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 |
* |
* |
|
****************************************************************************/ |
|
|
|
|
|
/***************************************************************************** |
|
|
* |
|
|
* 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 |
****************************************************************************/ |
****************************************************************************/ |
28 |
#include <string.h> |
#include <string.h> |
29 |
#include <time.h> |
#include <time.h> |
30 |
|
|
31 |
|
#if !defined(_WIN32) |
32 |
|
#include <unistd.h> |
33 |
|
#endif |
34 |
|
|
35 |
#include "xvid.h" |
#include "xvid.h" |
36 |
#include "decoder.h" |
#include "decoder.h" |
37 |
#include "encoder.h" |
#include "encoder.h" |
40 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
41 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
42 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
43 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
44 |
#include "utils/mbfunctions.h" |
#include "utils/mbfunctions.h" |
45 |
#include "quant/quant_h263.h" |
#include "quant/quant.h" |
|
#include "quant/quant_mpeg4.h" |
|
46 |
#include "motion/motion.h" |
#include "motion/motion.h" |
47 |
|
#include "motion/gmc.h" |
48 |
#include "motion/sad.h" |
#include "motion/sad.h" |
49 |
#include "utils/emms.h" |
#include "utils/emms.h" |
50 |
#include "utils/timer.h" |
#include "utils/timer.h" |
51 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
52 |
|
#include "image/qpel.h" |
53 |
|
#include "image/postprocessing.h" |
54 |
|
|
55 |
#if defined(ARCH_IS_IA32) |
#if defined(_DEBUG) |
56 |
|
unsigned int xvid_debug = 0; /* xvid debug mask */ |
57 |
|
#endif |
58 |
|
|
59 |
#if defined(_MSC_VER) |
#if (defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)) && defined(_MSC_VER) |
60 |
# include <windows.h> |
# include <windows.h> |
61 |
#else |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || defined(ARCH_IS_PPC) |
62 |
# include <signal.h> |
# include <signal.h> |
63 |
# include <setjmp.h> |
# include <setjmp.h> |
64 |
|
|
73 |
|
|
74 |
|
|
75 |
/* |
/* |
76 |
calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
* Calls the funcptr, and returns whether SIGILL (illegal instruction) was |
77 |
return values: |
* signalled |
78 |
-1 : could not determine |
* |
79 |
0 : SIGILL was *not* signalled |
* Return values: |
80 |
1 : SIGILL was signalled |
* -1 : could not determine |
81 |
|
* 0 : SIGILL was *not* signalled |
82 |
|
* 1 : SIGILL was signalled |
83 |
*/ |
*/ |
84 |
|
#if (defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)) && defined(_MSC_VER) |
85 |
int |
static int |
86 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
87 |
{ |
{ |
|
#if defined(_MSC_VER) |
|
88 |
_try { |
_try { |
89 |
func(); |
func(); |
90 |
} |
} _except(EXCEPTION_EXECUTE_HANDLER) { |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
|
91 |
|
|
92 |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
93 |
return 1; |
return(1); |
94 |
} |
} |
95 |
return 0; |
return(0); |
96 |
#else |
} |
97 |
|
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || defined(ARCH_IS_PPC) |
98 |
|
static int |
99 |
|
sigill_check(void (*func)()) |
100 |
|
{ |
101 |
void * old_handler; |
void * old_handler; |
102 |
int jmpret; |
int jmpret; |
103 |
|
|
104 |
|
/* Set our SIGILL handler */ |
105 |
old_handler = signal(SIGILL, sigill_handler); |
old_handler = signal(SIGILL, sigill_handler); |
106 |
if (old_handler == SIG_ERR) |
|
107 |
{ |
/* Check for error */ |
108 |
return -1; |
if (old_handler == SIG_ERR) { |
109 |
|
return(-1); |
110 |
} |
} |
111 |
|
|
112 |
|
/* Save stack context, so if func triggers a SIGILL, we can still roll |
113 |
|
* back to a valid CPU state */ |
114 |
jmpret = setjmp(mark); |
jmpret = setjmp(mark); |
115 |
if (jmpret == 0) |
|
116 |
{ |
/* If setjmp returned directly, then its returned value is 0, and we still |
117 |
|
* have to test the passed func. Otherwise it means the stack context has |
118 |
|
* been restored by a longjmp() call, which in our case happens only in the |
119 |
|
* signal handler */ |
120 |
|
if (jmpret == 0) { |
121 |
func(); |
func(); |
122 |
} |
} |
123 |
|
|
124 |
|
/* Restore old signal handler */ |
125 |
signal(SIGILL, old_handler); |
signal(SIGILL, old_handler); |
126 |
|
|
127 |
return jmpret; |
return(jmpret); |
|
#endif |
|
128 |
} |
} |
129 |
#endif |
#endif |
130 |
|
|
131 |
|
|
132 |
/* detect cpu flags */ |
/* detect cpu flags */ |
133 |
static unsigned int |
static unsigned int |
134 |
detect_cpu_flags() |
detect_cpu_flags(void) |
135 |
{ |
{ |
136 |
/* enable native assembly optimizations by default */ |
/* enable native assembly optimizations by default */ |
137 |
unsigned int cpu_flags = XVID_CPU_ASM; |
unsigned int cpu_flags = XVID_CPU_ASM; |
138 |
|
|
139 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
140 |
cpu_flags |= check_cpu_features(); |
cpu_flags |= check_cpu_features(); |
141 |
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
142 |
cpu_flags &= ~XVID_CPU_SSE; |
cpu_flags &= ~XVID_CPU_SSE; |
143 |
|
|
144 |
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
if ((cpu_flags & (XVID_CPU_SSE2|XVID_CPU_SSE3|XVID_CPU_SSE41)) && sigill_check(sse2_os_trigger)) |
145 |
cpu_flags &= ~XVID_CPU_SSE2; |
cpu_flags &= ~(XVID_CPU_SSE2|XVID_CPU_SSE3|XVID_CPU_SSE41); |
146 |
#endif |
#endif |
147 |
|
|
148 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
149 |
#if defined(ARCH_IS_PPC_ALTIVEC) |
if (!sigill_check(altivec_trigger)) |
150 |
cpu_flags |= XVID_CPU_ALTIVEC; |
cpu_flags |= XVID_CPU_ALTIVEC; |
151 |
#endif |
#endif |
|
#endif |
|
152 |
|
|
153 |
return cpu_flags; |
return cpu_flags; |
154 |
} |
} |
170 |
|
|
171 |
|
|
172 |
static |
static |
173 |
int xvid_init_init(XVID_INIT_PARAM * init_param) |
int xvid_gbl_init(xvid_gbl_init_t * init) |
174 |
{ |
{ |
175 |
int cpu_flags; |
unsigned 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)) |
|
|
{ |
|
|
init_param->cpu_flags = cpu_flags; |
|
|
return XVID_ERR_OK; |
|
|
} |
|
176 |
|
|
177 |
init_param->cpu_flags = cpu_flags; |
if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */ |
178 |
|
return XVID_ERR_VERSION; |
179 |
|
|
180 |
|
cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
181 |
|
|
182 |
/* Initialize the function pointers */ |
/* Initialize the function pointers */ |
183 |
idct_int32_init(); |
idct_int32_init(); |
188 |
idct = idct_int32; |
idct = idct_int32; |
189 |
|
|
190 |
/* Only needed on PPC Altivec archs */ |
/* Only needed on PPC Altivec archs */ |
191 |
sadInit = 0; |
sadInit = NULL; |
192 |
|
|
193 |
/* Restore FPU context : emms_c is a nop functions */ |
/* Restore FPU context : emms_c is a nop functions */ |
194 |
emms = emms_c; |
emms = emms_c; |
195 |
|
|
196 |
|
/* Qpel stuff */ |
197 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_C; |
198 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_C; |
199 |
|
xvid_Init_QP(); |
200 |
|
|
201 |
/* Quantization functions */ |
/* Quantization functions */ |
202 |
quant_intra = quant_intra_c; |
quant_h263_intra = quant_h263_intra_c; |
203 |
dequant_intra = dequant_intra_c; |
quant_h263_inter = quant_h263_inter_c; |
204 |
quant_inter = quant_inter_c; |
dequant_h263_intra = dequant_h263_intra_c; |
205 |
dequant_inter = dequant_inter_c; |
dequant_h263_inter = dequant_h263_inter_c; |
206 |
|
|
207 |
quant4_intra = quant4_intra_c; |
quant_mpeg_intra = quant_mpeg_intra_c; |
208 |
dequant4_intra = dequant4_intra_c; |
quant_mpeg_inter = quant_mpeg_inter_c; |
209 |
quant4_inter = quant4_inter_c; |
dequant_mpeg_intra = dequant_mpeg_intra_c; |
210 |
dequant4_inter = dequant4_inter_c; |
dequant_mpeg_inter = dequant_mpeg_inter_c; |
211 |
|
|
212 |
/* Block transfer related functions */ |
/* Block transfer related functions */ |
213 |
transfer_8to16copy = transfer_8to16copy_c; |
transfer_8to16copy = transfer_8to16copy_c; |
215 |
transfer_8to16sub = transfer_8to16sub_c; |
transfer_8to16sub = transfer_8to16sub_c; |
216 |
transfer_8to16subro = transfer_8to16subro_c; |
transfer_8to16subro = transfer_8to16subro_c; |
217 |
transfer_8to16sub2 = transfer_8to16sub2_c; |
transfer_8to16sub2 = transfer_8to16sub2_c; |
218 |
|
transfer_8to16sub2ro = transfer_8to16sub2ro_c; |
219 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
220 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
221 |
|
transfer8x4_copy = transfer8x4_copy_c; |
222 |
|
|
223 |
/* Interlacing functions */ |
/* Interlacing functions */ |
224 |
MBFieldTest = MBFieldTest_c; |
MBFieldTest = MBFieldTest_c; |
228 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
229 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
230 |
|
|
231 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_c; |
232 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_c; |
233 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_c; |
234 |
|
|
235 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_c; |
236 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_c; |
237 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_c; |
238 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_c; |
239 |
|
|
240 |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
241 |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
242 |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
251 |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
252 |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
253 |
|
|
254 |
/* reduced resoltuion */ |
/* postprocessing */ |
255 |
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; |
|
256 |
|
|
257 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
258 |
colorspace_init(); |
colorspace_init(); |
261 |
yv12_to_yv12 = yv12_to_yv12_c; |
yv12_to_yv12 = yv12_to_yv12_c; |
262 |
rgb555_to_yv12 = rgb555_to_yv12_c; |
rgb555_to_yv12 = rgb555_to_yv12_c; |
263 |
rgb565_to_yv12 = rgb565_to_yv12_c; |
rgb565_to_yv12 = rgb565_to_yv12_c; |
264 |
|
rgb_to_yv12 = rgb_to_yv12_c; |
265 |
bgr_to_yv12 = bgr_to_yv12_c; |
bgr_to_yv12 = bgr_to_yv12_c; |
266 |
bgra_to_yv12 = bgra_to_yv12_c; |
bgra_to_yv12 = bgra_to_yv12_c; |
267 |
abgr_to_yv12 = abgr_to_yv12_c; |
abgr_to_yv12 = abgr_to_yv12_c; |
268 |
rgba_to_yv12 = rgba_to_yv12_c; |
rgba_to_yv12 = rgba_to_yv12_c; |
269 |
|
argb_to_yv12 = argb_to_yv12_c; |
270 |
yuyv_to_yv12 = yuyv_to_yv12_c; |
yuyv_to_yv12 = yuyv_to_yv12_c; |
271 |
uyvy_to_yv12 = uyvy_to_yv12_c; |
uyvy_to_yv12 = uyvy_to_yv12_c; |
272 |
|
|
276 |
bgrai_to_yv12 = bgrai_to_yv12_c; |
bgrai_to_yv12 = bgrai_to_yv12_c; |
277 |
abgri_to_yv12 = abgri_to_yv12_c; |
abgri_to_yv12 = abgri_to_yv12_c; |
278 |
rgbai_to_yv12 = rgbai_to_yv12_c; |
rgbai_to_yv12 = rgbai_to_yv12_c; |
279 |
|
argbi_to_yv12 = argbi_to_yv12_c; |
280 |
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
281 |
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
282 |
|
|
|
|
|
283 |
/* All colorspace transformation functions YV12->User format */ |
/* All colorspace transformation functions YV12->User format */ |
284 |
yv12_to_rgb555 = yv12_to_rgb555_c; |
yv12_to_rgb555 = yv12_to_rgb555_c; |
285 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
286 |
|
yv12_to_rgb = yv12_to_rgb_c; |
287 |
yv12_to_bgr = yv12_to_bgr_c; |
yv12_to_bgr = yv12_to_bgr_c; |
288 |
yv12_to_bgra = yv12_to_bgra_c; |
yv12_to_bgra = yv12_to_bgra_c; |
289 |
yv12_to_abgr = yv12_to_abgr_c; |
yv12_to_abgr = yv12_to_abgr_c; |
290 |
yv12_to_rgba = yv12_to_rgba_c; |
yv12_to_rgba = yv12_to_rgba_c; |
291 |
|
yv12_to_argb = yv12_to_argb_c; |
292 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
293 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
294 |
|
|
298 |
yv12_to_bgrai = yv12_to_bgrai_c; |
yv12_to_bgrai = yv12_to_bgrai_c; |
299 |
yv12_to_abgri = yv12_to_abgri_c; |
yv12_to_abgri = yv12_to_abgri_c; |
300 |
yv12_to_rgbai = yv12_to_rgbai_c; |
yv12_to_rgbai = yv12_to_rgbai_c; |
301 |
|
yv12_to_argbi = yv12_to_argbi_c; |
302 |
yv12_to_yuyvi = yv12_to_yuyvi_c; |
yv12_to_yuyvi = yv12_to_yuyvi_c; |
303 |
yv12_to_uyvyi = yv12_to_uyvyi_c; |
yv12_to_uyvyi = yv12_to_uyvyi_c; |
304 |
|
|
310 |
sad8bi = sad8bi_c; |
sad8bi = sad8bi_c; |
311 |
dev16 = dev16_c; |
dev16 = dev16_c; |
312 |
sad16v = sad16v_c; |
sad16v = sad16v_c; |
313 |
|
sse8_16bit = sse8_16bit_c; |
314 |
|
sse8_8bit = sse8_8bit_c; |
315 |
|
|
316 |
// Halfpel8_Refine = Halfpel8_Refine_c; |
init_GMC(cpu_flags); |
317 |
|
|
318 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
|
|
|
|
if ((cpu_flags & XVID_CPU_ASM)) |
|
|
{ |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
} |
|
319 |
|
|
320 |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
321 |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
322 |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2) || |
323 |
|
(cpu_flags & XVID_CPU_SSE3) || (cpu_flags & XVID_CPU_SSE41)) |
324 |
{ |
{ |
325 |
/* Restore FPU context : emms_c is a nop functions */ |
/* Restore FPU context : emms_c is a nop functions */ |
326 |
emms = emms_mmx; |
emms = emms_mmx; |
329 |
if ((cpu_flags & XVID_CPU_MMX)) { |
if ((cpu_flags & XVID_CPU_MMX)) { |
330 |
|
|
331 |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
332 |
fdct = fdct_mmx; |
fdct = fdct_mmx_skal; |
333 |
idct = idct_mmx; |
idct = idct_mmx; |
334 |
|
|
335 |
|
/* Qpel stuff */ |
336 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_mmx; |
337 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx; |
338 |
|
|
339 |
/* Quantization related functions */ |
/* Quantization related functions */ |
340 |
quant_intra = quant_intra_mmx; |
quant_h263_intra = quant_h263_intra_mmx; |
341 |
dequant_intra = dequant_intra_mmx; |
quant_h263_inter = quant_h263_inter_mmx; |
342 |
quant_inter = quant_inter_mmx; |
dequant_h263_intra = dequant_h263_intra_mmx; |
343 |
dequant_inter = dequant_inter_mmx; |
dequant_h263_inter = dequant_h263_inter_mmx; |
344 |
|
quant_mpeg_intra = quant_mpeg_intra_mmx; |
345 |
quant4_intra = quant4_intra_mmx; |
quant_mpeg_inter = quant_mpeg_inter_mmx; |
346 |
dequant4_intra = dequant4_intra_mmx; |
dequant_mpeg_intra = dequant_mpeg_intra_mmx; |
347 |
quant4_inter = quant4_inter_mmx; |
dequant_mpeg_inter = dequant_mpeg_inter_mmx; |
348 |
dequant4_inter = dequant4_inter_mmx; |
|
349 |
|
|
350 |
/* Block related functions */ |
/* Block related functions */ |
351 |
transfer_8to16copy = transfer_8to16copy_mmx; |
transfer_8to16copy = transfer_8to16copy_mmx; |
355 |
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
transfer_8to16sub2 = transfer_8to16sub2_mmx; |
356 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
357 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
358 |
|
transfer8x4_copy = transfer8x4_copy_mmx; |
359 |
|
|
360 |
/* Interlacing Functions */ |
/* Interlacing Functions */ |
361 |
MBFieldTest = MBFieldTest_mmx; |
MBFieldTest = MBFieldTest_mmx; |
365 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
366 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
367 |
|
|
368 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_mmx; |
369 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_mmx; |
370 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_mmx; |
371 |
|
|
372 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_mmx; |
373 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_mmx; |
374 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_mmx; |
375 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_mmx; |
376 |
|
|
377 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
378 |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
379 |
|
|
380 |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
381 |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
382 |
|
|
383 |
/* reduced resolution */ |
/* postprocessing */ |
384 |
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; |
|
385 |
|
|
386 |
/* image input xxx_to_yv12 related functions */ |
/* image input xxx_to_yv12 related functions */ |
387 |
|
|
388 |
yv12_to_yv12 = yv12_to_yv12_mmx; |
yv12_to_yv12 = yv12_to_yv12_mmx; |
389 |
|
|
390 |
bgr_to_yv12 = bgr_to_yv12_mmx; |
bgr_to_yv12 = bgr_to_yv12_mmx; |
391 |
|
rgb_to_yv12 = rgb_to_yv12_mmx; |
392 |
bgra_to_yv12 = bgra_to_yv12_mmx; |
bgra_to_yv12 = bgra_to_yv12_mmx; |
393 |
|
rgba_to_yv12 = rgba_to_yv12_mmx; |
394 |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
395 |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
396 |
|
|
411 |
sad8bi = sad8bi_mmx; |
sad8bi = sad8bi_mmx; |
412 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
413 |
sad16v = sad16v_mmx; |
sad16v = sad16v_mmx; |
414 |
|
sse8_16bit = sse8_16bit_mmx; |
415 |
|
sse8_8bit = sse8_8bit_mmx; |
416 |
} |
} |
417 |
|
|
418 |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
426 |
|
|
427 |
yuyv_to_yv12 = yuyv_to_yv12_3dn; |
yuyv_to_yv12 = yuyv_to_yv12_3dn; |
428 |
uyvy_to_yv12 = uyvy_to_yv12_3dn; |
uyvy_to_yv12 = uyvy_to_yv12_3dn; |
429 |
|
|
430 |
} |
} |
431 |
|
|
432 |
|
|
433 |
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
434 |
|
|
435 |
/* Inverse DCT */ |
/* DCT */ |
436 |
|
fdct = fdct_xmm_skal; |
437 |
idct = idct_xmm; |
idct = idct_xmm; |
438 |
|
|
439 |
/* Interpolation */ |
/* Interpolation */ |
441 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
442 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
443 |
|
|
444 |
/* reduced resolution */ |
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_xmm; |
445 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_xmm; |
446 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_xmm; |
447 |
|
|
448 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_xmm; |
449 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm; |
450 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm; |
451 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm; |
452 |
|
|
453 |
/* Quantization */ |
/* Quantization */ |
454 |
quant4_intra = quant4_intra_xmm; |
quant_mpeg_inter = quant_mpeg_inter_xmm; |
|
quant4_inter = quant4_inter_xmm; |
|
455 |
|
|
456 |
dequant_intra = dequant_intra_xmm; |
dequant_h263_intra = dequant_h263_intra_xmm; |
457 |
dequant_inter = dequant_inter_xmm; |
dequant_h263_inter = dequant_h263_inter_xmm; |
458 |
|
|
459 |
/* Buffer transfer */ |
/* Buffer transfer */ |
460 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
461 |
|
transfer_8to16sub2ro = transfer_8to16sub2ro_xmm; |
462 |
|
|
463 |
/* Colorspace transformation */ |
/* Colorspace transformation */ |
464 |
yv12_to_yv12 = yv12_to_yv12_xmm; |
/* yv12_to_yv12 = yv12_to_yv12_xmm; */ /* appears to be slow on many machines */ |
465 |
yuyv_to_yv12 = yuyv_to_yv12_xmm; |
yuyv_to_yv12 = yuyv_to_yv12_xmm; |
466 |
uyvy_to_yv12 = uyvy_to_yv12_xmm; |
uyvy_to_yv12 = uyvy_to_yv12_xmm; |
467 |
|
|
480 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
481 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; |
482 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
483 |
|
|
484 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dn; |
485 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dn; |
486 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dn; |
487 |
} |
} |
488 |
|
|
489 |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
490 |
|
|
|
/* Inverse DCT */ |
|
|
idct = idct_3dne; |
|
|
|
|
491 |
/* Buffer transfer */ |
/* Buffer transfer */ |
492 |
transfer_8to16copy = transfer_8to16copy_3dne; |
transfer_8to16copy = transfer_8to16copy_3dne; |
493 |
transfer_16to8copy = transfer_16to8copy_3dne; |
transfer_16to8copy = transfer_16to8copy_3dne; |
494 |
transfer_8to16sub = transfer_8to16sub_3dne; |
transfer_8to16sub = transfer_8to16sub_3dne; |
495 |
transfer_8to16subro = transfer_8to16subro_3dne; |
transfer_8to16subro = transfer_8to16subro_3dne; |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
|
496 |
transfer_16to8add = transfer_16to8add_3dne; |
transfer_16to8add = transfer_16to8add_3dne; |
497 |
transfer8x8_copy = transfer8x8_copy_3dne; |
transfer8x8_copy = transfer8x8_copy_3dne; |
498 |
|
transfer8x4_copy = transfer8x4_copy_3dne; |
499 |
|
|
500 |
|
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
501 |
|
/* Inverse DCT */ |
502 |
|
idct = idct_3dne; |
503 |
|
|
504 |
|
/* Buffer transfer */ |
505 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
506 |
|
|
507 |
|
/* Interpolation */ |
508 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
509 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
510 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
511 |
|
|
512 |
|
interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dne; |
513 |
|
interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dne; |
514 |
|
interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dne; |
515 |
|
|
516 |
/* Quantization */ |
/* Quantization */ |
517 |
dequant4_intra = dequant4_intra_3dne; |
quant_h263_intra = quant_h263_intra_3dne; /* cmov only */ |
518 |
dequant4_inter = dequant4_inter_3dne; |
quant_h263_inter = quant_h263_inter_3dne; |
519 |
quant_intra = quant_intra_3dne; |
dequant_mpeg_intra = dequant_mpeg_intra_3dne; /* cmov only */ |
520 |
quant_inter = quant_inter_3dne; |
dequant_mpeg_inter = dequant_mpeg_inter_3dne; |
521 |
dequant_intra = dequant_intra_3dne; |
dequant_h263_intra = dequant_h263_intra_3dne; |
522 |
dequant_inter = dequant_inter_3dne; |
dequant_h263_inter = dequant_h263_inter_3dne; |
523 |
|
|
524 |
/* ME functions */ |
/* ME functions */ |
|
calc_cbp = calc_cbp_3dne; |
|
525 |
sad16 = sad16_3dne; |
sad16 = sad16_3dne; |
526 |
sad8 = sad8_3dne; |
sad8 = sad8_3dne; |
527 |
sad16bi = sad16bi_3dne; |
sad16bi = sad16bi_3dne; |
528 |
sad8bi = sad8bi_3dne; |
sad8bi = sad8bi_3dne; |
529 |
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; |
|
530 |
} |
} |
531 |
|
} |
532 |
|
|
533 |
if ((cpu_flags & XVID_CPU_SSE2)) { |
if ((cpu_flags & XVID_CPU_SSE2)) { |
534 |
|
|
535 |
calc_cbp = calc_cbp_sse2; |
calc_cbp = calc_cbp_sse2; |
536 |
|
|
537 |
/* Quantization */ |
/* Quantization */ |
538 |
quant_intra = quant_intra_sse2; |
quant_h263_intra = quant_h263_intra_sse2; |
539 |
dequant_intra = dequant_intra_sse2; |
quant_h263_inter = quant_h263_inter_sse2; |
540 |
quant_inter = quant_inter_sse2; |
dequant_h263_intra = dequant_h263_intra_sse2; |
541 |
dequant_inter = dequant_inter_sse2; |
dequant_h263_inter = dequant_h263_inter_sse2; |
542 |
|
|
543 |
#if defined(EXPERIMENTAL_SSE2_CODE) |
/* SAD operators */ |
|
/* ME; slower than xmm */ |
|
544 |
sad16 = sad16_sse2; |
sad16 = sad16_sse2; |
545 |
dev16 = dev16_sse2; |
dev16 = dev16_sse2; |
546 |
#endif |
|
547 |
/* Forward and Inverse DCT */ |
/* DCT operators */ |
548 |
idct = idct_sse2; |
fdct = fdct_sse2_skal; |
549 |
fdct = fdct_sse2; |
idct = idct_sse2_skal; /* Is now IEEE1180 and Walken compliant. */ |
550 |
|
|
551 |
|
/* postprocessing */ |
552 |
|
image_brightness = image_brightness_sse2; |
553 |
|
|
554 |
} |
} |
555 |
#endif |
|
556 |
|
if ((cpu_flags & XVID_CPU_SSE3)) { |
557 |
|
|
558 |
|
/* SAD operators */ |
559 |
|
sad16 = sad16_sse3; |
560 |
|
dev16 = dev16_sse3; |
561 |
|
} |
562 |
|
|
563 |
|
#endif /* ARCH_IS_IA32 */ |
564 |
|
|
565 |
#if defined(ARCH_IS_IA64) |
#if defined(ARCH_IS_IA64) |
566 |
if ((cpu_flags & XVID_CPU_ASM)) { //use assembler routines? |
if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
567 |
idct_ia64_init(); |
idct_ia64_init(); |
568 |
fdct = fdct_ia64; |
fdct = fdct_ia64; |
569 |
idct = idct_ia64; //not yet working, crashes |
idct = idct_ia64; /*not yet working, crashes */ |
570 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
571 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; |
572 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; |
574 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
575 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
576 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
577 |
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
/* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
578 |
quant_intra = quant_intra_ia64; |
quant_h263_intra = quant_h263_intra_ia64; |
579 |
dequant_intra = dequant_intra_ia64; |
quant_h263_inter = quant_h263_inter_ia64; |
580 |
quant_inter = quant_inter_ia64; |
dequant_h263_intra = dequant_h263_intra_ia64; |
581 |
dequant_inter = dequant_inter_ia64; |
dequant_h263_inter = dequant_h263_inter_ia64; |
582 |
transfer_8to16copy = transfer_8to16copy_ia64; |
transfer_8to16copy = transfer_8to16copy_ia64; |
583 |
transfer_16to8copy = transfer_16to8copy_ia64; |
transfer_16to8copy = transfer_16to8copy_ia64; |
584 |
transfer_8to16sub = transfer_8to16sub_ia64; |
transfer_8to16sub = transfer_8to16sub_ia64; |
585 |
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
transfer_8to16sub2 = transfer_8to16sub2_ia64; |
586 |
transfer_16to8add = transfer_16to8add_ia64; |
transfer_16to8add = transfer_16to8add_ia64; |
587 |
transfer8x8_copy = transfer8x8_copy_ia64; |
transfer8x8_copy = transfer8x8_copy_ia64; |
|
DEBUG("Using IA-64 assembler routines.\n"); |
|
588 |
} |
} |
589 |
#endif |
#endif |
590 |
|
|
591 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
592 |
if ((cpu_flags & XVID_CPU_ASM)) |
if ((cpu_flags & XVID_CPU_ALTIVEC)) { |
593 |
{ |
/* sad operators */ |
594 |
calc_cbp = calc_cbp_ppc; |
sad16 = sad16_altivec_c; |
595 |
} |
sad16bi = sad16bi_altivec_c; |
596 |
|
sad8 = sad8_altivec_c; |
597 |
if ((cpu_flags & XVID_CPU_ALTIVEC)) |
dev16 = dev16_altivec_c; |
598 |
{ |
|
599 |
calc_cbp = calc_cbp_altivec; |
sse8_16bit = sse8_16bit_altivec_c; |
600 |
fdct = fdct_altivec; |
|
601 |
idct = idct_altivec; |
/* mem transfer */ |
602 |
sadInit = sadInit_altivec; |
transfer_8to16copy = transfer_8to16copy_altivec_c; |
603 |
sad16 = sad16_altivec; |
transfer_16to8copy = transfer_16to8copy_altivec_c; |
604 |
sad8 = sad8_altivec; |
transfer_8to16sub = transfer_8to16sub_altivec_c; |
605 |
dev16 = dev16_altivec; |
transfer_8to16subro = transfer_8to16subro_altivec_c; |
606 |
} |
transfer_8to16sub2 = transfer_8to16sub2_altivec_c; |
607 |
#endif |
transfer_16to8add = transfer_16to8add_altivec_c; |
608 |
|
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; |
|
|
} |
|
609 |
|
|
610 |
|
/* Inverse DCT */ |
611 |
|
idct = idct_altivec_c; |
612 |
|
|
613 |
|
/* Interpolation */ |
614 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c; |
615 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c; |
616 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c; |
617 |
|
|
618 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; |
619 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; |
620 |
|
|
621 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c; |
622 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c; |
623 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c; |
624 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c; |
625 |
|
|
626 |
|
/* Colorspace conversion */ |
627 |
|
bgra_to_yv12 = bgra_to_yv12_altivec_c; |
628 |
|
abgr_to_yv12 = abgr_to_yv12_altivec_c; |
629 |
|
rgba_to_yv12 = rgba_to_yv12_altivec_c; |
630 |
|
argb_to_yv12 = argb_to_yv12_altivec_c; |
631 |
|
|
632 |
void fill8(uint8_t * block, int size, int value) |
yuyv_to_yv12 = yuyv_to_yv12_altivec_c; |
633 |
{ |
uyvy_to_yv12 = uyvy_to_yv12_altivec_c; |
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = value; |
|
|
} |
|
634 |
|
|
635 |
void fill16(int16_t * block, int size, int value) |
yv12_to_yuyv = yv12_to_yuyv_altivec_c; |
636 |
{ |
yv12_to_uyvy = yv12_to_uyvy_altivec_c; |
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = value; |
|
|
} |
|
637 |
|
|
638 |
#define RANDOM(min,max) min + (rand() % (max-min)) |
/* Quantization */ |
639 |
|
quant_h263_intra = quant_h263_intra_altivec_c; |
640 |
|
quant_h263_inter = quant_h263_inter_altivec_c; |
641 |
|
dequant_h263_intra = dequant_h263_intra_altivec_c; |
642 |
|
dequant_h263_inter = dequant_h263_inter_altivec_c; |
643 |
|
|
644 |
void random8(uint8_t * block, int size, int min, int max) |
dequant_mpeg_intra = dequant_mpeg_intra_altivec_c; |
645 |
{ |
dequant_mpeg_inter = dequant_mpeg_inter_altivec_c; |
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = RANDOM(min,max); |
|
|
} |
|
646 |
|
|
647 |
void random16(int16_t * block, int size, int min, int max) |
/* Qpel stuff */ |
648 |
{ |
xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C; |
649 |
int i; |
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C; |
|
for (i = 0; i < size; i++) |
|
|
block[i] = RANDOM(min,max); |
|
650 |
} |
} |
651 |
|
#endif |
652 |
|
|
653 |
int compare16(const int16_t * blockA, const int16_t * blockB, int size) |
#if defined(_DEBUG) |
654 |
{ |
xvid_debug = init->debug; |
655 |
int i; |
#endif |
|
for (i = 0; i < size; i++) |
|
|
if (blockA[i] != blockB[i]) |
|
|
return 1; |
|
|
|
|
|
return 0; |
|
|
} |
|
656 |
|
|
657 |
int diff16(const int16_t * blockA, const int16_t * blockB, int size) |
return(0); |
|
{ |
|
|
int i, diff = 0; |
|
|
for (i = 0; i < size; i++) |
|
|
diff += ABS(blockA[i]-blockB[i]); |
|
|
return diff; |
|
658 |
} |
} |
659 |
|
|
660 |
|
|
661 |
#define XVID_TEST_RANDOM 0x00000001 /* random input data */ |
static int |
662 |
#define XVID_TEST_VERBOSE 0x00000002 /* verbose error output */ |
xvid_gbl_info(xvid_gbl_info_t * info) |
|
|
|
|
|
|
|
#define TEST_FORWARD 0x00000001 /* intra */ |
|
|
#define TEST_FDCT (TEST_FORWARD) |
|
|
#define TEST_IDCT (0) |
|
|
|
|
|
static 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 */ |
|
663 |
{ |
{ |
664 |
min = -256; |
if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */ |
665 |
max = 255; |
return XVID_ERR_VERSION; |
|
}else{ /* inverse */ |
|
|
min = -2048; |
|
|
max = 2047; |
|
|
} |
|
666 |
|
|
667 |
for (i = 0; i < 64*64; i++) |
info->actual_version = XVID_VERSION; |
668 |
{ |
info->build = "xvid-1.3.0-dev"; |
669 |
if ((flags & XVID_TEST_RANDOM)) |
info->cpu_flags = detect_cpu_flags(); |
670 |
{ |
info->num_threads = 0; |
|
random16(arrayA, 64, min, max); |
|
|
}else{ |
|
|
fill16(arrayA, 64, i); |
|
|
} |
|
|
memcpy(arrayB, arrayA, 64*sizeof(int16_t)); |
|
671 |
|
|
672 |
if ((test & TEST_FORWARD)) |
#if defined(_WIN32) |
|
{ |
|
|
timeSTART = read_counter(); |
|
|
((fdctFunc*)funcA)(arrayA); |
|
|
timeA += read_counter() - timeSTART; |
|
673 |
|
|
|
timeSTART = read_counter(); |
|
|
((fdctFunc*)funcB)(arrayB); |
|
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
else |
|
674 |
{ |
{ |
675 |
timeSTART = read_counter(); |
SYSTEM_INFO siSysInfo; |
676 |
((idctFunc*)funcA)(arrayA); |
GetSystemInfo(&siSysInfo); |
677 |
timeA += read_counter() - timeSTART; |
info->num_threads = siSysInfo.dwNumberOfProcessors; /* number of _logical_ cores */ |
|
|
|
|
timeSTART = read_counter(); |
|
|
((idctFunc*)funcB)(arrayB); |
|
|
timeB += read_counter() - timeSTART; |
|
678 |
} |
} |
679 |
|
|
680 |
tmp = diff16(arrayA, arrayB, 64) / 64; |
#else |
|
if (tmp > max_error) |
|
|
max_error = tmp; |
|
|
if (tmp < min_error) |
|
|
min_error = tmp; |
|
681 |
|
|
682 |
count++; |
info->num_threads = sysconf(_SC_NPROCESSORS_CONF); |
|
} |
|
683 |
|
|
684 |
/* print the "average difference" of best/worst transforms */ |
#endif |
|
printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error); |
|
685 |
|
|
686 |
return 0; |
return 0; |
687 |
} |
} |
688 |
|
|
689 |
|
|
690 |
#define TEST_QUANT 0x00000001 /* forward quantization */ |
static int |
691 |
#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) |
|
|
|
|
|
static 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 */ |
|
692 |
{ |
{ |
693 |
timeSTART = read_counter(); |
int width; |
694 |
((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q); |
int height; |
695 |
timeA += read_counter() - timeSTART; |
int width2; |
696 |
|
int height2; |
697 |
|
IMAGE img; |
698 |
|
|
699 |
timeSTART = read_counter(); |
if (XVID_VERSION_MAJOR(convert->version) != 1) /* v1.x.x */ |
700 |
((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; |
|
701 |
|
|
702 |
timeSTART = read_counter(); |
#if 0 |
703 |
retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q); |
const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
704 |
timeB += read_counter() - timeSTART; |
#endif |
705 |
} |
width = convert->width; |
706 |
|
height = convert->height; |
707 |
|
width2 = convert->width/2; |
708 |
|
height2 = convert->height/2; |
709 |
|
|
710 |
/* 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)) |
|
711 |
{ |
{ |
712 |
errors++; |
case XVID_CSP_YV12 : |
713 |
if ((flags & XVID_TEST_VERBOSE)) |
img.y = convert->input.plane[0]; |
714 |
printf("%s error: q=%i, i=%i\n", nameB, q, i); |
img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; |
715 |
} |
img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; |
716 |
|
image_output(&img, width, height, width, |
717 |
|
(uint8_t**)convert->output.plane, convert->output.stride, |
718 |
|
convert->output.csp, convert->interlacing); |
719 |
|
break; |
720 |
|
|
721 |
count++; |
default : |
722 |
} |
return XVID_ERR_FORMAT; |
723 |
} |
} |
724 |
|
|
|
printf("%s:\t%i", nameB, (int)(timeB / count)); |
|
|
if (errors>0) |
|
|
printf("\t(%i errors out of %i)", errors, count); |
|
|
printf("\n"); |
|
725 |
|
|
726 |
|
emms(); |
727 |
return 0; |
return 0; |
728 |
} |
} |
729 |
|
|
730 |
|
/***************************************************************************** |
731 |
|
* XviD Global Entry point |
732 |
int xvid_init_test(int flags) |
* |
733 |
{ |
* Well this function initialize all internal function pointers according |
734 |
int cpu_flags; |
* to the CPU features forced by the library client or autodetected (depending |
735 |
|
* on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all |
736 |
srand(time(0)); |
* image colorspace transformation tables. |
737 |
|
* |
738 |
printf("xvid_init_test\n"); |
****************************************************************************/ |
|
|
|
|
#if defined(ARCH_IS_IA32) |
|
|
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; |
|
|
} |
|
739 |
|
|
740 |
|
|
741 |
int |
int |
742 |
xvid_init(void *handle, |
xvid_global(void *handle, |
743 |
int opt, |
int opt, |
744 |
void *param1, |
void *param1, |
745 |
void *param2) |
void *param2) |
746 |
{ |
{ |
747 |
switch(opt) |
switch(opt) |
748 |
{ |
{ |
749 |
case XVID_INIT_INIT : |
case XVID_GBL_INIT : |
750 |
return xvid_init_init((XVID_INIT_PARAM*)param1); |
return xvid_gbl_init((xvid_gbl_init_t*)param1); |
751 |
|
|
752 |
case XVID_INIT_CONVERT : |
case XVID_GBL_INFO : |
753 |
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
return xvid_gbl_info((xvid_gbl_info_t*)param1); |
754 |
|
|
755 |
case XVID_INIT_TEST : |
case XVID_GBL_CONVERT : |
756 |
return xvid_init_test((int)param1); |
return xvid_gbl_convert((xvid_gbl_convert_t*)param1); |
757 |
|
|
758 |
default : |
default : |
759 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
777 |
void *param2) |
void *param2) |
778 |
{ |
{ |
779 |
switch (opt) { |
switch (opt) { |
|
case XVID_DEC_DECODE: |
|
|
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2); |
|
|
|
|
780 |
case XVID_DEC_CREATE: |
case XVID_DEC_CREATE: |
781 |
return decoder_create((XVID_DEC_PARAM *) param1); |
return decoder_create((xvid_dec_create_t *) param1); |
782 |
|
|
783 |
case XVID_DEC_DESTROY: |
case XVID_DEC_DESTROY: |
784 |
return decoder_destroy((DECODER *) handle); |
return decoder_destroy((DECODER *) handle); |
785 |
|
|
786 |
|
case XVID_DEC_DECODE: |
787 |
|
return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); |
788 |
|
|
789 |
default: |
default: |
790 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
791 |
} |
} |
811 |
switch (opt) { |
switch (opt) { |
812 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
813 |
|
|
814 |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
return enc_encode((Encoder *) handle, |
815 |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
(xvid_enc_frame_t *) param1, |
816 |
(XVID_ENC_STATS *) param2); |
(xvid_enc_stats_t *) param2); |
|
else |
|
|
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
|
|
(XVID_ENC_STATS *) param2); |
|
817 |
|
|
818 |
case XVID_ENC_CREATE: |
case XVID_ENC_CREATE: |
819 |
return encoder_create((XVID_ENC_PARAM *) param1); |
return enc_create((xvid_enc_create_t *) param1); |
820 |
|
|
821 |
case XVID_ENC_DESTROY: |
case XVID_ENC_DESTROY: |
822 |
return encoder_destroy((Encoder *) handle); |
return enc_destroy((Encoder *) handle); |
823 |
|
|
824 |
default: |
default: |
825 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |