */ |
/* image input xxx_to_yv12 related functions */ |
382 |
yv12_to_yv12 = yv12_to_yv12_mmx; |
yv12_to_yv12 = yv12_to_yv12_mmx; |
383 |
bgr_to_yv12 = bgr_to_yv12_mmx; |
bgr_to_yv12 = bgr_to_yv12_mmx; |
384 |
|
rgb_to_yv12 = rgb_to_yv12_mmx; |
385 |
bgra_to_yv12 = bgra_to_yv12_mmx; |
bgra_to_yv12 = bgra_to_yv12_mmx; |
386 |
|
rgba_to_yv12 = rgba_to_yv12_mmx; |
387 |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
388 |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
389 |
|
|
405 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
406 |
sad16v = sad16v_mmx; |
sad16v = sad16v_mmx; |
407 |
sse8_16bit = sse8_16bit_mmx; |
sse8_16bit = sse8_16bit_mmx; |
408 |
|
sse8_8bit = sse8_8bit_mmx; |
409 |
} |
} |
410 |
|
|
411 |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
433 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
434 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
435 |
|
|
436 |
/* reduced resolution */ |
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_xmm; |
437 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_xmm; |
438 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_xmm; |
439 |
|
|
440 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_xmm; |
441 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm; |
442 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm; |
443 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm; |
444 |
|
|
445 |
/* Quantization */ |
/* Quantization */ |
|
quant_mpeg_intra = quant_mpeg_intra_xmm; |
|
446 |
quant_mpeg_inter = quant_mpeg_inter_xmm; |
quant_mpeg_inter = quant_mpeg_inter_xmm; |
447 |
|
|
448 |
dequant_h263_intra = dequant_h263_intra_xmm; |
dequant_h263_intra = dequant_h263_intra_xmm; |
450 |
|
|
451 |
/* Buffer transfer */ |
/* Buffer transfer */ |
452 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
453 |
|
transfer_8to16sub2ro = transfer_8to16sub2ro_xmm; |
454 |
|
|
455 |
/* Colorspace transformation */ |
/* Colorspace transformation */ |
456 |
yv12_to_yv12 = yv12_to_yv12_xmm; |
yv12_to_yv12 = yv12_to_yv12_xmm; |
472 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
473 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; |
474 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
475 |
|
|
476 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dn; |
477 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dn; |
478 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dn; |
479 |
} |
} |
480 |
|
|
481 |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
482 |
|
|
|
/* Inverse DCT */ |
|
|
idct = idct_3dne; |
|
|
|
|
483 |
/* Buffer transfer */ |
/* Buffer transfer */ |
484 |
transfer_8to16copy = transfer_8to16copy_3dne; |
transfer_8to16copy = transfer_8to16copy_3dne; |
485 |
transfer_16to8copy = transfer_16to8copy_3dne; |
transfer_16to8copy = transfer_16to8copy_3dne; |
486 |
transfer_8to16sub = transfer_8to16sub_3dne; |
transfer_8to16sub = transfer_8to16sub_3dne; |
487 |
transfer_8to16subro = transfer_8to16subro_3dne; |
transfer_8to16subro = transfer_8to16subro_3dne; |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
|
488 |
transfer_16to8add = transfer_16to8add_3dne; |
transfer_16to8add = transfer_16to8add_3dne; |
489 |
transfer8x8_copy = transfer8x8_copy_3dne; |
transfer8x8_copy = transfer8x8_copy_3dne; |
490 |
|
transfer8x4_copy = transfer8x4_copy_3dne; |
491 |
|
|
492 |
|
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
493 |
|
/* Inverse DCT */ |
494 |
|
idct = idct_3dne; |
495 |
|
|
496 |
|
/* Buffer transfer */ |
497 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
498 |
|
|
499 |
|
/* Interpolation */ |
500 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
501 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
502 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
503 |
|
|
504 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dne; |
505 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dne; |
506 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dne; |
507 |
|
|
508 |
/* Quantization */ |
/* Quantization */ |
509 |
quant_h263_intra = quant_h263_intra_3dne; |
quant_h263_intra = quant_h263_intra_3dne; /* cmov only */ |
510 |
quant_h263_inter = quant_h263_inter_3dne; |
quant_h263_inter = quant_h263_inter_3dne; |
511 |
dequant_mpeg_intra = dequant_mpeg_intra_3dne; |
dequant_mpeg_intra = dequant_mpeg_intra_3dne; /* cmov only */ |
512 |
dequant_mpeg_inter = dequant_mpeg_inter_3dne; |
dequant_mpeg_inter = dequant_mpeg_inter_3dne; |
513 |
dequant_h263_intra = dequant_h263_intra_3dne; |
dequant_h263_intra = dequant_h263_intra_3dne; |
514 |
dequant_h263_inter = dequant_h263_inter_3dne; |
dequant_h263_inter = dequant_h263_inter_3dne; |
515 |
|
|
516 |
/* ME functions */ |
/* ME functions */ |
517 |
calc_cbp = calc_cbp_3dne; |
calc_cbp = calc_cbp_3dne; |
518 |
|
|
519 |
sad16 = sad16_3dne; |
sad16 = sad16_3dne; |
520 |
sad8 = sad8_3dne; |
sad8 = sad8_3dne; |
521 |
sad16bi = sad16bi_3dne; |
sad16bi = sad16bi_3dne; |
522 |
sad8bi = sad8bi_3dne; |
sad8bi = sad8bi_3dne; |
523 |
dev16 = dev16_3dne; |
dev16 = dev16_3dne; |
524 |
|
} |
|
/* Interpolation */ |
|
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
|
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
|
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
|
525 |
} |
} |
526 |
|
|
527 |
if ((cpu_flags & XVID_CPU_SSE2)) { |
if ((cpu_flags & XVID_CPU_SSE2)) { |
538 |
sad16 = sad16_sse2; |
sad16 = sad16_sse2; |
539 |
dev16 = dev16_sse2; |
dev16 = dev16_sse2; |
540 |
|
|
541 |
/* DCT operators |
/* DCT operators */ |
|
* no iDCT because it's not "Walken matching" */ |
|
542 |
fdct = fdct_sse2_skal; |
fdct = fdct_sse2_skal; |
543 |
|
/* idct = idct_sse2_skal; */ /* Is now IEEE1180 and Walken compliant. Disabled until fully tested. */ |
544 |
|
|
545 |
|
/* postprocessing */ |
546 |
|
image_brightness = image_brightness_sse2; |
547 |
} |
} |
548 |
|
|
549 |
|
#if 0 // TODO: test... |
550 |
|
if ((cpu_flags & XVID_CPU_SSE3)) { |
551 |
|
|
552 |
|
/* SAD operators */ |
553 |
|
sad16 = sad16_sse3; |
554 |
|
dev16 = dev16_sse3; |
555 |
|
} |
556 |
|
#endif |
557 |
#endif /* ARCH_IS_IA32 */ |
#endif /* ARCH_IS_IA32 */ |
558 |
|
|
559 |
#if defined(ARCH_IS_IA64) |
#if defined(ARCH_IS_IA64) |
583 |
#endif |
#endif |
584 |
|
|
585 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
586 |
if ((cpu_flags & XVID_CPU_ASM)) |
if ((cpu_flags & XVID_CPU_ALTIVEC)) { |
587 |
{ |
/* sad operators */ |
588 |
calc_cbp = calc_cbp_ppc; |
sad16 = sad16_altivec_c; |
589 |
|
sad16bi = sad16bi_altivec_c; |
590 |
|
sad8 = sad8_altivec_c; |
591 |
|
dev16 = dev16_altivec_c; |
592 |
|
|
593 |
|
sse8_16bit = sse8_16bit_altivec_c; |
594 |
|
|
595 |
|
/* mem transfer */ |
596 |
|
transfer_8to16copy = transfer_8to16copy_altivec_c; |
597 |
|
transfer_16to8copy = transfer_16to8copy_altivec_c; |
598 |
|
transfer_8to16sub = transfer_8to16sub_altivec_c; |
599 |
|
transfer_8to16subro = transfer_8to16subro_altivec_c; |
600 |
|
transfer_8to16sub2 = transfer_8to16sub2_altivec_c; |
601 |
|
transfer_16to8add = transfer_16to8add_altivec_c; |
602 |
|
transfer8x8_copy = transfer8x8_copy_altivec_c; |
603 |
|
|
604 |
|
/* Inverse DCT */ |
605 |
|
idct = idct_altivec_c; |
606 |
|
|
607 |
|
/* Interpolation */ |
608 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c; |
609 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c; |
610 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c; |
611 |
|
|
612 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; |
613 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; |
614 |
|
|
615 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c; |
616 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c; |
617 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c; |
618 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c; |
619 |
|
|
620 |
|
/* Colorspace conversion */ |
621 |
|
bgra_to_yv12 = bgra_to_yv12_altivec_c; |
622 |
|
abgr_to_yv12 = abgr_to_yv12_altivec_c; |
623 |
|
rgba_to_yv12 = rgba_to_yv12_altivec_c; |
624 |
|
argb_to_yv12 = argb_to_yv12_altivec_c; |
625 |
|
|
626 |
|
yuyv_to_yv12 = yuyv_to_yv12_altivec_c; |
627 |
|
uyvy_to_yv12 = uyvy_to_yv12_altivec_c; |
628 |
|
|
629 |
|
yv12_to_yuyv = yv12_to_yuyv_altivec_c; |
630 |
|
yv12_to_uyvy = yv12_to_uyvy_altivec_c; |
631 |
|
|
632 |
|
/* Quantization */ |
633 |
|
quant_h263_intra = quant_h263_intra_altivec_c; |
634 |
|
quant_h263_inter = quant_h263_inter_altivec_c; |
635 |
|
dequant_h263_intra = dequant_h263_intra_altivec_c; |
636 |
|
dequant_h263_inter = dequant_h263_inter_altivec_c; |
637 |
|
|
638 |
|
dequant_mpeg_intra = dequant_mpeg_intra_altivec_c; |
639 |
|
dequant_mpeg_inter = dequant_mpeg_inter_altivec_c; |
640 |
|
|
641 |
|
/* Qpel stuff */ |
642 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C; |
643 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C; |
644 |
} |
} |
645 |
|
#endif |
646 |
|
|
647 |
if ((cpu_flags & XVID_CPU_ALTIVEC)) |
#if defined(ARCH_IS_X86_64) |
648 |
{ |
/* For now, only XVID_CPU_ASM is looked for, so user can still |
649 |
calc_cbp = calc_cbp_altivec; |
* disable asm usage the usual way. When Intel EMT64 cpus will |
650 |
fdct = fdct_altivec; |
* be out, maybe we'll have to check more precisely what cpu |
651 |
idct = idct_altivec; |
* features there really are. */ |
652 |
sadInit = sadInit_altivec; |
if (cpu_flags & XVID_CPU_ASM) { |
653 |
sad16 = sad16_altivec; |
/* SIMD state flusher */ |
654 |
sad8 = sad8_altivec; |
emms = emms_mmx; |
655 |
dev16 = dev16_altivec; |
|
656 |
|
/* DCT operators */ |
657 |
|
fdct = fdct_skal_x86_64; |
658 |
|
idct = idct_x86_64; |
659 |
|
|
660 |
|
/* SAD operators */ |
661 |
|
sad16 = sad16_x86_64; |
662 |
|
sad8 = sad8_x86_64; |
663 |
|
sad16bi = sad16bi_x86_64; |
664 |
|
sad8bi = sad8bi_x86_64; |
665 |
|
dev16 = dev16_x86_64; |
666 |
|
sad16v = sad16v_x86_64; |
667 |
|
sse8_16bit = sse8_16bit_x86_64; |
668 |
|
sse8_8bit = sse8_8bit_x86_64; |
669 |
|
|
670 |
|
/* Interpolation operators */ |
671 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_x86_64; |
672 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_x86_64; |
673 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_x86_64; |
674 |
|
|
675 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_x86_64; |
676 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_x86_64; |
677 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_x86_64; |
678 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_x86_64; |
679 |
|
|
680 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_x86_64; |
681 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_x86_64; |
682 |
|
|
683 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_x86_64; |
684 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_x86_64; |
685 |
|
|
686 |
|
/* Quantization related functions */ |
687 |
|
quant_h263_intra = quant_h263_intra_x86_64; |
688 |
|
quant_h263_inter = quant_h263_inter_x86_64; |
689 |
|
dequant_h263_intra = dequant_h263_intra_x86_64; |
690 |
|
dequant_h263_inter = dequant_h263_inter_x86_64; |
691 |
|
/*quant_mpeg_intra = quant_mpeg_intra_x86_64; fix me! */ |
692 |
|
quant_mpeg_inter = quant_mpeg_inter_x86_64; |
693 |
|
dequant_mpeg_intra = dequant_mpeg_intra_x86_64; |
694 |
|
dequant_mpeg_inter = dequant_mpeg_inter_x86_64; |
695 |
|
|
696 |
|
/* Block related functions */ |
697 |
|
transfer_8to16copy = transfer_8to16copy_x86_64; |
698 |
|
transfer_16to8copy = transfer_16to8copy_x86_64; |
699 |
|
transfer_8to16sub = transfer_8to16sub_x86_64; |
700 |
|
transfer_8to16subro = transfer_8to16subro_x86_64; |
701 |
|
transfer_8to16sub2 = transfer_8to16sub2_x86_64; |
702 |
|
transfer_8to16sub2ro= transfer_8to16sub2ro_x86_64; |
703 |
|
transfer_16to8add = transfer_16to8add_x86_64; |
704 |
|
transfer8x8_copy = transfer8x8_copy_x86_64; |
705 |
|
|
706 |
|
/* Qpel stuff */ |
707 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_x86_64; |
708 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_x86_64; |
709 |
|
|
710 |
|
/* Interlacing Functions */ |
711 |
|
MBFieldTest = MBFieldTest_x86_64; |
712 |
} |
} |
713 |
#endif |
#endif |
714 |
|
|
716 |
xvid_debug = init->debug; |
xvid_debug = init->debug; |
717 |
#endif |
#endif |
718 |
|
|
719 |
return 0; |
return(0); |
720 |
} |
} |
721 |
|
|
722 |
|
|
727 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
728 |
|
|
729 |
info->actual_version = XVID_VERSION; |
info->actual_version = XVID_VERSION; |
730 |
info->build = "xvid-1.0.0"; |
info->build = "xvid-1.2.0-dev"; |
731 |
info->cpu_flags = detect_cpu_flags(); |
info->cpu_flags = detect_cpu_flags(); |
|
|
|
|
#if defined(_SMP) && defined(WIN32) |
|
|
info->num_threads = pthread_num_processors_np();; |
|
|
#else |
|
732 |
info->num_threads = 0; |
info->num_threads = 0; |
733 |
|
|
734 |
|
#if defined(WIN32) |
735 |
|
{ |
736 |
|
DWORD dwProcessAffinityMask, dwSystemAffinityMask; |
737 |
|
if (GetProcessAffinityMask(GetCurrentProcess(), &dwProcessAffinityMask, &dwSystemAffinityMask)) { |
738 |
|
int i; |
739 |
|
for(i=0; i<32; i++) { |
740 |
|
if ((dwProcessAffinityMask & (1<<i))) |
741 |
|
info->num_threads++; |
742 |
|
} |
743 |
|
} |
744 |
|
} |
745 |
#endif |
#endif |
746 |
|
|
747 |
return 0; |
return 0; |