3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Native API implementation - |
* - 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 |
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 |
* |
* |
|
****************************************************************************/ |
|
|
|
|
|
/***************************************************************************** |
|
|
* |
|
|
* History |
|
|
* |
|
|
* - 23.06.2002 added XVID_CPU_CHKONLY |
|
|
* - 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> |
|
|
* |
|
22 |
* $Id$ |
* $Id$ |
23 |
* |
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
47 |
#include "utils/timer.h" |
#include "utils/timer.h" |
48 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
49 |
|
|
50 |
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
#if defined(_DEBUG) |
51 |
|
unsigned int xvid_debug = 0; /* xvid debug mask */ |
52 |
|
#endif |
53 |
|
|
54 |
|
#if defined(ARCH_IS_IA32) |
55 |
|
|
56 |
#ifdef WIN32 |
#if defined(_MSC_VER) |
57 |
#include <windows.h> |
#include <windows.h> |
58 |
#else |
#else |
59 |
#include <signal.h> |
#include <signal.h> |
60 |
#include <setjmp.h> |
#include <setjmp.h> |
|
#endif |
|
|
|
|
|
|
|
|
#ifndef WIN32 |
|
61 |
|
|
62 |
static jmp_buf mark; |
static jmp_buf mark; |
63 |
|
|
70 |
|
|
71 |
|
|
72 |
/* |
/* |
73 |
calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
* Calls the funcptr, and returns whether SIGILL (illegal instruction) was |
74 |
return values: |
* signalled |
75 |
-1 : could not determine |
* |
76 |
0 : SIGILL was *not* signalled |
* Return values: |
77 |
1 : SIGILL was signalled |
* -1 : could not determine |
78 |
|
* 0 : SIGILL was *not* signalled |
79 |
|
* 1 : SIGILL was signalled |
80 |
*/ |
*/ |
81 |
|
|
82 |
int |
int |
83 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
84 |
{ |
{ |
85 |
#ifdef WIN32 |
#if defined(_MSC_VER) |
86 |
_try { |
_try { |
87 |
func(); |
func(); |
88 |
} |
} |
116 |
} |
} |
117 |
#endif |
#endif |
118 |
|
|
119 |
|
|
120 |
|
/* detect cpu flags */ |
121 |
|
static unsigned int |
122 |
|
detect_cpu_flags() |
123 |
|
{ |
124 |
|
/* enable native assembly optimizations by default */ |
125 |
|
unsigned int cpu_flags = XVID_CPU_ASM; |
126 |
|
|
127 |
|
#if defined(ARCH_IS_IA32) |
128 |
|
cpu_flags |= check_cpu_features(); |
129 |
|
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
130 |
|
cpu_flags &= ~XVID_CPU_SSE; |
131 |
|
|
132 |
|
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
133 |
|
cpu_flags &= ~XVID_CPU_SSE2; |
134 |
|
#endif |
135 |
|
|
136 |
|
#if defined(ARCH_IS_PPC) |
137 |
|
#if defined(ARCH_IS_PPC_ALTIVEC) |
138 |
|
cpu_flags |= XVID_CPU_ALTIVEC; |
139 |
|
#endif |
140 |
|
#endif |
141 |
|
|
142 |
|
return cpu_flags; |
143 |
|
} |
144 |
|
|
145 |
|
|
146 |
/***************************************************************************** |
/***************************************************************************** |
147 |
* XviD Init Entry point |
* XviD Init Entry point |
148 |
* |
* |
159 |
|
|
160 |
|
|
161 |
static |
static |
162 |
int xvid_init_init(XVID_INIT_PARAM * init_param) |
int xvid_gbl_init(xvid_gbl_init_t * init) |
163 |
{ |
{ |
164 |
int cpu_flags; |
unsigned int cpu_flags; |
|
|
|
|
/* Inform the client the API version */ |
|
|
init_param->api_version = API_VERSION; |
|
165 |
|
|
166 |
/* Inform the client the core build - unused because we're still alpha */ |
if (XVID_MAJOR(init->version) != 1) /* v1.x.x */ |
167 |
init_param->core_build = 1000; |
return XVID_ERR_VERSION; |
|
|
|
|
/* Do we have to force CPU features ? */ |
|
|
if ((init_param->cpu_flags & XVID_CPU_FORCE)) { |
|
|
|
|
|
cpu_flags = init_param->cpu_flags; |
|
|
|
|
|
} else { |
|
|
|
|
|
cpu_flags = check_cpu_features(); |
|
|
|
|
|
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
|
|
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
|
|
cpu_flags &= ~XVID_CPU_SSE; |
|
|
|
|
|
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
|
|
cpu_flags &= ~XVID_CPU_SSE2; |
|
|
#endif |
|
|
} |
|
|
|
|
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
|
|
{ |
|
|
init_param->cpu_flags = cpu_flags; |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
init_param->cpu_flags = cpu_flags; |
|
168 |
|
|
169 |
|
cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
170 |
|
|
171 |
/* Initialize the function pointers */ |
/* Initialize the function pointers */ |
172 |
idct_int32_init(); |
idct_int32_init(); |
197 |
transfer_8to16copy = transfer_8to16copy_c; |
transfer_8to16copy = transfer_8to16copy_c; |
198 |
transfer_16to8copy = transfer_16to8copy_c; |
transfer_16to8copy = transfer_16to8copy_c; |
199 |
transfer_8to16sub = transfer_8to16sub_c; |
transfer_8to16sub = transfer_8to16sub_c; |
200 |
|
transfer_8to16subro = transfer_8to16subro_c; |
201 |
transfer_8to16sub2 = transfer_8to16sub2_c; |
transfer_8to16sub2 = transfer_8to16sub2_c; |
202 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
203 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
225 |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
226 |
|
|
227 |
/* reduced resoltuion */ |
/* reduced resoltuion */ |
|
|
|
228 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
229 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
|
#ifdef ARCH_X86 |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
#else |
|
230 |
vfilter_31 = xvid_VFilter_31_C; |
vfilter_31 = xvid_VFilter_31_C; |
231 |
hfilter_31 = xvid_HFilter_31_C; |
hfilter_31 = xvid_HFilter_31_C; |
|
#endif |
|
232 |
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
233 |
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
234 |
|
|
284 |
dev16 = dev16_c; |
dev16 = dev16_c; |
285 |
sad16v = sad16v_c; |
sad16v = sad16v_c; |
286 |
|
|
287 |
// Halfpel8_Refine = Halfpel8_Refine_c; |
/* Halfpel8_Refine = Halfpel8_Refine_c; */ |
288 |
|
|
289 |
|
#if defined(ARCH_IS_IA32) |
290 |
|
|
291 |
#ifdef ARCH_X86 |
if ((cpu_flags & XVID_CPU_ASM)) { |
292 |
|
vfilter_31 = xvid_VFilter_31_x86; |
293 |
|
hfilter_31 = xvid_HFilter_31_x86; |
294 |
|
} |
295 |
|
|
296 |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
297 |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
301 |
emms = emms_mmx; |
emms = emms_mmx; |
302 |
} |
} |
303 |
|
|
304 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX)) { |
305 |
|
|
306 |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
307 |
fdct = fdct_mmx; |
fdct = fdct_mmx; |
308 |
idct = idct_mmx; |
idct = simple_idct_mmx; |
309 |
|
|
310 |
/* Quantization related functions */ |
/* Quantization related functions */ |
311 |
quant_intra = quant_intra_mmx; |
quant_intra = quant_intra_mmx; |
322 |
transfer_8to16copy = transfer_8to16copy_mmx; |
transfer_8to16copy = transfer_8to16copy_mmx; |
323 |
transfer_16to8copy = transfer_16to8copy_mmx; |
transfer_16to8copy = transfer_16to8copy_mmx; |
324 |
transfer_8to16sub = transfer_8to16sub_mmx; |
transfer_8to16sub = transfer_8to16sub_mmx; |
325 |
|
transfer_8to16subro = transfer_8to16subro_mmx; |
326 |
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
327 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
328 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
372 |
sad8bi = sad8bi_mmx; |
sad8bi = sad8bi_mmx; |
373 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
374 |
sad16v = sad16v_mmx; |
sad16v = sad16v_mmx; |
|
|
|
375 |
} |
} |
376 |
|
|
377 |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
378 |
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
if ((cpu_flags & XVID_CPU_3DNOW)) { |
379 |
|
|
380 |
|
emms = emms_3dn; |
381 |
|
|
382 |
/* ME functions */ |
/* ME functions */ |
383 |
sad16bi = sad16bi_3dn; |
sad16bi = sad16bi_3dn; |
388 |
} |
} |
389 |
|
|
390 |
|
|
391 |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
392 |
|
|
393 |
/* Inverse DCT */ |
/* Inverse DCT */ |
394 |
|
#if 0 /* We don't use Walken idct anymore! */ |
395 |
idct = idct_xmm; |
idct = idct_xmm; |
396 |
|
#endif |
397 |
|
|
398 |
/* Interpolation */ |
/* Interpolation */ |
399 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; |
428 |
sad16v = sad16v_xmm; |
sad16v = sad16v_xmm; |
429 |
} |
} |
430 |
|
|
431 |
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
if ((cpu_flags & XVID_CPU_3DNOW)) { |
432 |
|
|
433 |
/* Interpolation */ |
/* Interpolation */ |
434 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
436 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
437 |
} |
} |
438 |
|
|
439 |
if ((cpu_flags & XVID_CPU_3DNOWEXT) > 0) { |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
440 |
|
|
441 |
/* Inverse DCT */ |
/* Inverse DCT */ |
442 |
|
#if 0 /* We don't use Walken idct anymore! */ |
443 |
idct = idct_3dne; |
idct = idct_3dne; |
444 |
|
#endif |
445 |
|
|
446 |
/* Buffer transfer */ |
/* Buffer transfer */ |
447 |
transfer_8to16copy = transfer_8to16copy_3dne; |
transfer_8to16copy = transfer_8to16copy_3dne; |
448 |
transfer_16to8copy = transfer_16to8copy_3dne; |
transfer_16to8copy = transfer_16to8copy_3dne; |
449 |
transfer_8to16sub = transfer_8to16sub_3dne; |
transfer_8to16sub = transfer_8to16sub_3dne; |
450 |
|
transfer_8to16subro = transfer_8to16subro_3dne; |
451 |
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
452 |
transfer_16to8add = transfer_16to8add_3dne; |
transfer_16to8add = transfer_16to8add_3dne; |
453 |
transfer8x8_copy = transfer8x8_copy_3dne; |
transfer8x8_copy = transfer8x8_copy_3dne; |
475 |
} |
} |
476 |
|
|
477 |
|
|
478 |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
if ((cpu_flags & XVID_CPU_SSE2)) { |
|
#ifdef EXPERIMENTAL_SSE2_CODE |
|
479 |
|
|
480 |
calc_cbp = calc_cbp_sse2; |
calc_cbp = calc_cbp_sse2; |
481 |
|
|
485 |
quant_inter = quant_inter_sse2; |
quant_inter = quant_inter_sse2; |
486 |
dequant_inter = dequant_inter_sse2; |
dequant_inter = dequant_inter_sse2; |
487 |
|
|
488 |
/* ME */ |
#if defined(EXPERIMENTAL_SSE2_CODE) |
489 |
|
/* ME; slower than xmm */ |
490 |
sad16 = sad16_sse2; |
sad16 = sad16_sse2; |
491 |
dev16 = dev16_sse2; |
dev16 = dev16_sse2; |
492 |
|
#endif |
493 |
/* Forward and Inverse DCT */ |
/* Forward and Inverse DCT */ |
494 |
|
#if 0 /* Both function are known to be unprecise, better keep them deactivated */ |
495 |
idct = idct_sse2; |
idct = idct_sse2; |
496 |
fdct = fdct_sse2; |
fdct = fdct_sse2; |
497 |
#endif |
#endif |
498 |
} |
} |
|
|
|
499 |
#endif |
#endif |
500 |
|
|
501 |
#ifdef ARCH_IA64 |
#if defined(ARCH_IS_IA64) |
502 |
if ((cpu_flags & XVID_CPU_IA64) > 0) { //use assembler routines? |
if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
503 |
idct_ia64_init(); |
idct_ia64_init(); |
504 |
fdct = fdct_ia64; |
fdct = fdct_ia64; |
505 |
idct = idct_ia64; //not yet working, crashes |
idct = idct_ia64; /*not yet working, crashes */ |
506 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
507 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
508 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
510 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
511 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
512 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
513 |
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
/* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
514 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
515 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
516 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |
521 |
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
522 |
transfer_16to8add = transfer_16to8add_ia64; |
transfer_16to8add = transfer_16to8add_ia64; |
523 |
transfer8x8_copy = transfer8x8_copy_ia64; |
transfer8x8_copy = transfer8x8_copy_ia64; |
524 |
DEBUG("Using IA-64 assembler routines.\n"); |
DPRINTF(DPRINTF_DEBUG, "Using IA-64 assembler routines."); |
525 |
} |
} |
526 |
#endif |
#endif |
527 |
|
|
528 |
#ifdef ARCH_PPC |
#if defined(ARCH_IS_PPC) |
529 |
#ifdef ARCH_PPC_ALTIVEC |
if ((cpu_flags & XVID_CPU_ASM)) |
530 |
|
{ |
531 |
|
calc_cbp = calc_cbp_ppc; |
532 |
|
} |
533 |
|
|
534 |
|
if ((cpu_flags & XVID_CPU_ALTIVEC)) |
535 |
|
{ |
536 |
calc_cbp = calc_cbp_altivec; |
calc_cbp = calc_cbp_altivec; |
537 |
fdct = fdct_altivec; |
fdct = fdct_altivec; |
538 |
idct = idct_altivec; |
idct = idct_altivec; |
540 |
sad16 = sad16_altivec; |
sad16 = sad16_altivec; |
541 |
sad8 = sad8_altivec; |
sad8 = sad8_altivec; |
542 |
dev16 = dev16_altivec; |
dev16 = dev16_altivec; |
543 |
#else |
} |
|
calc_cbp = calc_cbp_ppc; |
|
544 |
#endif |
#endif |
545 |
|
|
546 |
|
#if defined(_DEBUG) |
547 |
|
xvid_debug = init->debug; |
548 |
#endif |
#endif |
549 |
|
|
550 |
return XVID_ERR_OK; |
return 0; |
551 |
} |
} |
552 |
|
|
553 |
|
|
554 |
|
static int |
555 |
|
xvid_gbl_info(xvid_gbl_info_t * info) |
556 |
|
{ |
557 |
|
if (XVID_MAJOR(info->version) != 1) /* v1.x.x */ |
558 |
|
return XVID_ERR_VERSION; |
559 |
|
|
560 |
|
info->actual_version = XVID_VERSION; |
561 |
|
info->build = "dev-api-4"; |
562 |
|
info->cpu_flags = detect_cpu_flags(); |
563 |
|
|
564 |
|
#if defined(_SMP) && defined(WIN32) |
565 |
|
info->num_threads = pthread_num_processors_np();; |
566 |
|
#else |
567 |
|
info->num_threads = 0; |
568 |
|
#endif |
569 |
|
|
570 |
|
return 0; |
571 |
|
} |
572 |
|
|
573 |
|
|
574 |
static int |
static int |
575 |
xvid_init_convert(XVID_INIT_CONVERTINFO* convert) |
xvid_gbl_convert(xvid_gbl_convert_t* convert) |
576 |
{ |
{ |
577 |
// const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
int width; |
578 |
const int width = convert->width; |
int height; |
579 |
const int height = convert->height; |
int width2; |
580 |
const int width2 = convert->width/2; |
int height2; |
|
const int height2 = convert->height/2; |
|
581 |
IMAGE img; |
IMAGE img; |
582 |
|
|
583 |
switch (convert->input.colorspace & ~XVID_CSP_VFLIP) |
if (XVID_MAJOR(convert->version) != 1) /* v1.x.x */ |
584 |
|
return XVID_ERR_VERSION; |
585 |
|
|
586 |
|
#if 0 |
587 |
|
const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
588 |
|
#endif |
589 |
|
width = convert->width; |
590 |
|
height = convert->height; |
591 |
|
width2 = convert->width/2; |
592 |
|
height2 = convert->height/2; |
593 |
|
|
594 |
|
switch (convert->input.csp & ~XVID_CSP_VFLIP) |
595 |
{ |
{ |
596 |
case XVID_CSP_YV12 : |
case XVID_CSP_YV12 : |
597 |
img.y = convert->input.y; |
img.y = convert->input.plane[0]; |
598 |
img.v = (uint8_t*)convert->input.y + width*height; |
img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; |
599 |
img.u = (uint8_t*)convert->input.y + width*height + width2*height2; |
img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; |
600 |
image_output(&img, width, height, width, |
image_output(&img, width, height, width, |
601 |
convert->output.y, convert->output.y_stride, |
(uint8_t**)convert->output.plane, convert->output.stride, |
602 |
convert->output.colorspace, convert->interlacing); |
convert->output.csp, convert->interlacing); |
603 |
break; |
break; |
604 |
|
|
605 |
default : |
default : |
608 |
|
|
609 |
|
|
610 |
emms(); |
emms(); |
611 |
return XVID_ERR_OK; |
return 0; |
612 |
} |
} |
613 |
|
|
614 |
|
|
657 |
{ |
{ |
658 |
int i, diff = 0; |
int i, diff = 0; |
659 |
for (i = 0; i < size; i++) |
for (i = 0; i < size; i++) |
660 |
diff += ABS(blockA[i]-blockB[i]); |
diff += abs(blockA[i]-blockB[i]); |
661 |
return diff; |
return diff; |
662 |
} |
} |
663 |
|
|
670 |
#define TEST_FDCT (TEST_FORWARD) |
#define TEST_FDCT (TEST_FORWARD) |
671 |
#define TEST_IDCT (0) |
#define TEST_IDCT (0) |
672 |
|
|
673 |
int test_transform(void * funcA, void * funcB, const char * nameB, |
static int test_transform(void * funcA, void * funcB, const char * nameB, |
674 |
int test, int flags) |
int test, int flags) |
675 |
{ |
{ |
676 |
int i; |
int i; |
750 |
#define TEST_DEQUANT_INTRA (TEST_INTRA) |
#define TEST_DEQUANT_INTRA (TEST_INTRA) |
751 |
#define TEST_DEQUANT_INTER (0) |
#define TEST_DEQUANT_INTER (0) |
752 |
|
|
753 |
int test_quant(void * funcA, void * funcB, const char * nameB, |
static int test_quant(void * funcA, void * funcB, const char * nameB, |
754 |
int test, int flags) |
int test, int flags) |
755 |
{ |
{ |
756 |
int q,i; |
int q,i; |
757 |
int64_t timeSTART; |
int64_t timeSTART; |
758 |
int64_t timeA = 0; |
int64_t timeA = 0; |
759 |
int64_t timeB = 0; |
int64_t timeB = 0; |
760 |
int retA, retB; |
int retA = 0, retB = 0; |
761 |
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
762 |
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
763 |
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
831 |
|
|
832 |
int xvid_init_test(int flags) |
int xvid_init_test(int flags) |
833 |
{ |
{ |
834 |
|
#if defined(ARCH_IS_IA32) |
835 |
int cpu_flags; |
int cpu_flags; |
836 |
|
#endif |
837 |
|
|
838 |
srand(time(0)); |
printf("XviD tests\n\n"); |
839 |
|
|
840 |
printf("xvid_init_test\n"); |
#if defined(ARCH_IS_IA32) |
841 |
|
cpu_flags = detect_cpu_flags(); |
842 |
|
#endif |
843 |
|
|
|
#if defined(ARCH_X86) |
|
|
cpu_flags = check_cpu_features(); |
|
844 |
idct_int32_init(); |
idct_int32_init(); |
845 |
emms_mmx(); |
emms(); |
846 |
|
|
847 |
|
srand(time(0)); |
848 |
|
|
849 |
|
/* fDCT test */ |
850 |
printf("--- fdct ---\n"); |
printf("--- fdct ---\n"); |
851 |
test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags); |
test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags); |
852 |
|
|
853 |
|
#if defined(ARCH_IS_IA32) |
854 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
855 |
test_transform(fdct_int32, fdct_mmx, "mmx", TEST_FDCT, flags); |
test_transform(fdct_int32, fdct_mmx, "mmx", TEST_FDCT, flags); |
856 |
if (cpu_flags & XVID_CPU_SSE2) |
if (cpu_flags & XVID_CPU_SSE2) |
857 |
test_transform(fdct_int32, fdct_sse2, "sse2", TEST_FDCT, flags); |
test_transform(fdct_int32, fdct_sse2, "sse2", TEST_FDCT, flags); |
858 |
|
#endif |
859 |
|
|
860 |
|
/* iDCT test */ |
861 |
printf("\n--- idct ---\n"); |
printf("\n--- idct ---\n"); |
862 |
test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags); |
test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags); |
863 |
|
|
864 |
|
#if defined(ARCH_IS_IA32) |
865 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
866 |
test_transform(idct_int32, idct_mmx, "mmx", TEST_IDCT, flags); |
test_transform(idct_int32, idct_mmx, "mmx", TEST_IDCT, flags); |
867 |
if (cpu_flags & XVID_CPU_MMXEXT) |
if (cpu_flags & XVID_CPU_MMXEXT) |
870 |
test_transform(idct_int32, idct_3dne, "3dne", TEST_IDCT, flags); |
test_transform(idct_int32, idct_3dne, "3dne", TEST_IDCT, flags); |
871 |
if (cpu_flags & XVID_CPU_SSE2) |
if (cpu_flags & XVID_CPU_SSE2) |
872 |
test_transform(idct_int32, idct_sse2, "sse2", TEST_IDCT, flags); |
test_transform(idct_int32, idct_sse2, "sse2", TEST_IDCT, flags); |
873 |
|
#endif |
874 |
|
|
875 |
|
/* Intra quantization test */ |
876 |
printf("\n--- quant intra ---\n"); |
printf("\n--- quant intra ---\n"); |
877 |
test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags); |
test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags); |
878 |
|
|
879 |
|
#if defined(ARCH_IS_IA32) |
880 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
881 |
test_quant(quant_intra_c, quant_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
test_quant(quant_intra_c, quant_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
882 |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
883 |
test_quant(quant_intra_c, quant_intra_3dne, "3dne", TEST_QUANT_INTRA, flags); |
test_quant(quant_intra_c, quant_intra_3dne, "3dne", TEST_QUANT_INTRA, flags); |
884 |
if (cpu_flags & XVID_CPU_SSE2) |
if (cpu_flags & XVID_CPU_SSE2) |
885 |
test_quant(quant_intra_c, quant_intra_sse2, "sse2", TEST_QUANT_INTRA, flags); |
test_quant(quant_intra_c, quant_intra_sse2, "sse2", TEST_QUANT_INTRA, flags); |
886 |
|
#endif |
887 |
|
|
888 |
|
/* Inter quantization test */ |
889 |
printf("\n--- quant inter ---\n"); |
printf("\n--- quant inter ---\n"); |
890 |
test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags); |
test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags); |
891 |
|
|
892 |
|
#if defined(ARCH_IS_IA32) |
893 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
894 |
test_quant(quant_inter_c, quant_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
test_quant(quant_inter_c, quant_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
895 |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
896 |
test_quant(quant_inter_c, quant_inter_3dne, "3dne", TEST_QUANT_INTER, flags); |
test_quant(quant_inter_c, quant_inter_3dne, "3dne", TEST_QUANT_INTER, flags); |
897 |
if (cpu_flags & XVID_CPU_SSE2) |
if (cpu_flags & XVID_CPU_SSE2) |
898 |
test_quant(quant_inter_c, quant_inter_sse2, "sse2", TEST_QUANT_INTER, flags); |
test_quant(quant_inter_c, quant_inter_sse2, "sse2", TEST_QUANT_INTER, flags); |
899 |
|
#endif |
900 |
|
|
901 |
|
/* Intra dequantization test */ |
902 |
printf("\n--- dequant intra ---\n"); |
printf("\n--- dequant intra ---\n"); |
903 |
test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
904 |
|
|
905 |
|
#if defined(ARCH_IS_IA32) |
906 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
907 |
test_quant(dequant_intra_c, dequant_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant_intra_c, dequant_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
908 |
if (cpu_flags & XVID_CPU_MMXEXT) |
if (cpu_flags & XVID_CPU_MMXEXT) |
911 |
test_quant(dequant_intra_c, dequant_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant_intra_c, dequant_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
912 |
if (cpu_flags & XVID_CPU_SSE2) |
if (cpu_flags & XVID_CPU_SSE2) |
913 |
test_quant(dequant_intra_c, dequant_intra_sse2, "sse2", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant_intra_c, dequant_intra_sse2, "sse2", TEST_DEQUANT_INTRA, flags); |
914 |
|
#endif |
915 |
|
|
916 |
|
/* Inter dequantization test */ |
917 |
printf("\n--- dequant inter ---\n"); |
printf("\n--- dequant inter ---\n"); |
918 |
test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags); |
test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags); |
919 |
|
|
920 |
|
#if defined(ARCH_IS_IA32) |
921 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
922 |
test_quant(dequant_inter_c, dequant_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
test_quant(dequant_inter_c, dequant_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
923 |
if (cpu_flags & XVID_CPU_MMXEXT) |
if (cpu_flags & XVID_CPU_MMXEXT) |
926 |
test_quant(dequant_inter_c, dequant_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
test_quant(dequant_inter_c, dequant_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
927 |
if (cpu_flags & XVID_CPU_SSE2) |
if (cpu_flags & XVID_CPU_SSE2) |
928 |
test_quant(dequant_inter_c, dequant_inter_sse2, "sse2", TEST_DEQUANT_INTER, flags); |
test_quant(dequant_inter_c, dequant_inter_sse2, "sse2", TEST_DEQUANT_INTER, flags); |
929 |
|
#endif |
930 |
|
|
931 |
printf("\n--- quant4_intra ---\n"); |
/* Intra quantization test */ |
932 |
|
printf("\n--- quant4 intra ---\n"); |
933 |
test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags); |
test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags); |
934 |
|
|
935 |
|
#if defined(ARCH_IS_IA32) |
936 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
937 |
test_quant(quant4_intra_c, quant4_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
test_quant(quant4_intra_c, quant4_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
938 |
if (cpu_flags & XVID_CPU_MMXEXT) |
if (cpu_flags & XVID_CPU_MMXEXT) |
939 |
test_quant(quant4_intra_c, quant4_intra_xmm, "xmm", TEST_QUANT_INTRA, flags); |
test_quant(quant4_intra_c, quant4_intra_xmm, "xmm", TEST_QUANT_INTRA, flags); |
940 |
|
#endif |
941 |
|
|
942 |
printf("\n--- quant4_inter ---\n"); |
/* Inter quantization test */ |
943 |
|
printf("\n--- quant4 inter ---\n"); |
944 |
test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags); |
test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags); |
945 |
|
|
946 |
|
#if defined(ARCH_IS_IA32) |
947 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
948 |
test_quant(quant4_inter_c, quant4_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
test_quant(quant4_inter_c, quant4_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
949 |
if (cpu_flags & XVID_CPU_MMXEXT) |
if (cpu_flags & XVID_CPU_MMXEXT) |
950 |
test_quant(quant4_inter_c, quant4_inter_xmm, "xmm", TEST_QUANT_INTER, flags); |
test_quant(quant4_inter_c, quant4_inter_xmm, "xmm", TEST_QUANT_INTER, flags); |
951 |
|
#endif |
952 |
|
|
953 |
printf("\n--- dequant4_intra ---\n"); |
/* Intra dequantization test */ |
954 |
|
printf("\n--- dequant4 intra ---\n"); |
955 |
test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
956 |
|
|
957 |
|
#if defined(ARCH_IS_IA32) |
958 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
959 |
test_quant(dequant4_intra_c, dequant4_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant4_intra_c, dequant4_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
960 |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
961 |
test_quant(dequant4_intra_c, dequant4_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
test_quant(dequant4_intra_c, dequant4_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
962 |
|
#endif |
963 |
|
|
964 |
printf("\n--- dequant4_inter ---\n"); |
/* Inter dequantization test */ |
965 |
|
printf("\n--- dequant4 inter ---\n"); |
966 |
test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags); |
test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags); |
967 |
|
|
968 |
|
#if defined(ARCH_IS_IA32) |
969 |
if (cpu_flags & XVID_CPU_MMX) |
if (cpu_flags & XVID_CPU_MMX) |
970 |
test_quant(dequant4_inter_c, dequant4_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
test_quant(dequant4_inter_c, dequant4_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
971 |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
if (cpu_flags & XVID_CPU_3DNOWEXT) |
972 |
test_quant(dequant4_inter_c, dequant4_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
test_quant(dequant4_inter_c, dequant4_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
|
|
|
|
emms_mmx(); |
|
|
|
|
973 |
#endif |
#endif |
974 |
|
|
975 |
return XVID_ERR_OK; |
emms(); |
976 |
|
|
977 |
|
return 0; |
978 |
} |
} |
979 |
|
|
980 |
|
|
981 |
|
/***************************************************************************** |
982 |
|
* XviD Global Entry point |
983 |
|
* |
984 |
|
* Well this function initialize all internal function pointers according |
985 |
|
* to the CPU features forced by the library client or autodetected (depending |
986 |
|
* on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all |
987 |
|
* image colorspace transformation tables. |
988 |
|
* |
989 |
|
****************************************************************************/ |
990 |
|
|
991 |
|
|
992 |
int |
int |
993 |
xvid_init(void *handle, |
xvid_global(void *handle, |
994 |
int opt, |
int opt, |
995 |
void *param1, |
void *param1, |
996 |
void *param2) |
void *param2) |
997 |
{ |
{ |
998 |
switch(opt) |
switch(opt) |
999 |
{ |
{ |
1000 |
case XVID_INIT_INIT : |
case XVID_GBL_INIT : |
1001 |
return xvid_init_init((XVID_INIT_PARAM*)param1); |
return xvid_gbl_init((xvid_gbl_init_t*)param1); |
1002 |
|
|
1003 |
case XVID_INIT_CONVERT : |
case XVID_GBL_INFO : |
1004 |
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
return xvid_gbl_info((xvid_gbl_info_t*)param1); |
1005 |
|
|
1006 |
case XVID_INIT_TEST : |
case XVID_GBL_CONVERT : |
1007 |
return xvid_init_test((int)param1); |
return xvid_gbl_convert((xvid_gbl_convert_t*)param1); |
1008 |
|
|
1009 |
|
case XVID_GBL_TEST : |
1010 |
|
{ |
1011 |
|
ptr_t flags = (ptr_t)param1; |
1012 |
|
return xvid_init_test((int)flags); |
1013 |
|
} |
1014 |
default : |
default : |
1015 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1016 |
} |
} |
1033 |
void *param2) |
void *param2) |
1034 |
{ |
{ |
1035 |
switch (opt) { |
switch (opt) { |
|
case XVID_DEC_DECODE: |
|
|
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2); |
|
|
|
|
1036 |
case XVID_DEC_CREATE: |
case XVID_DEC_CREATE: |
1037 |
return decoder_create((XVID_DEC_PARAM *) param1); |
return decoder_create((xvid_dec_create_t *) param1); |
1038 |
|
|
1039 |
case XVID_DEC_DESTROY: |
case XVID_DEC_DESTROY: |
1040 |
return decoder_destroy((DECODER *) handle); |
return decoder_destroy((DECODER *) handle); |
1041 |
|
|
1042 |
|
case XVID_DEC_DECODE: |
1043 |
|
return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); |
1044 |
|
|
1045 |
default: |
default: |
1046 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1047 |
} |
} |
1067 |
switch (opt) { |
switch (opt) { |
1068 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
1069 |
|
|
1070 |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
return enc_encode((Encoder *) handle, |
1071 |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
(xvid_enc_frame_t *) param1, |
1072 |
(XVID_ENC_STATS *) param2); |
(xvid_enc_stats_t *) param2); |
|
else |
|
|
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
|
|
(XVID_ENC_STATS *) param2); |
|
1073 |
|
|
1074 |
case XVID_ENC_CREATE: |
case XVID_ENC_CREATE: |
1075 |
return encoder_create((XVID_ENC_PARAM *) param1); |
return enc_create((xvid_enc_create_t *) param1); |
1076 |
|
|
1077 |
case XVID_ENC_DESTROY: |
case XVID_ENC_DESTROY: |
1078 |
return encoder_destroy((Encoder *) handle); |
return enc_destroy((Encoder *) handle); |
1079 |
|
|
1080 |
default: |
default: |
1081 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |