[cvs] / xvidcore / src / xvid.c Repository:
ViewVC logotype

Diff of /xvidcore/src/xvid.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.42, Sun Feb 16 05:11:39 2003 UTC revision 1.50, Mon Apr 5 20:36:36 2004 UTC
# Line 3  Line 3 
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
# Line 26  Line 19 
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   ****************************************************************************/   ****************************************************************************/
# Line 57  Line 39 
39  #include "image/reduced.h"  #include "image/reduced.h"
40  #include "utils/mem_transfer.h"  #include "utils/mem_transfer.h"
41  #include "utils/mbfunctions.h"  #include "utils/mbfunctions.h"
42  #include "quant/quant_h263.h"  #include "quant/quant.h"
 #include "quant/quant_mpeg4.h"  
43  #include "motion/motion.h"  #include "motion/motion.h"
44  #include "motion/sad.h"  #include "motion/sad.h"
45  #include "utils/emms.h"  #include "utils/emms.h"
46  #include "utils/timer.h"  #include "utils/timer.h"
47  #include "bitstream/mbcoding.h"  #include "bitstream/mbcoding.h"
48    #include "image/qpel.h"
49    #include "image/postprocessing.h"
50    
51  #if defined(ARCH_IS_IA32)  #if defined(_DEBUG)
52    unsigned int xvid_debug = 0; /* xvid debug mask */
53    #endif
54    
55  #if defined(_MSC_VER)  #if defined(ARCH_IS_IA32) && defined(_MSC_VER)
56  #       include <windows.h>  #       include <windows.h>
57  #else  #elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC)
58  #       include <signal.h>  #       include <signal.h>
59  #       include <setjmp.h>  #       include <setjmp.h>
60    
# Line 84  Line 69 
69    
70    
71  /*  /*
72  calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled   * Calls the funcptr, and returns whether SIGILL (illegal instruction) was
73  return values:   * signalled
74  -1 : could not determine   *
75  0  : SIGILL was *not* signalled   * Return values:
76  1  : SIGILL was signalled   *  -1 : could not determine
77     *   0 : SIGILL was *not* signalled
78     *   1 : SIGILL was signalled
79  */  */
80    #if defined(ARCH_IS_IA32) && defined(_MSC_VER)
81  int  static int
82  sigill_check(void (*func)())  sigill_check(void (*func)())
83  {  {
 #if defined(_MSC_VER)  
84          _try {          _try {
85                  func();                  func();
86          }          } _except(EXCEPTION_EXECUTE_HANDLER) {
         _except(EXCEPTION_EXECUTE_HANDLER) {  
87    
88                  if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION)                  if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION)
89                          return 1;                          return(1);
90          }          }
91          return 0;          return(0);
92  #else  }
93    #elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC)
94    static int
95    sigill_check(void (*func)())
96    {
97      void * old_handler;      void * old_handler;
98      int jmpret;      int jmpret;
99    
100        /* Set our SIGILL handler */
101      old_handler = signal(SIGILL, sigill_handler);      old_handler = signal(SIGILL, sigill_handler);
102      if (old_handler == SIG_ERR)  
103      {      /* Check for error */
104          return -1;      if (old_handler == SIG_ERR) {
105            return(-1);
106      }      }
107    
108        /* Save stack context, so if func triggers a SIGILL, we can still roll
109             * back to a valid CPU state */
110      jmpret = setjmp(mark);      jmpret = setjmp(mark);
111      if (jmpret == 0)  
112      {          /* If setjmp returned directly, then its returned value is 0, and we still
113             * have to test the passed func. Otherwise it means the stack context has
114             * been restored by a longjmp() call, which in our case happens only in the
115             * signal handler */
116        if (jmpret == 0) {
117          func();          func();
118      }      }
119    
120        /* Restore old signal handler */
121      signal(SIGILL, old_handler);      signal(SIGILL, old_handler);
122    
123      return jmpret;      return(jmpret);
 #endif  
124  }  }
125  #endif  #endif
126    
# Line 146  Line 142 
142  #endif  #endif
143    
144  #if defined(ARCH_IS_PPC)  #if defined(ARCH_IS_PPC)
145  #if defined(ARCH_IS_PPC_ALTIVEC)          if (!sigill_check(altivec_trigger))
146          cpu_flags |= XVID_CPU_ALTIVEC;          cpu_flags |= XVID_CPU_ALTIVEC;
147  #endif  #endif
 #endif  
148    
149          return cpu_flags;          return cpu_flags;
150  }  }
# Line 171  Line 166 
166    
167    
168  static  static
169  int xvid_init_init(XVID_INIT_PARAM * init_param)  int xvid_gbl_init(xvid_gbl_init_t * init)
170  {  {
171          int cpu_flags;          unsigned int cpu_flags;
   
         /* Inform the client the API version */  
         init_param->api_version = API_VERSION;  
172    
173          /* Inform the client the core build - unused because we're still alpha */          if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */
174          init_param->core_build = 1000;                  return XVID_ERR_VERSION;
   
         /* 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;  
         }  
   
         init_param->cpu_flags = cpu_flags;  
175    
176            cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags();
177    
178          /* Initialize the function pointers */          /* Initialize the function pointers */
179          idct_int32_init();          idct_int32_init();
# Line 214  Line 189 
189          /* Restore FPU context : emms_c is a nop functions */          /* Restore FPU context : emms_c is a nop functions */
190          emms = emms_c;          emms = emms_c;
191    
192            /* Qpel stuff */
193            xvid_QP_Funcs = &xvid_QP_Funcs_C;
194            xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_C;
195            xvid_Init_QP();
196    
197          /* Quantization functions */          /* Quantization functions */
198          quant_intra   = quant_intra_c;          quant_h263_intra   = quant_h263_intra_c;
199          dequant_intra = dequant_intra_c;          quant_h263_inter   = quant_h263_inter_c;
200          quant_inter   = quant_inter_c;          dequant_h263_intra = dequant_h263_intra_c;
201          dequant_inter = dequant_inter_c;          dequant_h263_inter = dequant_h263_inter_c;
202    
203          quant4_intra   = quant4_intra_c;          quant_mpeg_intra   = quant_mpeg_intra_c;
204          dequant4_intra = dequant4_intra_c;          quant_mpeg_inter   = quant_mpeg_inter_c;
205          quant4_inter   = quant4_inter_c;          dequant_mpeg_intra = dequant_mpeg_intra_c;
206          dequant4_inter = dequant4_inter_c;          dequant_mpeg_inter = dequant_mpeg_inter_c;
207    
208          /* Block transfer related functions */          /* Block transfer related functions */
209          transfer_8to16copy = transfer_8to16copy_c;          transfer_8to16copy = transfer_8to16copy_c;
# Line 256  Line 236 
236          interpolate8x8_avg2 = interpolate8x8_avg2_c;          interpolate8x8_avg2 = interpolate8x8_avg2_c;
237          interpolate8x8_avg4 = interpolate8x8_avg4_c;          interpolate8x8_avg4 = interpolate8x8_avg4_c;
238    
239          /* reduced resoltuion */          /* postprocessing */
240            image_brightness = image_brightness_c;
241    
242            /* reduced resolution */
243          copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C;          copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C;
244          add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C;          add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C;
245          vfilter_31 = xvid_VFilter_31_C;          vfilter_31 = xvid_VFilter_31_C;
# Line 275  Line 258 
258          bgra_to_yv12    = bgra_to_yv12_c;          bgra_to_yv12    = bgra_to_yv12_c;
259          abgr_to_yv12    = abgr_to_yv12_c;          abgr_to_yv12    = abgr_to_yv12_c;
260          rgba_to_yv12    = rgba_to_yv12_c;          rgba_to_yv12    = rgba_to_yv12_c;
261            argb_to_yv12    = argb_to_yv12_c;
262          yuyv_to_yv12    = yuyv_to_yv12_c;          yuyv_to_yv12    = yuyv_to_yv12_c;
263          uyvy_to_yv12    = uyvy_to_yv12_c;          uyvy_to_yv12    = uyvy_to_yv12_c;
264    
# Line 284  Line 268 
268          bgrai_to_yv12   = bgrai_to_yv12_c;          bgrai_to_yv12   = bgrai_to_yv12_c;
269          abgri_to_yv12   = abgri_to_yv12_c;          abgri_to_yv12   = abgri_to_yv12_c;
270          rgbai_to_yv12   = rgbai_to_yv12_c;          rgbai_to_yv12   = rgbai_to_yv12_c;
271            argbi_to_yv12   = argbi_to_yv12_c;
272          yuyvi_to_yv12   = yuyvi_to_yv12_c;          yuyvi_to_yv12   = yuyvi_to_yv12_c;
273          uyvyi_to_yv12   = uyvyi_to_yv12_c;          uyvyi_to_yv12   = uyvyi_to_yv12_c;
274    
   
275          /* All colorspace transformation functions YV12->User format */          /* All colorspace transformation functions YV12->User format */
276          yv12_to_rgb555  = yv12_to_rgb555_c;          yv12_to_rgb555  = yv12_to_rgb555_c;
277          yv12_to_rgb565  = yv12_to_rgb565_c;          yv12_to_rgb565  = yv12_to_rgb565_c;
# Line 295  Line 279 
279          yv12_to_bgra    = yv12_to_bgra_c;          yv12_to_bgra    = yv12_to_bgra_c;
280          yv12_to_abgr    = yv12_to_abgr_c;          yv12_to_abgr    = yv12_to_abgr_c;
281          yv12_to_rgba    = yv12_to_rgba_c;          yv12_to_rgba    = yv12_to_rgba_c;
282            yv12_to_argb    = yv12_to_argb_c;
283          yv12_to_yuyv    = yv12_to_yuyv_c;          yv12_to_yuyv    = yv12_to_yuyv_c;
284          yv12_to_uyvy    = yv12_to_uyvy_c;          yv12_to_uyvy    = yv12_to_uyvy_c;
285    
# Line 304  Line 289 
289          yv12_to_bgrai   = yv12_to_bgrai_c;          yv12_to_bgrai   = yv12_to_bgrai_c;
290          yv12_to_abgri   = yv12_to_abgri_c;          yv12_to_abgri   = yv12_to_abgri_c;
291          yv12_to_rgbai   = yv12_to_rgbai_c;          yv12_to_rgbai   = yv12_to_rgbai_c;
292            yv12_to_argbi   = yv12_to_argbi_c;
293          yv12_to_yuyvi   = yv12_to_yuyvi_c;          yv12_to_yuyvi   = yv12_to_yuyvi_c;
294          yv12_to_uyvyi   = yv12_to_uyvyi_c;          yv12_to_uyvyi   = yv12_to_uyvyi_c;
295    
# Line 315  Line 301 
301          sad8bi   = sad8bi_c;          sad8bi   = sad8bi_c;
302          dev16    = dev16_c;          dev16    = dev16_c;
303          sad16v   = sad16v_c;          sad16v   = sad16v_c;
304            sse8_16bit = sse8_16bit_c;
 //      Halfpel8_Refine = Halfpel8_Refine_c;  
305    
306  #if defined(ARCH_IS_IA32)  #if defined(ARCH_IS_IA32)
307    
308          if ((cpu_flags & XVID_CPU_ASM))          if ((cpu_flags & XVID_CPU_ASM)) {
         {  
309                  vfilter_31 = xvid_VFilter_31_x86;                  vfilter_31 = xvid_VFilter_31_x86;
310                  hfilter_31 = xvid_HFilter_31_x86;                  hfilter_31 = xvid_HFilter_31_x86;
311          }          }
# Line 337  Line 321 
321          if ((cpu_flags & XVID_CPU_MMX)) {          if ((cpu_flags & XVID_CPU_MMX)) {
322    
323                  /* Forward and Inverse Discrete Cosine Transformation functions */                  /* Forward and Inverse Discrete Cosine Transformation functions */
324                  fdct = fdct_mmx;                  fdct = fdct_mmx_skal;
325                  idct = idct_mmx;                  idct = idct_mmx;
326    
327                    /* Qpel stuff */
328                    xvid_QP_Funcs = &xvid_QP_Funcs_mmx;
329                    xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx;
330    
331                  /* Quantization related functions */                  /* Quantization related functions */
332                  quant_intra   = quant_intra_mmx;                  quant_h263_intra   = quant_h263_intra_mmx;
333                  dequant_intra = dequant_intra_mmx;                  quant_h263_inter   = quant_h263_inter_mmx;
334                  quant_inter   = quant_inter_mmx;                  dequant_h263_intra = dequant_h263_intra_mmx;
335                  dequant_inter = dequant_inter_mmx;                  dequant_h263_inter = dequant_h263_inter_mmx;
336    
337                  quant4_intra   = quant4_intra_mmx;                  quant_mpeg_intra   = quant_mpeg_intra_mmx;
338                  dequant4_intra = dequant4_intra_mmx;                  quant_mpeg_inter   = quant_mpeg_inter_mmx;
339                  quant4_inter   = quant4_inter_mmx;                  dequant_mpeg_intra = dequant_mpeg_intra_mmx;
340                  dequant4_inter = dequant4_inter_mmx;                  dequant_mpeg_inter = dequant_mpeg_inter_mmx;
341    
342                  /* Block related functions */                  /* Block related functions */
343                  transfer_8to16copy = transfer_8to16copy_mmx;                  transfer_8to16copy = transfer_8to16copy_mmx;
# Line 374  Line 362 
362                  interpolate8x8_avg2 = interpolate8x8_avg2_mmx;                  interpolate8x8_avg2 = interpolate8x8_avg2_mmx;
363                  interpolate8x8_avg4 = interpolate8x8_avg4_mmx;                  interpolate8x8_avg4 = interpolate8x8_avg4_mmx;
364    
365                    /* postprocessing */
366                    image_brightness = image_brightness_mmx;
367    
368                  /* reduced resolution */                  /* reduced resolution */
369                  copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx;                  copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx;
370                  add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx;                  add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx;
# Line 405  Line 396 
396                  sad8bi  = sad8bi_mmx;                  sad8bi  = sad8bi_mmx;
397                  dev16    = dev16_mmx;                  dev16    = dev16_mmx;
398                  sad16v   = sad16v_mmx;                  sad16v   = sad16v_mmx;
399                    sse8_16bit = sse8_16bit_mmx;
400          }          }
401    
402          /* these 3dnow functions are faster than mmx, but slower than xmm. */          /* these 3dnow functions are faster than mmx, but slower than xmm. */
# Line 423  Line 415 
415    
416          if ((cpu_flags & XVID_CPU_MMXEXT)) {          if ((cpu_flags & XVID_CPU_MMXEXT)) {
417    
418                  /* Inverse DCT */                  /* DCT */
419                    fdct = fdct_xmm_skal;
420                  idct = idct_xmm;                  idct = idct_xmm;
421    
422                  /* Interpolation */                  /* Interpolation */
# Line 436  Line 429 
429                  add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm;                  add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm;
430    
431                  /* Quantization */                  /* Quantization */
432                  quant4_intra = quant4_intra_xmm;                  quant_mpeg_intra = quant_mpeg_intra_xmm;
433                  quant4_inter = quant4_inter_xmm;                  quant_mpeg_inter = quant_mpeg_inter_xmm;
434    
435                  dequant_intra = dequant_intra_xmm;                  dequant_h263_intra = dequant_h263_intra_xmm;
436                  dequant_inter = dequant_inter_xmm;                  dequant_h263_inter = dequant_h263_inter_xmm;
437    
438                  /* Buffer transfer */                  /* Buffer transfer */
439                  transfer_8to16sub2 = transfer_8to16sub2_xmm;                  transfer_8to16sub2 = transfer_8to16sub2_xmm;
# Line 482  Line 475 
475                  transfer8x8_copy = transfer8x8_copy_3dne;                  transfer8x8_copy = transfer8x8_copy_3dne;
476    
477                  /* Quantization */                  /* Quantization */
478                  dequant4_intra = dequant4_intra_3dne;                  quant_h263_intra = quant_h263_intra_3dne;
479                  dequant4_inter = dequant4_inter_3dne;                  quant_h263_inter = quant_h263_inter_3dne;
480                  quant_intra = quant_intra_3dne;                  dequant_mpeg_intra = dequant_mpeg_intra_3dne;
481                  quant_inter = quant_inter_3dne;                  dequant_mpeg_inter = dequant_mpeg_inter_3dne;
482                  dequant_intra = dequant_intra_3dne;                  dequant_h263_intra = dequant_h263_intra_3dne;
483                  dequant_inter = dequant_inter_3dne;                  dequant_h263_inter = dequant_h263_inter_3dne;
484    
485                  /* ME functions */                  /* ME functions */
486                  calc_cbp = calc_cbp_3dne;                  calc_cbp = calc_cbp_3dne;
# Line 503  Line 496 
496                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;
497          }          }
498    
   
499          if ((cpu_flags & XVID_CPU_SSE2)) {          if ((cpu_flags & XVID_CPU_SSE2)) {
500    
501                  calc_cbp = calc_cbp_sse2;                  calc_cbp = calc_cbp_sse2;
502    
503                  /* Quantization */                  /* Quantization */
504                  quant_intra   = quant_intra_sse2;                  quant_h263_intra   = quant_h263_intra_sse2;
505                  dequant_intra = dequant_intra_sse2;                  quant_h263_inter   = quant_h263_inter_sse2;
506                  quant_inter   = quant_inter_sse2;                  dequant_h263_intra = dequant_h263_intra_sse2;
507                  dequant_inter = dequant_inter_sse2;                  dequant_h263_inter = dequant_h263_inter_sse2;
508    
509  #if defined(EXPERIMENTAL_SSE2_CODE)                  /* SAD operators */
                 /* ME; slower than xmm */  
510                  sad16    = sad16_sse2;                  sad16    = sad16_sse2;
511                  dev16    = dev16_sse2;                  dev16    = dev16_sse2;
512  #endif  
513                  /* Forward and Inverse DCT */                  /* DCT operators
514                  idct  = idct_sse2;                   * no iDCT because it's not "Walken matching" */
515                  fdct = fdct_sse2;                  fdct = fdct_sse2_skal;
516          }          }
517  #endif  #endif /* ARCH_IS_IA32 */
518    
519  #if defined(ARCH_IS_IA64)  #if defined(ARCH_IS_IA64)
520          if ((cpu_flags & XVID_CPU_ASM)) { //use assembler routines?          if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */
521            idct_ia64_init();            idct_ia64_init();
522            fdct = fdct_ia64;            fdct = fdct_ia64;
523            idct = idct_ia64;   //not yet working, crashes            idct = idct_ia64;   /*not yet working, crashes */
524            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
525            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
526            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
# Line 537  Line 528 
528            sad16bi = sad16bi_ia64;            sad16bi = sad16bi_ia64;
529            sad8 = sad8_ia64;            sad8 = sad8_ia64;
530            dev16 = dev16_ia64;            dev16 = dev16_ia64;
531  //        Halfpel8_Refine = Halfpel8_Refine_ia64;  /*        Halfpel8_Refine = Halfpel8_Refine_ia64; */
532            quant_intra = quant_intra_ia64;            quant_h263_intra = quant_h263_intra_ia64;
533            dequant_intra = dequant_intra_ia64;            quant_h263_inter = quant_h263_inter_ia64;
534            quant_inter = quant_inter_ia64;            dequant_h263_intra = dequant_h263_intra_ia64;
535            dequant_inter = dequant_inter_ia64;            dequant_h263_inter = dequant_h263_inter_ia64;
536            transfer_8to16copy = transfer_8to16copy_ia64;            transfer_8to16copy = transfer_8to16copy_ia64;
537            transfer_16to8copy = transfer_16to8copy_ia64;            transfer_16to8copy = transfer_16to8copy_ia64;
538            transfer_8to16sub = transfer_8to16sub_ia64;            transfer_8to16sub = transfer_8to16sub_ia64;
539            transfer_8to16sub2 = transfer_8to16sub2_ia64;            transfer_8to16sub2 = transfer_8to16sub2_ia64;
540            transfer_16to8add = transfer_16to8add_ia64;            transfer_16to8add = transfer_16to8add_ia64;
541            transfer8x8_copy = transfer8x8_copy_ia64;            transfer8x8_copy = transfer8x8_copy_ia64;
           DEBUG("Using IA-64 assembler routines.\n");  
542          }          }
543  #endif  #endif
544    
545  #if defined(ARCH_IS_PPC)  #if defined(ARCH_IS_PPC)
546          if ((cpu_flags & XVID_CPU_ASM))          if ((cpu_flags & XVID_CPU_ALTIVEC)) {
547          {            /* sad operators */
548                  calc_cbp = calc_cbp_ppc;            sad16 = sad16_altivec_c;
549          }            sad16bi = sad16bi_altivec_c;
550              sad8 = sad8_altivec_c;
551          if ((cpu_flags & XVID_CPU_ALTIVEC))            dev16 = dev16_altivec_c;
552          {  
553                  calc_cbp = calc_cbp_altivec;            sse8_16bit = sse8_16bit_altivec_c;
554                  fdct = fdct_altivec;  
555                  idct = idct_altivec;            /* mem transfer */
556                  sadInit = sadInit_altivec;            transfer_8to16copy = transfer_8to16copy_altivec_c;
557                  sad16 = sad16_altivec;            transfer_16to8copy = transfer_16to8copy_altivec_c;
558                  sad8 = sad8_altivec;            transfer_8to16sub = transfer_8to16sub_altivec_c;
559                  dev16 = dev16_altivec;            transfer_8to16subro = transfer_8to16subro_altivec_c;
560          }            transfer_8to16sub2 = transfer_8to16sub2_altivec_c;
561  #endif            transfer_16to8add = transfer_16to8add_altivec_c;
562              transfer8x8_copy = transfer8x8_copy_altivec_c;
         return XVID_ERR_OK;  
 }  
   
   
   
 static int  
 xvid_init_convert(XVID_INIT_CONVERTINFO* convert)  
 {  
         // const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP);  
         const int width = convert->width;  
         const int height = convert->height;  
         const int width2 = convert->width/2;  
         const int height2 = convert->height/2;  
         IMAGE img;  
   
         switch (convert->input.colorspace & ~XVID_CSP_VFLIP)  
         {  
                 case XVID_CSP_YV12 :  
                         img.y = convert->input.y;  
                         img.v = (uint8_t*)convert->input.y + width*height;  
                         img.u = (uint8_t*)convert->input.y + width*height + width2*height2;  
                         image_output(&img, width, height, width,  
                                                 convert->output.y, convert->output.y_stride,  
                                                 convert->output.colorspace, convert->interlacing);  
                         break;  
   
                 default :  
                         return XVID_ERR_FORMAT;  
         }  
   
   
         emms();  
         return XVID_ERR_OK;  
 }  
   
   
   
 void fill8(uint8_t * block, int size, int value)  
 {  
         int i;  
         for (i = 0; i < size; i++)  
                 block[i] = value;  
 }  
   
 void fill16(int16_t * block, int size, int value)  
 {  
         int i;  
         for (i = 0; i < size; i++)  
                 block[i] = value;  
 }  
   
 #define RANDOM(min,max) min + (rand() % (max-min))  
563    
564  void random8(uint8_t * block, int size, int min, int max)            /* Inverse DCT */
565  {            idct = idct_altivec_c;
         int i;  
         for (i = 0; i < size; i++)  
                 block[i] = RANDOM(min,max);  
 }  
566    
567  void random16(int16_t * block, int size, int min, int max)            /* Interpolation */
568  {            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c;
569          int i;            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c;
570          for (i = 0; i < size; i++)            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c;
571                  block[i] = RANDOM(min,max);  
572  }            interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c;
573              interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c;
574    
575              interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_altivec_c;
576    
577              /* Colorspace conversion */
578              bgra_to_yv12 = bgra_to_yv12_altivec_c;
579              abgr_to_yv12 = abgr_to_yv12_altivec_c;
580              rgba_to_yv12 = rgba_to_yv12_altivec_c;
581              argb_to_yv12 = argb_to_yv12_altivec_c;
582    
583  int compare16(const int16_t * blockA, const int16_t * blockB, int size)            yuyv_to_yv12 = yuyv_to_yv12_altivec_c;
584  {            uyvy_to_yv12 = uyvy_to_yv12_altivec_c;
         int i;  
         for (i = 0; i < size; i++)  
                 if (blockA[i] != blockB[i])  
                         return 1;  
585    
586          return 0;            yv12_to_yuyv = yv12_to_yuyv_altivec_c;
587  }            yv12_to_uyvy = yv12_to_uyvy_altivec_c;
588    
589  int diff16(const int16_t * blockA, const int16_t * blockB, int size)            /* Quantization */
590  {            quant_h263_intra = quant_h263_intra_altivec_c;
591          int i, diff = 0;            quant_h263_inter = quant_h263_inter_altivec_c;
592          for (i = 0; i < size; i++)            dequant_h263_intra = dequant_h263_intra_altivec_c;
593                  diff += ABS(blockA[i]-blockB[i]);            dequant_h263_inter = dequant_h263_inter_altivec_c;
         return diff;  
594  }  }
595    #endif
596    
597    #if defined(_DEBUG)
598        xvid_debug = init->debug;
599    #endif
600    
601  #define XVID_TEST_RANDOM        0x00000001      /* random input data */      return(0);
 #define XVID_TEST_VERBOSE       0x00000002      /* verbose error output */  
   
   
 #define TEST_FORWARD    0x00000001      /* intra */  
 #define TEST_FDCT  (TEST_FORWARD)  
 #define TEST_IDCT  (0)  
   
 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 */  
         {  
                 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);  
602                  }                  }
                 memcpy(arrayB, arrayA, 64*sizeof(int16_t));  
603    
                 if ((test & TEST_FORWARD))  
                 {  
                         timeSTART = read_counter();  
                         ((fdctFunc*)funcA)(arrayA);  
                         timeA += read_counter() - timeSTART;  
604    
605                          timeSTART = read_counter();  static int
606                          ((fdctFunc*)funcB)(arrayB);  xvid_gbl_info(xvid_gbl_info_t * info)
                         timeB += read_counter() - timeSTART;  
                 }  
                 else  
607                  {                  {
608                          timeSTART = read_counter();          if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */
609                          ((idctFunc*)funcA)(arrayA);                  return XVID_ERR_VERSION;
                         timeA += read_counter() - timeSTART;  
610    
611                          timeSTART = read_counter();          info->actual_version = XVID_VERSION;
612                          ((idctFunc*)funcB)(arrayB);          info->build = "xvid-1.0.0";
613                          timeB += read_counter() - timeSTART;          info->cpu_flags = detect_cpu_flags();
                 }  
   
                 tmp = diff16(arrayA, arrayB, 64) / 64;  
                 if (tmp > max_error)  
                         max_error = tmp;  
                 if (tmp < min_error)  
                         min_error = tmp;  
614    
615                  count++;  #if defined(_SMP) && defined(WIN32)
616          }          info->num_threads = pthread_num_processors_np();;
617    #else
618          /* print the "average difference" of best/worst transforms */          info->num_threads = 0;
619          printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error);  #endif
620    
621          return 0;          return 0;
622  }  }
623    
624    
625  #define TEST_QUANT      0x00000001      /* forward quantization */  static int
626  #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 */  
627                          {                          {
628                                  timeSTART = read_counter();          int width;
629                                  ((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q);          int height;
630                                  timeA += read_counter() - timeSTART;          int width2;
631            int height2;
632            IMAGE img;
633    
634                                  timeSTART = read_counter();          if (XVID_VERSION_MAJOR(convert->version) != 1)   /* v1.x.x */
635                                  ((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;  
636    
637                                  timeSTART = read_counter();  #if 0
638                                  retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q);          const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP);
639                                  timeB += read_counter() - timeSTART;  #endif
640                          }          width = convert->width;
641            height = convert->height;
642            width2 = convert->width/2;
643            height2 = convert->height/2;
644    
645                          /* 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))  
646                          {                          {
647                                  errors++;                  case XVID_CSP_YV12 :
648                                  if ((flags & XVID_TEST_VERBOSE))                          img.y = convert->input.plane[0];
649                                          printf("%s error: q=%i, i=%i\n", nameB, q, i);                          img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height;
650                          }                          img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2;
651                            image_output(&img, width, height, width,
652                                                    (uint8_t**)convert->output.plane, convert->output.stride,
653                                                    convert->output.csp, convert->interlacing);
654                            break;
655    
656                          count++;                  default :
657                  }                          return XVID_ERR_FORMAT;
658          }          }
659    
         printf("%s:\t%i", nameB, (int)(timeB / count));  
         if (errors>0)  
                 printf("\t(%i errors out of %i)", errors, count);  
         printf("\n");  
660    
661            emms();
662          return 0;          return 0;
663  }  }
664    
665    /*****************************************************************************
666     * XviD Global Entry point
667  int xvid_init_test(int flags)   *
668  {   * Well this function initialize all internal function pointers according
669          int cpu_flags;   * to the CPU features forced by the library client or autodetected (depending
670     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
671          srand(time(0));   * image colorspace transformation tables.
672     *
673          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;  
 }  
674    
675    
676  int  int
677  xvid_init(void *handle,  xvid_global(void *handle,
678                    int opt,                    int opt,
679                    void *param1,                    void *param1,
680                    void *param2)                    void *param2)
681  {  {
682          switch(opt)          switch(opt)
683          {          {
684                  case XVID_INIT_INIT :                  case XVID_GBL_INIT :
685                          return xvid_init_init((XVID_INIT_PARAM*)param1);                          return xvid_gbl_init((xvid_gbl_init_t*)param1);
686    
687                  case XVID_INIT_CONVERT :          case XVID_GBL_INFO :
688                          return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1);              return xvid_gbl_info((xvid_gbl_info_t*)param1);
689    
690                  case XVID_INIT_TEST :                  case XVID_GBL_CONVERT :
691                          return xvid_init_test((int)param1);                          return xvid_gbl_convert((xvid_gbl_convert_t*)param1);
692    
693                  default :                  default :
694                          return XVID_ERR_FAIL;                          return XVID_ERR_FAIL;
# Line 969  Line 712 
712                          void *param2)                          void *param2)
713  {  {
714          switch (opt) {          switch (opt) {
         case XVID_DEC_DECODE:  
                 return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2);  
   
715          case XVID_DEC_CREATE:          case XVID_DEC_CREATE:
716                  return decoder_create((XVID_DEC_PARAM *) param1);                  return decoder_create((xvid_dec_create_t *) param1);
717    
718          case XVID_DEC_DESTROY:          case XVID_DEC_DESTROY:
719                  return decoder_destroy((DECODER *) handle);                  return decoder_destroy((DECODER *) handle);
720    
721            case XVID_DEC_DECODE:
722                    return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2);
723    
724          default:          default:
725                  return XVID_ERR_FAIL;                  return XVID_ERR_FAIL;
726          }          }
# Line 1003  Line 746 
746          switch (opt) {          switch (opt) {
747          case XVID_ENC_ENCODE:          case XVID_ENC_ENCODE:
748    
749                  if (((Encoder *) handle)->mbParam.max_bframes >= 0)                  return enc_encode((Encoder *) handle,
750                  return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1,                                                            (xvid_enc_frame_t *) param1,
751                                                            (XVID_ENC_STATS *) param2);                                                            (xvid_enc_stats_t *) param2);
                 else  
                 return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1,  
                                                           (XVID_ENC_STATS *) param2);  
752    
753          case XVID_ENC_CREATE:          case XVID_ENC_CREATE:
754                  return encoder_create((XVID_ENC_PARAM *) param1);                  return enc_create((xvid_enc_create_t *) param1);
755    
756          case XVID_ENC_DESTROY:          case XVID_ENC_DESTROY:
757                  return encoder_destroy((Encoder *) handle);                  return enc_destroy((Encoder *) handle);
758    
759          default:          default:
760                  return XVID_ERR_FAIL;                  return XVID_ERR_FAIL;

Legend:
Removed from v.1.42  
changed lines
  Added in v.1.50

No admin address has been configured
ViewVC Help
Powered by ViewVC 1.0.4