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> |
|
|
* |
|
|
* $Id$ |
|
|
* |
|
|
****************************************************************************/ |
|
|
|
|
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 */ |
|
|
|
|
|
#ifdef ARCH_X86 |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
#else |
|
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
|
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
|
|
vfilter_31 = xvid_VFilter_31_C; |
|
|
hfilter_31 = xvid_HFilter_31_C; |
|
|
#endif |
|
|
|
|
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; |
|
|
|
|
|
/* 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 */ |
322 |
dequant_intra = dequant_intra_xmm; |
dequant_intra = dequant_intra_xmm; |
323 |
dequant_inter = dequant_inter_xmm; |
dequant_inter = dequant_inter_xmm; |
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) { |
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; |
|
|
} |
|
|
|
|
|
|
|
|
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); |
|
|
|
|
|
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 |
|
|