3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Native API implementation - |
* - Native API implementation - |
5 |
* |
* |
6 |
|
* Copyright(C) 2001-2002 Peter Ross <pross@cs.rmit.edu.au> |
7 |
|
* |
8 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
9 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
10 |
* to use this software module in hardware or software products are |
* to use this software module in hardware or software products are |
30 |
* |
* |
31 |
****************************************************************************/ |
****************************************************************************/ |
32 |
|
|
|
/***************************************************************************** |
|
|
* |
|
|
* 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> |
|
|
* |
|
|
* $Id$ |
|
|
* |
|
|
****************************************************************************/ |
|
|
|
|
|
#include <stdio.h> |
|
|
#include <stdlib.h> |
|
|
#include <string.h> |
|
|
#include <time.h> |
|
|
|
|
33 |
#include "xvid.h" |
#include "xvid.h" |
34 |
#include "decoder.h" |
#include "decoder.h" |
35 |
#include "encoder.h" |
#include "encoder.h" |
38 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
39 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
40 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
41 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
|
#include "utils/mbfunctions.h" |
|
42 |
#include "quant/quant_h263.h" |
#include "quant/quant_h263.h" |
43 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_mpeg4.h" |
44 |
#include "motion/motion.h" |
#include "motion/motion.h" |
70 |
|
|
71 |
|
|
72 |
/* |
/* |
73 |
calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
* Calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
74 |
return values: |
* Return values: |
75 |
-1 : could not determine |
* -1 : could not determine |
76 |
0 : SIGILL was *not* signalled |
* 0 : SIGILL was *not* signalled |
77 |
1 : SIGILL was signalled |
* 1 : SIGILL was signalled |
78 |
*/ |
*/ |
79 |
|
|
80 |
int |
int |
128 |
* |
* |
129 |
****************************************************************************/ |
****************************************************************************/ |
130 |
|
|
131 |
|
int |
132 |
static |
xvid_init(void *handle, |
133 |
int xvid_init_init(XVID_INIT_PARAM * init_param) |
int opt, |
134 |
|
void *param1, |
135 |
|
void *param2) |
136 |
{ |
{ |
137 |
int cpu_flags; |
int cpu_flags; |
138 |
|
XVID_INIT_PARAM *init_param; |
139 |
|
|
140 |
|
init_param = (XVID_INIT_PARAM *) param1; |
141 |
|
|
142 |
/* Inform the client the API version */ |
/* Inform the client the API version */ |
143 |
init_param->api_version = API_VERSION; |
init_param->api_version = API_VERSION; |
205 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
206 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
207 |
|
|
|
/* Interlacing functions */ |
|
|
MBFieldTest = MBFieldTest_c; |
|
|
|
|
208 |
/* Image interpolation related functions */ |
/* Image interpolation related functions */ |
209 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
210 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
211 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
212 |
|
|
|
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
|
|
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
|
|
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
|
|
|
|
|
interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; |
|
|
interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; |
|
|
interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; |
|
|
|
|
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; |
|
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; |
|
|
|
|
|
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
|
|
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
|
|
|
|
|
/* reduced resoltuion */ |
|
|
|
|
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
|
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
|
|
#ifdef ARCH_X86 |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
#else |
|
|
vfilter_31 = xvid_VFilter_31_C; |
|
|
hfilter_31 = xvid_HFilter_31_C; |
|
|
#endif |
|
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
|
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
|
|
|
|
213 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
214 |
colorspace_init(); |
colorspace_init(); |
215 |
|
|
216 |
/* All colorspace transformation functions User Format->YV12 */ |
/* All colorspace transformation functions User Format->YV12 */ |
|
yv12_to_yv12 = yv12_to_yv12_c; |
|
217 |
rgb555_to_yv12 = rgb555_to_yv12_c; |
rgb555_to_yv12 = rgb555_to_yv12_c; |
218 |
rgb565_to_yv12 = rgb565_to_yv12_c; |
rgb565_to_yv12 = rgb565_to_yv12_c; |
219 |
bgr_to_yv12 = bgr_to_yv12_c; |
rgb24_to_yv12 = rgb24_to_yv12_c; |
220 |
bgra_to_yv12 = bgra_to_yv12_c; |
rgb32_to_yv12 = rgb32_to_yv12_c; |
221 |
abgr_to_yv12 = abgr_to_yv12_c; |
yuv_to_yv12 = yuv_to_yv12_c; |
|
rgba_to_yv12 = rgba_to_yv12_c; |
|
222 |
yuyv_to_yv12 = yuyv_to_yv12_c; |
yuyv_to_yv12 = yuyv_to_yv12_c; |
223 |
uyvy_to_yv12 = uyvy_to_yv12_c; |
uyvy_to_yv12 = uyvy_to_yv12_c; |
224 |
|
|
|
rgb555i_to_yv12 = rgb555i_to_yv12_c; |
|
|
rgb565i_to_yv12 = rgb565i_to_yv12_c; |
|
|
bgri_to_yv12 = bgri_to_yv12_c; |
|
|
bgrai_to_yv12 = bgrai_to_yv12_c; |
|
|
abgri_to_yv12 = abgri_to_yv12_c; |
|
|
rgbai_to_yv12 = rgbai_to_yv12_c; |
|
|
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
|
|
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
|
|
|
|
|
|
|
225 |
/* All colorspace transformation functions YV12->User format */ |
/* All colorspace transformation functions YV12->User format */ |
226 |
yv12_to_rgb555 = yv12_to_rgb555_c; |
yv12_to_rgb555 = yv12_to_rgb555_c; |
227 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
228 |
yv12_to_bgr = yv12_to_bgr_c; |
yv12_to_rgb24 = yv12_to_rgb24_c; |
229 |
yv12_to_bgra = yv12_to_bgra_c; |
yv12_to_rgb32 = yv12_to_rgb32_c; |
230 |
yv12_to_abgr = yv12_to_abgr_c; |
yv12_to_yuv = yv12_to_yuv_c; |
|
yv12_to_rgba = yv12_to_rgba_c; |
|
231 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
232 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
233 |
|
|
|
yv12_to_rgb555i = yv12_to_rgb555i_c; |
|
|
yv12_to_rgb565i = yv12_to_rgb565i_c; |
|
|
yv12_to_bgri = yv12_to_bgri_c; |
|
|
yv12_to_bgrai = yv12_to_bgrai_c; |
|
|
yv12_to_abgri = yv12_to_abgri_c; |
|
|
yv12_to_rgbai = yv12_to_rgbai_c; |
|
|
yv12_to_yuyvi = yv12_to_yuyvi_c; |
|
|
yv12_to_uyvyi = yv12_to_uyvyi_c; |
|
|
|
|
234 |
/* Functions used in motion estimation algorithms */ |
/* Functions used in motion estimation algorithms */ |
235 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
236 |
sad16 = sad16_c; |
sad16 = sad16_c; |
238 |
sad16bi = sad16bi_c; |
sad16bi = sad16bi_c; |
239 |
sad8bi = sad8bi_c; |
sad8bi = sad8bi_c; |
240 |
dev16 = dev16_c; |
dev16 = dev16_c; |
|
sad16v = sad16v_c; |
|
241 |
|
|
242 |
// Halfpel8_Refine = Halfpel8_Refine_c; |
Halfpel8_Refine = Halfpel8_Refine_c; |
243 |
|
|
244 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
|
|
|
|
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
|
|
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
|
|
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
|
|
{ |
|
|
/* Restore FPU context : emms_c is a nop functions */ |
|
|
emms = emms_mmx; |
|
|
} |
|
|
|
|
245 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
246 |
|
|
247 |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
248 |
fdct = fdct_mmx; |
fdct = fdct_mmx; |
249 |
idct = idct_mmx; |
idct = idct_mmx; |
250 |
|
|
251 |
|
/* To restore FPU context after mmx use */ |
252 |
|
emms = emms_mmx; |
253 |
|
|
254 |
/* Quantization related functions */ |
/* Quantization related functions */ |
255 |
quant_intra = quant_intra_mmx; |
quant_intra = quant_intra_mmx; |
256 |
dequant_intra = dequant_intra_mmx; |
dequant_intra = dequant_intra_mmx; |
270 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
271 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
272 |
|
|
|
/* Interlacing Functions */ |
|
|
MBFieldTest = MBFieldTest_mmx; |
|
273 |
|
|
274 |
/* Image Interpolation related functions */ |
/* Image Interpolation related functions */ |
275 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
276 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
277 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
278 |
|
|
279 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
/* Image RGB->YV12 related functions */ |
280 |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
281 |
|
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
282 |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
yuv_to_yv12 = yuv_to_yv12_mmx; |
|
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
|
|
|
|
|
/* reduced resolution */ |
|
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_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; |
|
|
|
|
|
/* image input xxx_to_yv12 related functions */ |
|
|
yv12_to_yv12 = yv12_to_yv12_mmx; |
|
|
bgr_to_yv12 = bgr_to_yv12_mmx; |
|
|
bgra_to_yv12 = bgra_to_yv12_mmx; |
|
283 |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
284 |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
285 |
|
|
286 |
/* image output yv12_to_xxx related functions */ |
/* Image YV12->RGB related functions */ |
287 |
yv12_to_bgr = yv12_to_bgr_mmx; |
yv12_to_rgb24 = yv12_to_rgb24_mmx; |
288 |
yv12_to_bgra = yv12_to_bgra_mmx; |
yv12_to_rgb32 = yv12_to_rgb32_mmx; |
289 |
yv12_to_yuyv = yv12_to_yuyv_mmx; |
yv12_to_yuyv = yv12_to_yuyv_mmx; |
290 |
yv12_to_uyvy = yv12_to_uyvy_mmx; |
yv12_to_uyvy = yv12_to_uyvy_mmx; |
291 |
|
|
|
yv12_to_yuyvi = yv12_to_yuyvi_mmx; |
|
|
yv12_to_uyvyi = yv12_to_uyvyi_mmx; |
|
|
|
|
292 |
/* Motion estimation related functions */ |
/* Motion estimation related functions */ |
293 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
294 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
296 |
sad16bi = sad16bi_mmx; |
sad16bi = sad16bi_mmx; |
297 |
sad8bi = sad8bi_mmx; |
sad8bi = sad8bi_mmx; |
298 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
|
sad16v = sad16v_mmx; |
|
299 |
|
|
300 |
} |
} |
301 |
|
|
305 |
/* ME functions */ |
/* ME functions */ |
306 |
sad16bi = sad16bi_3dn; |
sad16bi = sad16bi_3dn; |
307 |
sad8bi = sad8bi_3dn; |
sad8bi = sad8bi_3dn; |
|
|
|
|
yuyv_to_yv12 = yuyv_to_yv12_3dn; |
|
|
uyvy_to_yv12 = uyvy_to_yv12_3dn; |
|
308 |
} |
} |
309 |
|
|
310 |
|
|
318 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
319 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
320 |
|
|
|
/* reduced resolution */ |
|
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
|
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
|
|
|
|
321 |
/* Quantization */ |
/* Quantization */ |
|
quant4_intra = quant4_intra_xmm; |
|
|
quant4_inter = quant4_inter_xmm; |
|
|
|
|
322 |
dequant_intra = dequant_intra_xmm; |
dequant_intra = dequant_intra_xmm; |
323 |
dequant_inter = dequant_inter_xmm; |
dequant_inter = dequant_inter_xmm; |
324 |
|
|
326 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
327 |
|
|
328 |
/* Colorspace transformation */ |
/* Colorspace transformation */ |
329 |
yv12_to_yv12 = yv12_to_yv12_xmm; |
yuv_to_yv12 = yuv_to_yv12_xmm; |
|
yuyv_to_yv12 = yuyv_to_yv12_xmm; |
|
|
uyvy_to_yv12 = uyvy_to_yv12_xmm; |
|
330 |
|
|
331 |
/* ME functions */ |
/* ME functions */ |
332 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
334 |
sad16bi = sad16bi_xmm; |
sad16bi = sad16bi_xmm; |
335 |
sad8bi = sad8bi_xmm; |
sad8bi = sad8bi_xmm; |
336 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
337 |
sad16v = sad16v_xmm; |
|
338 |
} |
} |
339 |
|
|
340 |
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
345 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
346 |
} |
} |
347 |
|
|
|
if ((cpu_flags & XVID_CPU_3DNOWEXT) > 0) { |
|
|
|
|
|
/* Inverse DCT */ |
|
|
idct = idct_3dne; |
|
|
|
|
|
/* Buffer transfer */ |
|
|
transfer_8to16copy = transfer_8to16copy_3dne; |
|
|
transfer_16to8copy = transfer_16to8copy_3dne; |
|
|
transfer_8to16sub = transfer_8to16sub_3dne; |
|
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
|
|
transfer_16to8add = transfer_16to8add_3dne; |
|
|
transfer8x8_copy = transfer8x8_copy_3dne; |
|
|
|
|
|
/* Quantization */ |
|
|
dequant4_intra = dequant4_intra_3dne; |
|
|
dequant4_inter = dequant4_inter_3dne; |
|
|
quant_intra = quant_intra_3dne; |
|
|
quant_inter = quant_inter_3dne; |
|
|
dequant_intra = dequant_intra_3dne; |
|
|
dequant_inter = dequant_inter_3dne; |
|
|
|
|
|
/* ME functions */ |
|
|
calc_cbp = calc_cbp_3dne; |
|
|
sad16 = sad16_3dne; |
|
|
sad8 = sad8_3dne; |
|
|
sad16bi = sad16bi_3dne; |
|
|
sad8bi = sad8bi_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; |
|
|
} |
|
|
|
|
|
|
|
348 |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
349 |
#ifdef EXPERIMENTAL_SSE2_CODE |
#ifdef EXPERIMENTAL_SSE2_CODE |
350 |
|
|
380 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
381 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
382 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
383 |
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
Halfpel8_Refine = Halfpel8_Refine_ia64; |
384 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
385 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
386 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |
412 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
413 |
} |
} |
414 |
|
|
|
|
|
|
|
|
|
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; |
|
|
} |
|
|
|
|
|
#define RANDOM(min,max) min + (rand() % (max-min)) |
|
|
|
|
|
void random8(uint8_t * block, int size, int min, int max) |
|
|
{ |
|
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = RANDOM(min,max); |
|
|
} |
|
|
|
|
|
void random16(int16_t * block, int size, int min, int max) |
|
|
{ |
|
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
block[i] = RANDOM(min,max); |
|
|
} |
|
|
|
|
|
int compare16(const int16_t * blockA, const int16_t * blockB, int size) |
|
|
{ |
|
|
int i; |
|
|
for (i = 0; i < size; i++) |
|
|
if (blockA[i] != blockB[i]) |
|
|
return 1; |
|
|
|
|
|
return 0; |
|
|
} |
|
|
|
|
|
int diff16(const int16_t * blockA, const int16_t * blockB, int size) |
|
|
{ |
|
|
int i, diff = 0; |
|
|
for (i = 0; i < size; i++) |
|
|
diff += ABS(blockA[i]-blockB[i]); |
|
|
return diff; |
|
|
} |
|
|
|
|
|
|
|
|
#define XVID_TEST_RANDOM 0x00000001 /* random input data */ |
|
|
#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); |
|
|
} |
|
|
memcpy(arrayB, arrayA, 64*sizeof(int16_t)); |
|
|
|
|
|
if ((test & TEST_FORWARD)) |
|
|
{ |
|
|
timeSTART = read_counter(); |
|
|
((fdctFunc*)funcA)(arrayA); |
|
|
timeA += read_counter() - timeSTART; |
|
|
|
|
|
timeSTART = read_counter(); |
|
|
((fdctFunc*)funcB)(arrayB); |
|
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
else |
|
|
{ |
|
|
timeSTART = read_counter(); |
|
|
((idctFunc*)funcA)(arrayA); |
|
|
timeA += read_counter() - timeSTART; |
|
|
|
|
|
timeSTART = read_counter(); |
|
|
((idctFunc*)funcB)(arrayB); |
|
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
|
|
|
tmp = diff16(arrayA, arrayB, 64) / 64; |
|
|
if (tmp > max_error) |
|
|
max_error = tmp; |
|
|
if (tmp < min_error) |
|
|
min_error = tmp; |
|
|
|
|
|
count++; |
|
|
} |
|
|
|
|
|
/* print the "average difference" of best/worst transforms */ |
|
|
printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error); |
|
|
|
|
|
return 0; |
|
|
} |
|
|
|
|
|
|
|
|
#define TEST_QUANT 0x00000001 /* forward quantization */ |
|
|
#define TEST_INTRA 0x00000002 /* intra */ |
|
|
#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 */ |
|
|
{ |
|
|
timeSTART = read_counter(); |
|
|
((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q); |
|
|
timeA += read_counter() - timeSTART; |
|
|
|
|
|
timeSTART = read_counter(); |
|
|
((quanth263_intraFunc*)funcB)(arrayB, arrayX, q, q); |
|
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
else /* inter */ |
|
|
{ |
|
|
timeSTART = read_counter(); |
|
|
retA = ((quanth263_interFunc*)funcA)(arrayA, arrayX, q); |
|
|
timeA += read_counter() - timeSTART; |
|
|
|
|
|
timeSTART = read_counter(); |
|
|
retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q); |
|
|
timeB += read_counter() - timeSTART; |
|
|
} |
|
|
|
|
|
/* compare return value from quant_inter, and compare (de)quantiz'd arrays */ |
|
|
if ( ((test&TEST_QUANT) && !(test&TEST_INTRA) && retA != retB ) || |
|
|
compare16(arrayA, arrayB, 64)) |
|
|
{ |
|
|
errors++; |
|
|
if ((flags & XVID_TEST_VERBOSE)) |
|
|
printf("%s error: q=%i, i=%i\n", nameB, q, i); |
|
|
} |
|
|
|
|
|
count++; |
|
|
} |
|
|
} |
|
|
|
|
|
printf("%s:\t%i", nameB, (int)(timeB / count)); |
|
|
if (errors>0) |
|
|
printf("\t(%i errors out of %i)", errors, count); |
|
|
printf("\n"); |
|
|
|
|
|
return 0; |
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
int xvid_init_test(int flags) |
|
|
{ |
|
|
int cpu_flags; |
|
|
|
|
|
srand(time(0)); |
|
|
|
|
|
printf("xvid_init_test\n"); |
|
|
|
|
|
#if defined(ARCH_X86) |
|
|
cpu_flags = check_cpu_features(); |
|
|
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; |
|
|
} |
|
|
|
|
|
|
|
|
int |
|
|
xvid_init(void *handle, |
|
|
int opt, |
|
|
void *param1, |
|
|
void *param2) |
|
|
{ |
|
|
switch(opt) |
|
|
{ |
|
|
case XVID_INIT_INIT : |
|
|
return xvid_init_init((XVID_INIT_PARAM*)param1); |
|
|
|
|
|
case XVID_INIT_CONVERT : |
|
|
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
|
|
|
|
|
case XVID_INIT_TEST : |
|
|
return xvid_init_test((int)param1); |
|
|
|
|
|
default : |
|
|
return XVID_ERR_FAIL; |
|
|
} |
|
|
} |
|
|
|
|
415 |
/***************************************************************************** |
/***************************************************************************** |
416 |
* XviD Native decoder entry point |
* XviD Native decoder entry point |
417 |
* |
* |
430 |
{ |
{ |
431 |
switch (opt) { |
switch (opt) { |
432 |
case XVID_DEC_DECODE: |
case XVID_DEC_DECODE: |
433 |
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2); |
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1); |
434 |
|
|
435 |
case XVID_DEC_CREATE: |
case XVID_DEC_CREATE: |
436 |
return decoder_create((XVID_DEC_PARAM *) param1); |
return decoder_create((XVID_DEC_PARAM *) param1); |
462 |
{ |
{ |
463 |
switch (opt) { |
switch (opt) { |
464 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
465 |
|
#ifdef BFRAMES |
466 |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
467 |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
468 |
(XVID_ENC_STATS *) param2); |
(XVID_ENC_STATS *) param2); |
469 |
else |
else |
470 |
|
#endif |
471 |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
472 |
(XVID_ENC_STATS *) param2); |
(XVID_ENC_STATS *) param2); |
473 |
|
|