[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 1.45.2.7, Mon Jun 9 19:42:08 2003 UTC revision 1.51, Mon Apr 12 14:05:08 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   *  Copyright(C) 2001-2003 Peter Ross <pross@xvid.org>   *  Copyright(C) 2001-2004 Peter Ross <pross@xvid.org>
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 39  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(_DEBUG)  #if defined(_DEBUG)
52  unsigned int xvid_debug = 0; /* xvid debug mask */  unsigned int xvid_debug = 0; /* xvid debug mask */
53  #endif  #endif
54    
55  #if defined(ARCH_IS_IA32)  #if defined(ARCH_IS_IA32) && defined(_MSC_VER)
   
 #if 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 78  Line 77 
77   *   0 : SIGILL was *not* signalled   *   0 : SIGILL was *not* signalled
78   *   1 : SIGILL was signalled   *   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 134  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 163  Line 170 
170  {  {
171          unsigned int cpu_flags;          unsigned int cpu_flags;
172    
173          if (XVID_MAJOR(init->version) != 1) /* v1.x.x */          if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */
174                  return XVID_ERR_VERSION;                  return XVID_ERR_VERSION;
175    
176          cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags();          cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags();
# Line 182  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 224  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 243  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 252  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 263  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 272  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 283  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    
# Line 304  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 = simple_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 341  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 372  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 390  Line 415 
415    
416          if ((cpu_flags & XVID_CPU_MMXEXT)) {          if ((cpu_flags & XVID_CPU_MMXEXT)) {
417    
418                  /* Inverse DCT */                  /* DCT */
419  #if 0 /* We don't use  Walken idct anymore! */                  fdct = fdct_xmm_skal;
420                  idct = idct_xmm;                  idct = idct_xmm;
 #endif  
421    
422                  /* Interpolation */                  /* Interpolation */
423                  interpolate8x8_halfpel_h  = interpolate8x8_halfpel_h_xmm;                  interpolate8x8_halfpel_h  = interpolate8x8_halfpel_h_xmm;
# Line 405  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 438  Line 462 
462    
463          if ((cpu_flags & XVID_CPU_3DNOWEXT)) {          if ((cpu_flags & XVID_CPU_3DNOWEXT)) {
464    
                 /* Inverse DCT */  
 #if 0 /* We don't use  Walken idct anymore! */  
                 idct =  idct_3dne;  
 #endif  
   
465                  /* Buffer transfer */                  /* Buffer transfer */
466                  transfer_8to16copy =  transfer_8to16copy_3dne;                  transfer_8to16copy =  transfer_8to16copy_3dne;
467                  transfer_16to8copy = transfer_16to8copy_3dne;                  transfer_16to8copy = transfer_16to8copy_3dne;
468                  transfer_8to16sub =  transfer_8to16sub_3dne;                  transfer_8to16sub =  transfer_8to16sub_3dne;
469                  transfer_8to16subro =  transfer_8to16subro_3dne;                  transfer_8to16subro =  transfer_8to16subro_3dne;
                 transfer_8to16sub2 =  transfer_8to16sub2_3dne;  
470                  transfer_16to8add = transfer_16to8add_3dne;                  transfer_16to8add = transfer_16to8add_3dne;
471                  transfer8x8_copy = transfer8x8_copy_3dne;                  transfer8x8_copy = transfer8x8_copy_3dne;
472    
473                    if ((cpu_flags & XVID_CPU_MMXEXT)) {
474                            /* Inverse DCT */
475                            idct =  idct_3dne;
476    
477                            /* Buffer transfer */
478                            transfer_8to16sub2 =  transfer_8to16sub2_3dne;
479    
480                            /* Interpolation */
481                            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne;
482                            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne;
483                            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;
484    
485                  /* Quantization */                  /* Quantization */
486                  dequant4_intra = dequant4_intra_3dne;                          quant_h263_intra = quant_h263_intra_3dne;               /* cmov only */
487                  dequant4_inter = dequant4_inter_3dne;                          quant_h263_inter = quant_h263_inter_3dne;
488                  quant_intra = quant_intra_3dne;                          dequant_mpeg_intra = dequant_mpeg_intra_3dne;   /* cmov only */
489                  quant_inter = quant_inter_3dne;                          dequant_mpeg_inter = dequant_mpeg_inter_3dne;
490                  dequant_intra = dequant_intra_3dne;                          dequant_h263_intra = dequant_h263_intra_3dne;
491                  dequant_inter = dequant_inter_3dne;                          dequant_h263_inter = dequant_h263_inter_3dne;
492    
493                  /* ME functions */                  /* ME functions */
494                  calc_cbp = calc_cbp_3dne;                  calc_cbp = calc_cbp_3dne;
495    
496                  sad16 = sad16_3dne;                  sad16 = sad16_3dne;
497                  sad8 = sad8_3dne;                  sad8 = sad8_3dne;
498                  sad16bi = sad16bi_3dne;                  sad16bi = sad16bi_3dne;
499                  sad8bi = sad8bi_3dne;                  sad8bi = sad8bi_3dne;
500                  dev16 = dev16_3dne;                  dev16 = dev16_3dne;
   
                 /* Interpolation */  
                 interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne;  
                 interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne;  
                 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;  
501          }          }
502            }
503    
504          if ((cpu_flags & XVID_CPU_SSE2)) {          if ((cpu_flags & XVID_CPU_SSE2)) {
505    
506                  calc_cbp = calc_cbp_sse2;                  calc_cbp = calc_cbp_sse2;
507    
508                  /* Quantization */                  /* Quantization */
509                  quant_intra   = quant_intra_sse2;                  quant_h263_intra   = quant_h263_intra_sse2;
510                  dequant_intra = dequant_intra_sse2;                  quant_h263_inter   = quant_h263_inter_sse2;
511                  quant_inter   = quant_inter_sse2;                  dequant_h263_intra = dequant_h263_intra_sse2;
512                  dequant_inter = dequant_inter_sse2;                  dequant_h263_inter = dequant_h263_inter_sse2;
513    
514  #if defined(EXPERIMENTAL_SSE2_CODE)                  /* SAD operators */
                 /* ME; slower than xmm */  
515                  sad16    = sad16_sse2;                  sad16    = sad16_sse2;
516                  dev16    = dev16_sse2;                  dev16    = dev16_sse2;
517  #endif  
518                  /* Forward and Inverse DCT */                  /* DCT operators
519  #if 0 /* Both function are known to be unprecise, better keep them deactivated */                   * no iDCT because it's not "Walken matching" */
520                  idct  = idct_sse2;                  fdct = fdct_sse2_skal;
                 fdct = fdct_sse2;  
 #endif  
521          }          }
522  #endif  #endif /* ARCH_IS_IA32 */
523    
524  #if defined(ARCH_IS_IA64)  #if defined(ARCH_IS_IA64)
525          if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */          if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */
# Line 511  Line 534 
534            sad8 = sad8_ia64;            sad8 = sad8_ia64;
535            dev16 = dev16_ia64;            dev16 = dev16_ia64;
536  /*        Halfpel8_Refine = Halfpel8_Refine_ia64; */  /*        Halfpel8_Refine = Halfpel8_Refine_ia64; */
537            quant_intra = quant_intra_ia64;            quant_h263_intra = quant_h263_intra_ia64;
538            dequant_intra = dequant_intra_ia64;            quant_h263_inter = quant_h263_inter_ia64;
539            quant_inter = quant_inter_ia64;            dequant_h263_intra = dequant_h263_intra_ia64;
540            dequant_inter = dequant_inter_ia64;            dequant_h263_inter = dequant_h263_inter_ia64;
541            transfer_8to16copy = transfer_8to16copy_ia64;            transfer_8to16copy = transfer_8to16copy_ia64;
542            transfer_16to8copy = transfer_16to8copy_ia64;            transfer_16to8copy = transfer_16to8copy_ia64;
543            transfer_8to16sub = transfer_8to16sub_ia64;            transfer_8to16sub = transfer_8to16sub_ia64;
544            transfer_8to16sub2 = transfer_8to16sub2_ia64;            transfer_8to16sub2 = transfer_8to16sub2_ia64;
545            transfer_16to8add = transfer_16to8add_ia64;            transfer_16to8add = transfer_16to8add_ia64;
546            transfer8x8_copy = transfer8x8_copy_ia64;            transfer8x8_copy = transfer8x8_copy_ia64;
           DPRINTF(DPRINTF_DEBUG, "Using IA-64 assembler routines.");  
547          }          }
548  #endif  #endif
549    
550  #if defined(ARCH_IS_PPC)  #if defined(ARCH_IS_PPC)
551          if ((cpu_flags & XVID_CPU_ASM))          if ((cpu_flags & XVID_CPU_ALTIVEC)) {
552          {            /* sad operators */
553                  calc_cbp = calc_cbp_ppc;            sad16 = sad16_altivec_c;
554          }            sad16bi = sad16bi_altivec_c;
555              sad8 = sad8_altivec_c;
556              dev16 = dev16_altivec_c;
557    
558              sse8_16bit = sse8_16bit_altivec_c;
559    
560              /* mem transfer */
561              transfer_8to16copy = transfer_8to16copy_altivec_c;
562              transfer_16to8copy = transfer_16to8copy_altivec_c;
563              transfer_8to16sub = transfer_8to16sub_altivec_c;
564              transfer_8to16subro = transfer_8to16subro_altivec_c;
565              transfer_8to16sub2 = transfer_8to16sub2_altivec_c;
566              transfer_16to8add = transfer_16to8add_altivec_c;
567              transfer8x8_copy = transfer8x8_copy_altivec_c;
568    
569          if ((cpu_flags & XVID_CPU_ALTIVEC))            /* Inverse DCT */
570          {            idct = idct_altivec_c;
571                  calc_cbp = calc_cbp_altivec;  
572                  fdct = fdct_altivec;            /* Interpolation */
573                  idct = idct_altivec;            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c;
574                  sadInit = sadInit_altivec;            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c;
575                  sad16 = sad16_altivec;            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c;
576                  sad8 = sad8_altivec;  
577                  dev16 = dev16_altivec;            interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c;
578              interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c;
579    
580              interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_altivec_c;
581    
582              /* Colorspace conversion */
583              bgra_to_yv12 = bgra_to_yv12_altivec_c;
584              abgr_to_yv12 = abgr_to_yv12_altivec_c;
585              rgba_to_yv12 = rgba_to_yv12_altivec_c;
586              argb_to_yv12 = argb_to_yv12_altivec_c;
587    
588              yuyv_to_yv12 = yuyv_to_yv12_altivec_c;
589              uyvy_to_yv12 = uyvy_to_yv12_altivec_c;
590    
591              yv12_to_yuyv = yv12_to_yuyv_altivec_c;
592              yv12_to_uyvy = yv12_to_uyvy_altivec_c;
593    
594              /* Quantization */
595              quant_h263_intra = quant_h263_intra_altivec_c;
596              quant_h263_inter = quant_h263_inter_altivec_c;
597              dequant_h263_intra = dequant_h263_intra_altivec_c;
598              dequant_h263_inter = dequant_h263_inter_altivec_c;
599          }          }
600  #endif  #endif
601    
# Line 547  Line 603 
603      xvid_debug = init->debug;      xvid_debug = init->debug;
604  #endif  #endif
605    
606      return 0;      return(0);
607  }  }
608    
609    
610  static int  static int
611  xvid_gbl_info(xvid_gbl_info_t * info)  xvid_gbl_info(xvid_gbl_info_t * info)
612  {  {
613          if (XVID_MAJOR(info->version) != 1) /* v1.x.x */          if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */
614                  return XVID_ERR_VERSION;                  return XVID_ERR_VERSION;
615    
616          info->actual_version = XVID_VERSION;          info->actual_version = XVID_VERSION;
617          info->build = "dev-api-4";          info->build = "xvid-1.0.0";
618          info->cpu_flags = detect_cpu_flags();          info->cpu_flags = detect_cpu_flags();
619    
620  #if defined(_SMP) && defined(WIN32)  #if defined(_SMP) && defined(WIN32)
# Line 580  Line 636 
636          int height2;          int height2;
637          IMAGE img;          IMAGE img;
638    
639          if (XVID_MAJOR(convert->version) != 1)   /* v1.x.x */          if (XVID_VERSION_MAJOR(convert->version) != 1)   /* v1.x.x */
640                return XVID_ERR_VERSION;                return XVID_ERR_VERSION;
641    
642  #if 0  #if 0
# Line 611  Line 667 
667          return 0;          return 0;
668  }  }
669    
   
   
 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)  
   
 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);  
                 }  
                 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)  
   
 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 = 0, retB = 0;  
         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)  
 {  
 #if defined(ARCH_IS_IA32)  
         int cpu_flags;  
 #endif  
   
         printf("XviD tests\n\n");  
   
 #if defined(ARCH_IS_IA32)  
         cpu_flags = detect_cpu_flags();  
 #endif  
   
         idct_int32_init();  
         emms();  
   
         srand(time(0));  
   
         /* fDCT test */  
         printf("--- fdct ---\n");  
                 test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* iDCT test */  
         printf("\n--- idct ---\n");  
                 test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Intra quantization test */  
         printf("\n--- quant intra ---\n");  
                 test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Inter quantization test */  
         printf("\n--- quant inter ---\n");  
                 test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Intra dequantization test */  
         printf("\n--- dequant intra ---\n");  
                 test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Inter dequantization test */  
         printf("\n--- dequant inter ---\n");  
                 test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Intra quantization test */  
         printf("\n--- quant4 intra ---\n");  
                 test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Inter quantization test */  
         printf("\n--- quant4 inter ---\n");  
                 test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Intra dequantization test */  
         printf("\n--- dequant4 intra ---\n");  
                 test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         /* Inter dequantization test */  
         printf("\n--- dequant4 inter ---\n");  
                 test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags);  
   
 #if defined(ARCH_IS_IA32)  
         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);  
 #endif  
   
         emms();  
   
         return 0;  
 }  
   
   
670  /*****************************************************************************  /*****************************************************************************
671   * XviD Global Entry point   * XviD Global Entry point
672   *   *
# Line 1006  Line 695 
695                  case XVID_GBL_CONVERT :                  case XVID_GBL_CONVERT :
696                          return xvid_gbl_convert((xvid_gbl_convert_t*)param1);                          return xvid_gbl_convert((xvid_gbl_convert_t*)param1);
697    
                 case XVID_GBL_TEST :  
                 {  
                         ptr_t flags = (ptr_t)param1;  
                         return xvid_init_test((int)flags);  
                 }  
698                  default :                  default :
699                          return XVID_ERR_FAIL;                          return XVID_ERR_FAIL;
700          }          }
Legend:
Removed from v.1.45.2.7  
changed lines
  Added in v.1.51
No admin address has been configured
 ViewVC Help
Powered by ViewVC 1.0.4