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) 2001-2004 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 |
* |
* |
22 |
****************************************************************************/ |
* $Id: xvid.c,v 1.58 2004-12-09 23:02:54 edgomez Exp $ |
|
|
|
|
/***************************************************************************** |
|
|
* |
|
|
* 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> |
|
|
* |
|
|
* $Id: xvid.c,v 1.41 2003-02-15 15:22:17 edgomez 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_h263.h" |
#include "quant/quant.h" |
|
#include "quant/quant_mpeg4.h" |
|
42 |
#include "motion/motion.h" |
#include "motion/motion.h" |
43 |
#include "motion/sad.h" |
#include "motion/sad.h" |
44 |
#include "utils/emms.h" |
#include "utils/emms.h" |
45 |
#include "utils/timer.h" |
#include "utils/timer.h" |
46 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
47 |
|
#include "image/qpel.h" |
48 |
|
#include "image/postprocessing.h" |
49 |
|
|
50 |
#if defined(ARCH_IS_IA32) |
#if defined(_DEBUG) |
51 |
|
unsigned int xvid_debug = 0; /* xvid debug mask */ |
52 |
|
#endif |
53 |
|
|
54 |
#if defined(_MSC_VER) |
#if defined(ARCH_IS_IA32) && defined(_MSC_VER) |
55 |
# include <windows.h> |
# include <windows.h> |
56 |
#else |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
57 |
# include <signal.h> |
# include <signal.h> |
58 |
# include <setjmp.h> |
# include <setjmp.h> |
59 |
|
|
68 |
|
|
69 |
|
|
70 |
/* |
/* |
71 |
calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
* Calls the funcptr, and returns whether SIGILL (illegal instruction) was |
72 |
return values: |
* signalled |
73 |
-1 : could not determine |
* |
74 |
0 : SIGILL was *not* signalled |
* Return values: |
75 |
1 : SIGILL was signalled |
* -1 : could not determine |
76 |
|
* 0 : SIGILL was *not* signalled |
77 |
|
* 1 : SIGILL was signalled |
78 |
*/ |
*/ |
79 |
|
#if defined(ARCH_IS_IA32) && defined(_MSC_VER) |
80 |
int |
static int |
81 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
82 |
{ |
{ |
|
#if defined(_MSC_VER) |
|
83 |
_try { |
_try { |
84 |
func(); |
func(); |
85 |
} |
} _except(EXCEPTION_EXECUTE_HANDLER) { |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
|
86 |
|
|
87 |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
88 |
return 1; |
return(1); |
89 |
} |
} |
90 |
return 0; |
return(0); |
91 |
#else |
} |
92 |
|
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
93 |
|
static int |
94 |
|
sigill_check(void (*func)()) |
95 |
|
{ |
96 |
void * old_handler; |
void * old_handler; |
97 |
int jmpret; |
int jmpret; |
98 |
|
|
99 |
|
/* Set our SIGILL handler */ |
100 |
old_handler = signal(SIGILL, sigill_handler); |
old_handler = signal(SIGILL, sigill_handler); |
101 |
if (old_handler == SIG_ERR) |
|
102 |
{ |
/* Check for error */ |
103 |
return -1; |
if (old_handler == SIG_ERR) { |
104 |
|
return(-1); |
105 |
} |
} |
106 |
|
|
107 |
|
/* Save stack context, so if func triggers a SIGILL, we can still roll |
108 |
|
* back to a valid CPU state */ |
109 |
jmpret = setjmp(mark); |
jmpret = setjmp(mark); |
110 |
if (jmpret == 0) |
|
111 |
{ |
/* If setjmp returned directly, then its returned value is 0, and we still |
112 |
|
* have to test the passed func. Otherwise it means the stack context has |
113 |
|
* been restored by a longjmp() call, which in our case happens only in the |
114 |
|
* signal handler */ |
115 |
|
if (jmpret == 0) { |
116 |
func(); |
func(); |
117 |
} |
} |
118 |
|
|
119 |
|
/* Restore old signal handler */ |
120 |
signal(SIGILL, old_handler); |
signal(SIGILL, old_handler); |
121 |
|
|
122 |
return jmpret; |
return(jmpret); |
|
#endif |
|
123 |
} |
} |
124 |
#endif |
#endif |
125 |
|
|
141 |
#endif |
#endif |
142 |
|
|
143 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
144 |
#if defined(ARCH_IS_PPC_ALTIVEC) |
if (!sigill_check(altivec_trigger)) |
145 |
cpu_flags |= XVID_CPU_ALTIVEC; |
cpu_flags |= XVID_CPU_ALTIVEC; |
146 |
#endif |
#endif |
|
#endif |
|
147 |
|
|
148 |
return cpu_flags; |
return cpu_flags; |
149 |
} |
} |
165 |
|
|
166 |
|
|
167 |
static |
static |
168 |
int xvid_init_init(XVID_INIT_PARAM * init_param) |
int xvid_gbl_init(xvid_gbl_init_t * init) |
|
{ |
|
|
int cpu_flags; |
|
|
|
|
|
/* 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; |
|
|
|
|
|
/* Do we have to force CPU features ? */ |
|
|
if ((init_param->cpu_flags & XVID_CPU_FORCE)) { |
|
|
|
|
|
cpu_flags = init_param->cpu_flags; |
|
|
|
|
|
} else { |
|
|
|
|
|
cpu_flags = detect_cpu_flags(); |
|
|
} |
|
|
|
|
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
|
169 |
{ |
{ |
170 |
init_param->cpu_flags = cpu_flags; |
unsigned int cpu_flags; |
|
return XVID_ERR_OK; |
|
|
} |
|
171 |
|
|
172 |
init_param->cpu_flags = cpu_flags; |
if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */ |
173 |
|
return XVID_ERR_VERSION; |
174 |
|
|
175 |
|
cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
176 |
|
|
177 |
/* Initialize the function pointers */ |
/* Initialize the function pointers */ |
178 |
idct_int32_init(); |
idct_int32_init(); |
188 |
/* Restore FPU context : emms_c is a nop functions */ |
/* Restore FPU context : emms_c is a nop functions */ |
189 |
emms = emms_c; |
emms = emms_c; |
190 |
|
|
191 |
|
/* Qpel stuff */ |
192 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_C; |
193 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_C; |
194 |
|
xvid_Init_QP(); |
195 |
|
|
196 |
/* Quantization functions */ |
/* Quantization functions */ |
197 |
quant_intra = quant_intra_c; |
quant_h263_intra = quant_h263_intra_c; |
198 |
dequant_intra = dequant_intra_c; |
quant_h263_inter = quant_h263_inter_c; |
199 |
quant_inter = quant_inter_c; |
dequant_h263_intra = dequant_h263_intra_c; |
200 |
dequant_inter = dequant_inter_c; |
dequant_h263_inter = dequant_h263_inter_c; |
201 |
|
|
202 |
quant4_intra = quant4_intra_c; |
quant_mpeg_intra = quant_mpeg_intra_c; |
203 |
dequant4_intra = dequant4_intra_c; |
quant_mpeg_inter = quant_mpeg_inter_c; |
204 |
quant4_inter = quant4_inter_c; |
dequant_mpeg_intra = dequant_mpeg_intra_c; |
205 |
dequant4_inter = dequant4_inter_c; |
dequant_mpeg_inter = dequant_mpeg_inter_c; |
206 |
|
|
207 |
/* Block transfer related functions */ |
/* Block transfer related functions */ |
208 |
transfer_8to16copy = transfer_8to16copy_c; |
transfer_8to16copy = transfer_8to16copy_c; |
221 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
222 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
223 |
|
|
224 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_c; |
225 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_c; |
226 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_c; |
227 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_c; |
228 |
|
|
229 |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
230 |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
231 |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
240 |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
241 |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
242 |
|
|
243 |
/* reduced resoltuion */ |
/* postprocessing */ |
244 |
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; |
|
245 |
|
|
246 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
247 |
colorspace_init(); |
colorspace_init(); |
254 |
bgra_to_yv12 = bgra_to_yv12_c; |
bgra_to_yv12 = bgra_to_yv12_c; |
255 |
abgr_to_yv12 = abgr_to_yv12_c; |
abgr_to_yv12 = abgr_to_yv12_c; |
256 |
rgba_to_yv12 = rgba_to_yv12_c; |
rgba_to_yv12 = rgba_to_yv12_c; |
257 |
|
argb_to_yv12 = argb_to_yv12_c; |
258 |
yuyv_to_yv12 = yuyv_to_yv12_c; |
yuyv_to_yv12 = yuyv_to_yv12_c; |
259 |
uyvy_to_yv12 = uyvy_to_yv12_c; |
uyvy_to_yv12 = uyvy_to_yv12_c; |
260 |
|
|
264 |
bgrai_to_yv12 = bgrai_to_yv12_c; |
bgrai_to_yv12 = bgrai_to_yv12_c; |
265 |
abgri_to_yv12 = abgri_to_yv12_c; |
abgri_to_yv12 = abgri_to_yv12_c; |
266 |
rgbai_to_yv12 = rgbai_to_yv12_c; |
rgbai_to_yv12 = rgbai_to_yv12_c; |
267 |
|
argbi_to_yv12 = argbi_to_yv12_c; |
268 |
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
269 |
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
270 |
|
|
|
|
|
271 |
/* All colorspace transformation functions YV12->User format */ |
/* All colorspace transformation functions YV12->User format */ |
272 |
yv12_to_rgb555 = yv12_to_rgb555_c; |
yv12_to_rgb555 = yv12_to_rgb555_c; |
273 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
275 |
yv12_to_bgra = yv12_to_bgra_c; |
yv12_to_bgra = yv12_to_bgra_c; |
276 |
yv12_to_abgr = yv12_to_abgr_c; |
yv12_to_abgr = yv12_to_abgr_c; |
277 |
yv12_to_rgba = yv12_to_rgba_c; |
yv12_to_rgba = yv12_to_rgba_c; |
278 |
|
yv12_to_argb = yv12_to_argb_c; |
279 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
280 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
281 |
|
|
285 |
yv12_to_bgrai = yv12_to_bgrai_c; |
yv12_to_bgrai = yv12_to_bgrai_c; |
286 |
yv12_to_abgri = yv12_to_abgri_c; |
yv12_to_abgri = yv12_to_abgri_c; |
287 |
yv12_to_rgbai = yv12_to_rgbai_c; |
yv12_to_rgbai = yv12_to_rgbai_c; |
288 |
|
yv12_to_argbi = yv12_to_argbi_c; |
289 |
yv12_to_yuyvi = yv12_to_yuyvi_c; |
yv12_to_yuyvi = yv12_to_yuyvi_c; |
290 |
yv12_to_uyvyi = yv12_to_uyvyi_c; |
yv12_to_uyvyi = yv12_to_uyvyi_c; |
291 |
|
|
297 |
sad8bi = sad8bi_c; |
sad8bi = sad8bi_c; |
298 |
dev16 = dev16_c; |
dev16 = dev16_c; |
299 |
sad16v = sad16v_c; |
sad16v = sad16v_c; |
300 |
|
sse8_16bit = sse8_16bit_c; |
301 |
// Halfpel8_Refine = Halfpel8_Refine_c; |
sse8_8bit = sse8_8bit_c; |
302 |
|
|
303 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) |
304 |
|
|
|
if ((cpu_flags & XVID_CPU_ASM)) |
|
|
{ |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
} |
|
|
|
|
305 |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
306 |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
307 |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
313 |
if ((cpu_flags & XVID_CPU_MMX)) { |
if ((cpu_flags & XVID_CPU_MMX)) { |
314 |
|
|
315 |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
316 |
fdct = fdct_mmx; |
fdct = fdct_mmx_skal; |
317 |
idct = idct_mmx; |
idct = idct_mmx; |
318 |
|
|
319 |
|
/* Qpel stuff */ |
320 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_mmx; |
321 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx; |
322 |
|
|
323 |
/* Quantization related functions */ |
/* Quantization related functions */ |
324 |
quant_intra = quant_intra_mmx; |
quant_h263_intra = quant_h263_intra_mmx; |
325 |
dequant_intra = dequant_intra_mmx; |
quant_h263_inter = quant_h263_inter_mmx; |
326 |
quant_inter = quant_inter_mmx; |
dequant_h263_intra = dequant_h263_intra_mmx; |
327 |
dequant_inter = dequant_inter_mmx; |
dequant_h263_inter = dequant_h263_inter_mmx; |
328 |
|
|
329 |
quant4_intra = quant4_intra_mmx; |
quant_mpeg_intra = quant_mpeg_intra_mmx; |
330 |
dequant4_intra = dequant4_intra_mmx; |
quant_mpeg_inter = quant_mpeg_inter_mmx; |
331 |
quant4_inter = quant4_inter_mmx; |
dequant_mpeg_intra = dequant_mpeg_intra_mmx; |
332 |
dequant4_inter = dequant4_inter_mmx; |
dequant_mpeg_inter = dequant_mpeg_inter_mmx; |
333 |
|
|
334 |
/* Block related functions */ |
/* Block related functions */ |
335 |
transfer_8to16copy = transfer_8to16copy_mmx; |
transfer_8to16copy = transfer_8to16copy_mmx; |
348 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
349 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
350 |
|
|
351 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_mmx; |
352 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_mmx; |
353 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_mmx; |
354 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_mmx; |
355 |
|
|
356 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
357 |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
358 |
|
|
359 |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
360 |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
361 |
|
|
362 |
/* reduced resolution */ |
/* postprocessing */ |
363 |
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; |
|
364 |
|
|
365 |
/* image input xxx_to_yv12 related functions */ |
/* image input xxx_to_yv12 related functions */ |
366 |
yv12_to_yv12 = yv12_to_yv12_mmx; |
yv12_to_yv12 = yv12_to_yv12_mmx; |
386 |
sad8bi = sad8bi_mmx; |
sad8bi = sad8bi_mmx; |
387 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
388 |
sad16v = sad16v_mmx; |
sad16v = sad16v_mmx; |
389 |
|
sse8_16bit = sse8_16bit_mmx; |
390 |
|
sse8_8bit = sse8_8bit_mmx; |
391 |
} |
} |
392 |
|
|
393 |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
406 |
|
|
407 |
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
408 |
|
|
409 |
/* Inverse DCT */ |
/* DCT */ |
410 |
|
fdct = fdct_xmm_skal; |
411 |
idct = idct_xmm; |
idct = idct_xmm; |
412 |
|
|
413 |
/* Interpolation */ |
/* Interpolation */ |
415 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
416 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
417 |
|
|
418 |
/* reduced resolution */ |
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_xmm; |
419 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm; |
420 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm; |
421 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm; |
422 |
|
|
423 |
/* Quantization */ |
/* Quantization */ |
424 |
quant4_intra = quant4_intra_xmm; |
quant_mpeg_intra = quant_mpeg_intra_xmm; |
425 |
quant4_inter = quant4_inter_xmm; |
quant_mpeg_inter = quant_mpeg_inter_xmm; |
426 |
|
|
427 |
dequant_intra = dequant_intra_xmm; |
dequant_h263_intra = dequant_h263_intra_xmm; |
428 |
dequant_inter = dequant_inter_xmm; |
dequant_h263_inter = dequant_h263_inter_xmm; |
429 |
|
|
430 |
/* Buffer transfer */ |
/* Buffer transfer */ |
431 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
454 |
|
|
455 |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
456 |
|
|
|
/* Inverse DCT */ |
|
|
idct = idct_3dne; |
|
|
|
|
457 |
/* Buffer transfer */ |
/* Buffer transfer */ |
458 |
transfer_8to16copy = transfer_8to16copy_3dne; |
transfer_8to16copy = transfer_8to16copy_3dne; |
459 |
transfer_16to8copy = transfer_16to8copy_3dne; |
transfer_16to8copy = transfer_16to8copy_3dne; |
460 |
transfer_8to16sub = transfer_8to16sub_3dne; |
transfer_8to16sub = transfer_8to16sub_3dne; |
461 |
transfer_8to16subro = transfer_8to16subro_3dne; |
transfer_8to16subro = transfer_8to16subro_3dne; |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
|
462 |
transfer_16to8add = transfer_16to8add_3dne; |
transfer_16to8add = transfer_16to8add_3dne; |
463 |
transfer8x8_copy = transfer8x8_copy_3dne; |
transfer8x8_copy = transfer8x8_copy_3dne; |
464 |
|
|
465 |
|
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
466 |
|
/* Inverse DCT */ |
467 |
|
idct = idct_3dne; |
468 |
|
|
469 |
|
/* Buffer transfer */ |
470 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
471 |
|
|
472 |
|
/* Interpolation */ |
473 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
474 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
475 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
476 |
|
|
477 |
/* Quantization */ |
/* Quantization */ |
478 |
dequant4_intra = dequant4_intra_3dne; |
quant_h263_intra = quant_h263_intra_3dne; /* cmov only */ |
479 |
dequant4_inter = dequant4_inter_3dne; |
quant_h263_inter = quant_h263_inter_3dne; |
480 |
quant_intra = quant_intra_3dne; |
dequant_mpeg_intra = dequant_mpeg_intra_3dne; /* cmov only */ |
481 |
quant_inter = quant_inter_3dne; |
dequant_mpeg_inter = dequant_mpeg_inter_3dne; |
482 |
dequant_intra = dequant_intra_3dne; |
dequant_h263_intra = dequant_h263_intra_3dne; |
483 |
dequant_inter = dequant_inter_3dne; |
dequant_h263_inter = dequant_h263_inter_3dne; |
484 |
|
|
485 |
/* ME functions */ |
/* ME functions */ |
486 |
calc_cbp = calc_cbp_3dne; |
calc_cbp = calc_cbp_3dne; |
487 |
|
|
488 |
sad16 = sad16_3dne; |
sad16 = sad16_3dne; |
489 |
sad8 = sad8_3dne; |
sad8 = sad8_3dne; |
490 |
sad16bi = sad16bi_3dne; |
sad16bi = sad16bi_3dne; |
491 |
sad8bi = sad8bi_3dne; |
sad8bi = sad8bi_3dne; |
492 |
dev16 = dev16_3dne; |
dev16 = dev16_3dne; |
|
|
|
|
/* Interpolation */ |
|
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
|
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
|
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
|
493 |
} |
} |
494 |
|
} |
495 |
|
|
496 |
if ((cpu_flags & XVID_CPU_SSE2)) { |
if ((cpu_flags & XVID_CPU_SSE2)) { |
497 |
|
|
498 |
calc_cbp = calc_cbp_sse2; |
calc_cbp = calc_cbp_sse2; |
499 |
|
|
500 |
/* Quantization */ |
/* Quantization */ |
501 |
quant_intra = quant_intra_sse2; |
quant_h263_intra = quant_h263_intra_sse2; |
502 |
dequant_intra = dequant_intra_sse2; |
quant_h263_inter = quant_h263_inter_sse2; |
503 |
quant_inter = quant_inter_sse2; |
dequant_h263_intra = dequant_h263_intra_sse2; |
504 |
dequant_inter = dequant_inter_sse2; |
dequant_h263_inter = dequant_h263_inter_sse2; |
505 |
|
|
506 |
#if defined(EXPERIMENTAL_SSE2_CODE) |
/* SAD operators */ |
|
/* ME; slower than xmm */ |
|
507 |
sad16 = sad16_sse2; |
sad16 = sad16_sse2; |
508 |
dev16 = dev16_sse2; |
dev16 = dev16_sse2; |
509 |
#endif |
|
510 |
/* Forward and Inverse DCT */ |
/* DCT operators |
511 |
idct = idct_sse2; |
* no iDCT because it's not "Walken matching" */ |
512 |
fdct = fdct_sse2; |
fdct = fdct_sse2_skal; |
513 |
|
|
514 |
|
/* postprocessing */ |
515 |
|
image_brightness = image_brightness_sse2; |
516 |
} |
} |
517 |
#endif |
#endif /* ARCH_IS_IA32 */ |
518 |
|
|
519 |
#if defined(ARCH_IS_IA64) |
#if defined(ARCH_IS_IA64) |
520 |
if ((cpu_flags & XVID_CPU_ASM)) { //use assembler routines? |
if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
521 |
idct_ia64_init(); |
idct_ia64_init(); |
522 |
fdct = fdct_ia64; |
fdct = fdct_ia64; |
523 |
idct = idct_ia64; //not yet working, crashes |
idct = idct_ia64; /*not yet working, crashes */ |
524 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
525 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
526 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
528 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
529 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
530 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
531 |
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
/* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
532 |
quant_intra = quant_intra_ia64; |
quant_h263_intra = quant_h263_intra_ia64; |
533 |
dequant_intra = dequant_intra_ia64; |
quant_h263_inter = quant_h263_inter_ia64; |
534 |
quant_inter = quant_inter_ia64; |
dequant_h263_intra = dequant_h263_intra_ia64; |
535 |
dequant_inter = dequant_inter_ia64; |
dequant_h263_inter = dequant_h263_inter_ia64; |
536 |
transfer_8to16copy = transfer_8to16copy_ia64; |
transfer_8to16copy = transfer_8to16copy_ia64; |
537 |
transfer_16to8copy = transfer_16to8copy_ia64; |
transfer_16to8copy = transfer_16to8copy_ia64; |
538 |
transfer_8to16sub = transfer_8to16sub_ia64; |
transfer_8to16sub = transfer_8to16sub_ia64; |
539 |
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
540 |
transfer_16to8add = transfer_16to8add_ia64; |
transfer_16to8add = transfer_16to8add_ia64; |
541 |
transfer8x8_copy = transfer8x8_copy_ia64; |
transfer8x8_copy = transfer8x8_copy_ia64; |
|
DEBUG("Using IA-64 assembler routines.\n"); |
|
542 |
} |
} |
543 |
#endif |
#endif |
544 |
|
|
545 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
546 |
if ((cpu_flags & XVID_CPU_ASM)) |
if ((cpu_flags & XVID_CPU_ALTIVEC)) { |
547 |
{ |
/* sad operators */ |
548 |
calc_cbp = calc_cbp_ppc; |
sad16 = sad16_altivec_c; |
549 |
} |
sad16bi = sad16bi_altivec_c; |
550 |
|
sad8 = sad8_altivec_c; |
551 |
if ((cpu_flags & XVID_CPU_ALTIVEC)) |
dev16 = dev16_altivec_c; |
552 |
{ |
|
553 |
calc_cbp = calc_cbp_altivec; |
sse8_16bit = sse8_16bit_altivec_c; |
554 |
fdct = fdct_altivec; |
|
555 |
idct = idct_altivec; |
/* mem transfer */ |
556 |
sadInit = sadInit_altivec; |
transfer_8to16copy = transfer_8to16copy_altivec_c; |
557 |
sad16 = sad16_altivec; |
transfer_16to8copy = transfer_16to8copy_altivec_c; |
558 |
sad8 = sad8_altivec; |
transfer_8to16sub = transfer_8to16sub_altivec_c; |
559 |
dev16 = dev16_altivec; |
transfer_8to16subro = transfer_8to16subro_altivec_c; |
560 |
} |
transfer_8to16sub2 = transfer_8to16sub2_altivec_c; |
561 |
#endif |
transfer_16to8add = transfer_16to8add_altivec_c; |
562 |
|
transfer8x8_copy = transfer8x8_copy_altivec_c; |
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
static int |
|
|
xvid_init_convert(XVID_INIT_CONVERTINFO* convert) |
|
|
{ |
|
|
// const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
|
|
const int width = convert->width; |
|
|
const int height = convert->height; |
|
|
const int width2 = convert->width/2; |
|
|
const int height2 = convert->height/2; |
|
|
IMAGE img; |
|
|
|
|
|
switch (convert->input.colorspace & ~XVID_CSP_VFLIP) |
|
|
{ |
|
|
case XVID_CSP_YV12 : |
|
|
img.y = convert->input.y; |
|
|
img.v = (uint8_t*)convert->input.y + width*height; |
|
|
img.u = (uint8_t*)convert->input.y + width*height + width2*height2; |
|
|
image_output(&img, width, height, width, |
|
|
convert->output.y, convert->output.y_stride, |
|
|
convert->output.colorspace, convert->interlacing); |
|
|
break; |
|
|
|
|
|
default : |
|
|
return XVID_ERR_FORMAT; |
|
|
} |
|
|
|
|
|
|
|
|
emms(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
void fill8(uint8_t * block, int size, int value) |
|
|
{ |
|
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = value; |
|
|
} |
|
|
|
|
|
void fill16(int16_t * block, int size, int value) |
|
|
{ |
|
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = value; |
|
|
} |
|
563 |
|
|
564 |
#define RANDOM(min,max) min + (rand() % (max-min)) |
/* Inverse DCT */ |
565 |
|
idct = idct_altivec_c; |
566 |
|
|
567 |
void random8(uint8_t * block, int size, int min, int max) |
/* Interpolation */ |
568 |
{ |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c; |
569 |
int i; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c; |
570 |
for (i = 0; i < size; i++) |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c; |
571 |
block[i] = RANDOM(min,max); |
|
572 |
} |
interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; |
573 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; |
574 |
|
|
575 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c; |
576 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c; |
577 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c; |
578 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c; |
579 |
|
|
580 |
|
/* Colorspace conversion */ |
581 |
|
bgra_to_yv12 = bgra_to_yv12_altivec_c; |
582 |
|
abgr_to_yv12 = abgr_to_yv12_altivec_c; |
583 |
|
rgba_to_yv12 = rgba_to_yv12_altivec_c; |
584 |
|
argb_to_yv12 = argb_to_yv12_altivec_c; |
585 |
|
|
586 |
void random16(int16_t * block, int size, int min, int max) |
yuyv_to_yv12 = yuyv_to_yv12_altivec_c; |
587 |
{ |
uyvy_to_yv12 = uyvy_to_yv12_altivec_c; |
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = RANDOM(min,max); |
|
|
} |
|
588 |
|
|
589 |
int compare16(const int16_t * blockA, const int16_t * blockB, int size) |
yv12_to_yuyv = yv12_to_yuyv_altivec_c; |
590 |
{ |
yv12_to_uyvy = yv12_to_uyvy_altivec_c; |
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
if (blockA[i] != blockB[i]) |
|
|
return 1; |
|
591 |
|
|
592 |
return 0; |
/* Quantization */ |
593 |
} |
quant_h263_intra = quant_h263_intra_altivec_c; |
594 |
|
quant_h263_inter = quant_h263_inter_altivec_c; |
595 |
|
dequant_h263_intra = dequant_h263_intra_altivec_c; |
596 |
|
dequant_h263_inter = dequant_h263_inter_altivec_c; |
597 |
|
|
598 |
int diff16(const int16_t * blockA, const int16_t * blockB, int size) |
/* Qpel stuff */ |
599 |
{ |
xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C; |
600 |
int i, diff = 0; |
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C; |
|
for (i = 0; i < size; i++) |
|
|
diff += ABS(blockA[i]-blockB[i]); |
|
|
return diff; |
|
601 |
} |
} |
602 |
|
#endif |
603 |
|
|
604 |
|
#if defined(_DEBUG) |
605 |
|
xvid_debug = init->debug; |
606 |
|
#endif |
607 |
|
|
608 |
#define XVID_TEST_RANDOM 0x00000001 /* random input data */ |
return(0); |
|
#define XVID_TEST_VERBOSE 0x00000002 /* verbose error output */ |
|
|
|
|
|
|
|
|
#define TEST_FORWARD 0x00000001 /* intra */ |
|
|
#define TEST_FDCT (TEST_FORWARD) |
|
|
#define TEST_IDCT (0) |
|
|
|
|
|
int test_transform(void * funcA, void * funcB, const char * nameB, |
|
|
int test, int flags) |
|
|
{ |
|
|
int i; |
|
|
int64_t timeSTART; |
|
|
int64_t timeA = 0; |
|
|
int64_t timeB = 0; |
|
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
|
|
int min, max; |
|
|
int count = 0; |
|
|
|
|
|
int tmp; |
|
|
int min_error = 0x10000*64; |
|
|
int max_error = 0; |
|
|
|
|
|
|
|
|
if ((test & TEST_FORWARD)) /* forward */ |
|
|
{ |
|
|
min = -256; |
|
|
max = 255; |
|
|
}else{ /* inverse */ |
|
|
min = -2048; |
|
|
max = 2047; |
|
|
} |
|
|
|
|
|
for (i = 0; i < 64*64; i++) |
|
|
{ |
|
|
if ((flags & XVID_TEST_RANDOM)) |
|
|
{ |
|
|
random16(arrayA, 64, min, max); |
|
|
}else{ |
|
|
fill16(arrayA, 64, i); |
|
609 |
} |
} |
|
memcpy(arrayB, arrayA, 64*sizeof(int16_t)); |
|
610 |
|
|
|
if ((test & TEST_FORWARD)) |
|
|
{ |
|
|
timeSTART = read_counter(); |
|
|
((fdctFunc*)funcA)(arrayA); |
|
|
timeA += read_counter() - timeSTART; |
|
611 |
|
|
612 |
timeSTART = read_counter(); |
static int |
613 |
((fdctFunc*)funcB)(arrayB); |
xvid_gbl_info(xvid_gbl_info_t * info) |
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
else |
|
614 |
{ |
{ |
615 |
timeSTART = read_counter(); |
if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */ |
616 |
((idctFunc*)funcA)(arrayA); |
return XVID_ERR_VERSION; |
|
timeA += read_counter() - timeSTART; |
|
|
|
|
|
timeSTART = read_counter(); |
|
|
((idctFunc*)funcB)(arrayB); |
|
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
617 |
|
|
618 |
tmp = diff16(arrayA, arrayB, 64) / 64; |
info->actual_version = XVID_VERSION; |
619 |
if (tmp > max_error) |
info->build = "xvid-1.1-cvshead"; |
620 |
max_error = tmp; |
info->cpu_flags = detect_cpu_flags(); |
|
if (tmp < min_error) |
|
|
min_error = tmp; |
|
621 |
|
|
622 |
count++; |
#if defined(_SMP) && defined(WIN32) |
623 |
} |
info->num_threads = pthread_num_processors_np();; |
624 |
|
#else |
625 |
/* print the "average difference" of best/worst transforms */ |
info->num_threads = 0; |
626 |
printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error); |
#endif |
627 |
|
|
628 |
return 0; |
return 0; |
629 |
} |
} |
630 |
|
|
631 |
|
|
632 |
#define TEST_QUANT 0x00000001 /* forward quantization */ |
static int |
633 |
#define TEST_INTRA 0x00000002 /* intra */ |
xvid_gbl_convert(xvid_gbl_convert_t* convert) |
|
#define TEST_QUANT_INTRA (TEST_QUANT|TEST_INTRA) |
|
|
#define TEST_QUANT_INTER (TEST_QUANT) |
|
|
#define TEST_DEQUANT_INTRA (TEST_INTRA) |
|
|
#define TEST_DEQUANT_INTER (0) |
|
|
|
|
|
int test_quant(void * funcA, void * funcB, const char * nameB, |
|
|
int test, int flags) |
|
|
{ |
|
|
int q,i; |
|
|
int64_t timeSTART; |
|
|
int64_t timeA = 0; |
|
|
int64_t timeB = 0; |
|
|
int retA, retB; |
|
|
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
|
|
int min, max; |
|
|
int count = 0; |
|
|
int errors = 0; |
|
|
|
|
|
if ((test & TEST_QUANT)) /* quant */ |
|
|
{ |
|
|
min = -2048; |
|
|
max = 2047; |
|
|
}else{ /* dequant */ |
|
|
min = -256; |
|
|
max = 255; |
|
|
} |
|
|
|
|
|
for (q = 1; q <= 31; q++) /* quantizer */ |
|
|
{ |
|
|
for (i = min; i < max; i++) /* input coeff */ |
|
|
{ |
|
|
if ((flags & XVID_TEST_RANDOM)) |
|
|
{ |
|
|
random16(arrayX, 64, min, max); |
|
|
}else{ |
|
|
fill16(arrayX, 64, i); |
|
|
} |
|
|
|
|
|
if ((test & TEST_INTRA)) /* intra */ |
|
634 |
{ |
{ |
635 |
timeSTART = read_counter(); |
int width; |
636 |
((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q); |
int height; |
637 |
timeA += read_counter() - timeSTART; |
int width2; |
638 |
|
int height2; |
639 |
|
IMAGE img; |
640 |
|
|
641 |
timeSTART = read_counter(); |
if (XVID_VERSION_MAJOR(convert->version) != 1) /* v1.x.x */ |
642 |
((quanth263_intraFunc*)funcB)(arrayB, arrayX, q, q); |
return XVID_ERR_VERSION; |
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
else /* inter */ |
|
|
{ |
|
|
timeSTART = read_counter(); |
|
|
retA = ((quanth263_interFunc*)funcA)(arrayA, arrayX, q); |
|
|
timeA += read_counter() - timeSTART; |
|
643 |
|
|
644 |
timeSTART = read_counter(); |
#if 0 |
645 |
retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q); |
const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
646 |
timeB += read_counter() - timeSTART; |
#endif |
647 |
} |
width = convert->width; |
648 |
|
height = convert->height; |
649 |
|
width2 = convert->width/2; |
650 |
|
height2 = convert->height/2; |
651 |
|
|
652 |
/* compare return value from quant_inter, and compare (de)quantiz'd arrays */ |
switch (convert->input.csp & ~XVID_CSP_VFLIP) |
|
if ( ((test&TEST_QUANT) && !(test&TEST_INTRA) && retA != retB ) || |
|
|
compare16(arrayA, arrayB, 64)) |
|
653 |
{ |
{ |
654 |
errors++; |
case XVID_CSP_YV12 : |
655 |
if ((flags & XVID_TEST_VERBOSE)) |
img.y = convert->input.plane[0]; |
656 |
printf("%s error: q=%i, i=%i\n", nameB, q, i); |
img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; |
657 |
} |
img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; |
658 |
|
image_output(&img, width, height, width, |
659 |
|
(uint8_t**)convert->output.plane, convert->output.stride, |
660 |
|
convert->output.csp, convert->interlacing); |
661 |
|
break; |
662 |
|
|
663 |
count++; |
default : |
664 |
} |
return XVID_ERR_FORMAT; |
665 |
} |
} |
666 |
|
|
|
printf("%s:\t%i", nameB, (int)(timeB / count)); |
|
|
if (errors>0) |
|
|
printf("\t(%i errors out of %i)", errors, count); |
|
|
printf("\n"); |
|
667 |
|
|
668 |
|
emms(); |
669 |
return 0; |
return 0; |
670 |
} |
} |
671 |
|
|
672 |
|
/***************************************************************************** |
673 |
|
* XviD Global Entry point |
674 |
int xvid_init_test(int flags) |
* |
675 |
{ |
* Well this function initialize all internal function pointers according |
676 |
int cpu_flags; |
* to the CPU features forced by the library client or autodetected (depending |
677 |
|
* on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all |
678 |
srand(time(0)); |
* image colorspace transformation tables. |
679 |
|
* |
680 |
printf("xvid_init_test\n"); |
****************************************************************************/ |
|
|
|
|
#if defined(ARCH_X86) |
|
|
cpu_flags = detect_cpu_flags(); |
|
|
idct_int32_init(); |
|
|
emms_mmx(); |
|
|
|
|
|
printf("--- fdct ---\n"); |
|
|
test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_transform(fdct_int32, fdct_mmx, "mmx", TEST_FDCT, flags); |
|
|
if (cpu_flags & XVID_CPU_SSE2) |
|
|
test_transform(fdct_int32, fdct_sse2, "sse2", TEST_FDCT, flags); |
|
|
|
|
|
printf("\n--- idct ---\n"); |
|
|
test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_transform(idct_int32, idct_mmx, "mmx", TEST_IDCT, flags); |
|
|
if (cpu_flags & XVID_CPU_MMXEXT) |
|
|
test_transform(idct_int32, idct_xmm, "xmm", TEST_IDCT, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_transform(idct_int32, idct_3dne, "3dne", TEST_IDCT, flags); |
|
|
if (cpu_flags & XVID_CPU_SSE2) |
|
|
test_transform(idct_int32, idct_sse2, "sse2", TEST_IDCT, flags); |
|
|
|
|
|
printf("\n--- quant intra ---\n"); |
|
|
test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(quant_intra_c, quant_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_quant(quant_intra_c, quant_intra_3dne, "3dne", TEST_QUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_SSE2) |
|
|
test_quant(quant_intra_c, quant_intra_sse2, "sse2", TEST_QUANT_INTRA, flags); |
|
|
|
|
|
printf("\n--- quant inter ---\n"); |
|
|
test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(quant_inter_c, quant_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_quant(quant_inter_c, quant_inter_3dne, "3dne", TEST_QUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_SSE2) |
|
|
test_quant(quant_inter_c, quant_inter_sse2, "sse2", TEST_QUANT_INTER, flags); |
|
|
|
|
|
printf("\n--- dequant intra ---\n"); |
|
|
test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(dequant_intra_c, dequant_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_MMXEXT) |
|
|
test_quant(dequant_intra_c, dequant_intra_xmm, "xmm", TEST_DEQUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_quant(dequant_intra_c, dequant_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_SSE2) |
|
|
test_quant(dequant_intra_c, dequant_intra_sse2, "sse2", TEST_DEQUANT_INTRA, flags); |
|
|
|
|
|
printf("\n--- dequant inter ---\n"); |
|
|
test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(dequant_inter_c, dequant_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_MMXEXT) |
|
|
test_quant(dequant_inter_c, dequant_inter_xmm, "xmm", TEST_DEQUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_quant(dequant_inter_c, dequant_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_SSE2) |
|
|
test_quant(dequant_inter_c, dequant_inter_sse2, "sse2", TEST_DEQUANT_INTER, flags); |
|
|
|
|
|
printf("\n--- quant4_intra ---\n"); |
|
|
test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(quant4_intra_c, quant4_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_MMXEXT) |
|
|
test_quant(quant4_intra_c, quant4_intra_xmm, "xmm", TEST_QUANT_INTRA, flags); |
|
|
|
|
|
printf("\n--- quant4_inter ---\n"); |
|
|
test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(quant4_inter_c, quant4_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_MMXEXT) |
|
|
test_quant(quant4_inter_c, quant4_inter_xmm, "xmm", TEST_QUANT_INTER, flags); |
|
|
|
|
|
printf("\n--- dequant4_intra ---\n"); |
|
|
test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(dequant4_intra_c, dequant4_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_quant(dequant4_intra_c, dequant4_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
|
|
|
|
|
printf("\n--- dequant4_inter ---\n"); |
|
|
test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_MMX) |
|
|
test_quant(dequant4_inter_c, dequant4_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
|
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
|
|
test_quant(dequant4_inter_c, dequant4_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
|
|
|
|
|
emms_mmx(); |
|
|
|
|
|
#endif |
|
|
|
|
|
return XVID_ERR_OK; |
|
|
} |
|
681 |
|
|
682 |
|
|
683 |
int |
int |
684 |
xvid_init(void *handle, |
xvid_global(void *handle, |
685 |
int opt, |
int opt, |
686 |
void *param1, |
void *param1, |
687 |
void *param2) |
void *param2) |
688 |
{ |
{ |
689 |
switch(opt) |
switch(opt) |
690 |
{ |
{ |
691 |
case XVID_INIT_INIT : |
case XVID_GBL_INIT : |
692 |
return xvid_init_init((XVID_INIT_PARAM*)param1); |
return xvid_gbl_init((xvid_gbl_init_t*)param1); |
693 |
|
|
694 |
case XVID_INIT_CONVERT : |
case XVID_GBL_INFO : |
695 |
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
return xvid_gbl_info((xvid_gbl_info_t*)param1); |
696 |
|
|
697 |
case XVID_INIT_TEST : |
case XVID_GBL_CONVERT : |
698 |
return xvid_init_test((int)param1); |
return xvid_gbl_convert((xvid_gbl_convert_t*)param1); |
699 |
|
|
700 |
default : |
default : |
701 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
719 |
void *param2) |
void *param2) |
720 |
{ |
{ |
721 |
switch (opt) { |
switch (opt) { |
|
case XVID_DEC_DECODE: |
|
|
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2); |
|
|
|
|
722 |
case XVID_DEC_CREATE: |
case XVID_DEC_CREATE: |
723 |
return decoder_create((XVID_DEC_PARAM *) param1); |
return decoder_create((xvid_dec_create_t *) param1); |
724 |
|
|
725 |
case XVID_DEC_DESTROY: |
case XVID_DEC_DESTROY: |
726 |
return decoder_destroy((DECODER *) handle); |
return decoder_destroy((DECODER *) handle); |
727 |
|
|
728 |
|
case XVID_DEC_DECODE: |
729 |
|
return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); |
730 |
|
|
731 |
default: |
default: |
732 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
733 |
} |
} |
753 |
switch (opt) { |
switch (opt) { |
754 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
755 |
|
|
756 |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
return enc_encode((Encoder *) handle, |
757 |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
(xvid_enc_frame_t *) param1, |
758 |
(XVID_ENC_STATS *) param2); |
(xvid_enc_stats_t *) param2); |
|
else |
|
|
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
|
|
(XVID_ENC_STATS *) param2); |
|
759 |
|
|
760 |
case XVID_ENC_CREATE: |
case XVID_ENC_CREATE: |
761 |
return encoder_create((XVID_ENC_PARAM *) param1); |
return enc_create((xvid_enc_create_t *) param1); |
762 |
|
|
763 |
case XVID_ENC_DESTROY: |
case XVID_ENC_DESTROY: |
764 |
return encoder_destroy((Encoder *) handle); |
return enc_destroy((Encoder *) handle); |
765 |
|
|
766 |
default: |
default: |
767 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |