27 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
28 |
* |
* |
29 |
****************************************************************************/ |
****************************************************************************/ |
30 |
|
|
31 |
/***************************************************************************** |
/***************************************************************************** |
32 |
* |
* |
33 |
* History |
* History |
34 |
* |
* |
35 |
|
* - 23.06.2002 added XVID_CPU_CHKONLY |
36 |
* - 17.03.2002 Added interpolate8x8_halfpel_hv_xmm |
* - 17.03.2002 Added interpolate8x8_halfpel_hv_xmm |
37 |
* - 22.12.2001 API change: added xvid_init() - Isibaar |
* - 22.12.2001 API change: added xvid_init() - Isibaar |
38 |
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
39 |
* |
* |
40 |
* $Id: xvid.c,v 1.19 2002-06-16 19:53:44 edgomez Exp $ |
* $Id: xvid.c,v 1.33.2.8 2002-10-10 12:15:48 Isibaar Exp $ |
41 |
* |
* |
42 |
****************************************************************************/ |
****************************************************************************/ |
43 |
|
|
50 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
51 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
52 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
53 |
|
#include "utils/mbfunctions.h" |
54 |
#include "quant/quant_h263.h" |
#include "quant/quant_h263.h" |
55 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_mpeg4.h" |
56 |
|
#include "motion/motion.h" |
57 |
#include "motion/sad.h" |
#include "motion/sad.h" |
58 |
#include "utils/emms.h" |
#include "utils/emms.h" |
59 |
#include "utils/timer.h" |
#include "utils/timer.h" |
60 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
61 |
|
|
62 |
|
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
63 |
|
|
64 |
|
#ifdef WIN32 |
65 |
|
#include <windows.h> |
66 |
|
#else |
67 |
|
#include <signal.h> |
68 |
|
#include <setjmp.h> |
69 |
|
#endif |
70 |
|
|
71 |
|
|
72 |
|
#ifndef WIN32 |
73 |
|
|
74 |
|
static jmp_buf mark; |
75 |
|
|
76 |
|
static void |
77 |
|
sigill_handler(int signal) |
78 |
|
{ |
79 |
|
longjmp(mark, 1); |
80 |
|
} |
81 |
|
#endif |
82 |
|
|
83 |
|
|
84 |
|
/* |
85 |
|
calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
86 |
|
return values: |
87 |
|
-1 : could not determine |
88 |
|
0 : SIGILL was *not* signalled |
89 |
|
1 : SIGILL was signalled |
90 |
|
*/ |
91 |
|
|
92 |
|
int |
93 |
|
sigill_check(void (*func)()) |
94 |
|
{ |
95 |
|
#ifdef WIN32 |
96 |
|
_try { |
97 |
|
func(); |
98 |
|
} |
99 |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
100 |
|
|
101 |
|
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
102 |
|
return 1; |
103 |
|
} |
104 |
|
return 0; |
105 |
|
#else |
106 |
|
void * old_handler; |
107 |
|
int jmpret; |
108 |
|
|
109 |
|
|
110 |
|
old_handler = signal(SIGILL, sigill_handler); |
111 |
|
if (old_handler == SIG_ERR) |
112 |
|
{ |
113 |
|
return -1; |
114 |
|
} |
115 |
|
|
116 |
|
jmpret = setjmp(mark); |
117 |
|
if (jmpret == 0) |
118 |
|
{ |
119 |
|
func(); |
120 |
|
} |
121 |
|
|
122 |
|
signal(SIGILL, old_handler); |
123 |
|
|
124 |
|
return jmpret; |
125 |
|
#endif |
126 |
|
} |
127 |
|
#endif |
128 |
|
|
129 |
/***************************************************************************** |
/***************************************************************************** |
130 |
* XviD Init Entry point |
* XviD Init Entry point |
131 |
* |
* |
151 |
|
|
152 |
init_param = (XVID_INIT_PARAM *) param1; |
init_param = (XVID_INIT_PARAM *) param1; |
153 |
|
|
154 |
|
/* Inform the client the API version */ |
155 |
|
init_param->api_version = API_VERSION; |
156 |
|
|
157 |
|
/* Inform the client the core build - unused because we're still alpha */ |
158 |
|
init_param->core_build = 1000; |
159 |
|
|
160 |
/* Do we have to force CPU features ? */ |
/* Do we have to force CPU features ? */ |
161 |
if ((init_param->cpu_flags & XVID_CPU_FORCE) > 0) { |
if ((init_param->cpu_flags & XVID_CPU_FORCE)) { |
162 |
|
|
163 |
cpu_flags = init_param->cpu_flags; |
cpu_flags = init_param->cpu_flags; |
164 |
|
|
165 |
} else { |
} else { |
166 |
|
|
|
#ifdef ARCH_X86 |
|
167 |
cpu_flags = check_cpu_features(); |
cpu_flags = check_cpu_features(); |
168 |
#else |
|
169 |
cpu_flags = 0; |
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
170 |
|
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
171 |
|
cpu_flags &= ~XVID_CPU_SSE; |
172 |
|
|
173 |
|
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
174 |
|
cpu_flags &= ~XVID_CPU_SSE2; |
175 |
#endif |
#endif |
176 |
|
} |
177 |
|
|
178 |
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
179 |
|
{ |
180 |
init_param->cpu_flags = cpu_flags; |
init_param->cpu_flags = cpu_flags; |
181 |
|
return XVID_ERR_OK; |
182 |
} |
} |
183 |
|
|
184 |
|
init_param->cpu_flags = cpu_flags; |
185 |
|
|
186 |
|
|
187 |
/* Initialize the function pointers */ |
/* Initialize the function pointers */ |
188 |
idct_int32_init(); |
idct_int32_init(); |
189 |
init_vlc_tables(); |
init_vlc_tables(); |
217 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
218 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
219 |
|
|
220 |
|
/* Interlacing functions */ |
221 |
|
MBFieldTest = MBFieldTest_c; |
222 |
|
|
223 |
/* Image interpolation related functions */ |
/* Image interpolation related functions */ |
224 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
225 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
226 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
227 |
|
|
228 |
|
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
229 |
|
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
230 |
|
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
231 |
|
|
232 |
|
interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; |
233 |
|
interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; |
234 |
|
interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; |
235 |
|
|
236 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; |
237 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; |
238 |
|
|
239 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
240 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
241 |
|
|
242 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
243 |
colorspace_init(); |
colorspace_init(); |
244 |
|
|
256 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
257 |
yv12_to_rgb24 = yv12_to_rgb24_c; |
yv12_to_rgb24 = yv12_to_rgb24_c; |
258 |
yv12_to_rgb32 = yv12_to_rgb32_c; |
yv12_to_rgb32 = yv12_to_rgb32_c; |
259 |
|
yv12_to_abgr = yv12_to_abgr_c; |
260 |
|
yv12_to_rgba = yv12_to_rgba_c; |
261 |
yv12_to_yuv = yv12_to_yuv_c; |
yv12_to_yuv = yv12_to_yuv_c; |
262 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
263 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
265 |
/* Functions used in motion estimation algorithms */ |
/* Functions used in motion estimation algorithms */ |
266 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
267 |
sad16 = sad16_c; |
sad16 = sad16_c; |
|
sad16bi = sad16bi_c; |
|
268 |
sad8 = sad8_c; |
sad8 = sad8_c; |
269 |
|
sad16bi = sad16bi_c; |
270 |
|
sad8bi = sad8bi_c; |
271 |
dev16 = dev16_c; |
dev16 = dev16_c; |
272 |
|
sad16v = sad16v_c; |
273 |
|
|
274 |
|
// Halfpel8_Refine = Halfpel8_Refine_c; |
275 |
|
|
276 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
277 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
298 |
transfer_8to16copy = transfer_8to16copy_mmx; |
transfer_8to16copy = transfer_8to16copy_mmx; |
299 |
transfer_16to8copy = transfer_16to8copy_mmx; |
transfer_16to8copy = transfer_16to8copy_mmx; |
300 |
transfer_8to16sub = transfer_8to16sub_mmx; |
transfer_8to16sub = transfer_8to16sub_mmx; |
301 |
|
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
302 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
303 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
304 |
|
|
305 |
|
/* Interlacing Functions */ |
306 |
|
MBFieldTest = MBFieldTest_mmx; |
307 |
|
|
308 |
/* Image Interpolation related functions */ |
/* Image Interpolation related functions */ |
309 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
310 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
311 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
312 |
|
|
313 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
314 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
315 |
|
|
316 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
317 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
318 |
|
|
319 |
/* Image RGB->YV12 related functions */ |
/* Image RGB->YV12 related functions */ |
320 |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
321 |
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
333 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
334 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
335 |
sad8 = sad8_mmx; |
sad8 = sad8_mmx; |
336 |
|
sad16bi = sad16bi_mmx; |
337 |
|
sad8bi = sad8bi_mmx; |
338 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
339 |
|
sad16v = sad16v_mmx; |
340 |
|
|
341 |
|
} |
342 |
|
|
343 |
|
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
344 |
|
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
345 |
|
|
346 |
|
/* ME functions */ |
347 |
|
sad16bi = sad16bi_3dn; |
348 |
|
sad8bi = sad8bi_3dn; |
349 |
} |
} |
350 |
|
|
351 |
|
|
352 |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
353 |
|
|
354 |
/* Inverse DCT */ |
/* Inverse DCT */ |
359 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
360 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
361 |
|
|
362 |
|
/* Quantization */ |
363 |
|
dequant_intra = dequant_intra_xmm; |
364 |
|
dequant_inter = dequant_inter_xmm; |
365 |
|
|
366 |
/* Buffer transfer */ |
/* Buffer transfer */ |
367 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
368 |
|
|
372 |
/* ME functions */ |
/* ME functions */ |
373 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
374 |
sad8 = sad8_xmm; |
sad8 = sad8_xmm; |
375 |
|
sad16bi = sad16bi_xmm; |
376 |
|
sad8bi = sad8bi_xmm; |
377 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
378 |
|
sad16v = sad16v_xmm; |
379 |
|
fprintf(stderr,"sad16v=XMM\n"); |
380 |
|
|
381 |
} |
} |
382 |
|
|
391 |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
392 |
#ifdef EXPERIMENTAL_SSE2_CODE |
#ifdef EXPERIMENTAL_SSE2_CODE |
393 |
|
|
394 |
|
calc_cbp = calc_cbp_sse2; |
395 |
|
|
396 |
/* Quantization */ |
/* Quantization */ |
397 |
quant_intra = quant_intra_sse2; |
quant_intra = quant_intra_sse2; |
398 |
dequant_intra = dequant_intra_sse2; |
dequant_intra = dequant_intra_sse2; |
400 |
dequant_inter = dequant_inter_sse2; |
dequant_inter = dequant_inter_sse2; |
401 |
|
|
402 |
/* ME */ |
/* ME */ |
|
calc_cbp = calc_cbp_sse2; |
|
403 |
sad16 = sad16_sse2; |
sad16 = sad16_sse2; |
404 |
dev16 = dev16_sse2; |
dev16 = dev16_sse2; |
405 |
|
|
423 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
424 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
425 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
426 |
|
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
427 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
428 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
429 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |
452 |
#endif |
#endif |
453 |
#endif |
#endif |
454 |
|
|
|
/* Inform the client the API version */ |
|
|
init_param->api_version = API_VERSION; |
|
|
|
|
|
/* Inform the client the core build - unused because we're still alpha */ |
|
|
init_param->core_build = 1000; |
|
|
|
|
455 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
456 |
} |
} |
457 |
|
|
505 |
{ |
{ |
506 |
switch (opt) { |
switch (opt) { |
507 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
508 |
|
#ifdef BFRAMES |
509 |
|
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
510 |
|
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
511 |
|
(XVID_ENC_STATS *) param2); |
512 |
|
else |
513 |
|
#endif |
514 |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
515 |
(XVID_ENC_STATS *) param2); |
(XVID_ENC_STATS *) param2); |
516 |
|
|