[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 1.70, Mon Jul 10 08:09:59 2006 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 54  Line 36 
36  #include "dct/fdct.h"  #include "dct/fdct.h"
37  #include "image/colorspace.h"  #include "image/colorspace.h"
38  #include "image/interpolate8x8.h"  #include "image/interpolate8x8.h"
 #include "image/reduced.h"  
39  #include "utils/mem_transfer.h"  #include "utils/mem_transfer.h"
40  #include "utils/mbfunctions.h"  #include "utils/mbfunctions.h"
41  #include "quant/quant_h263.h"  #include "quant/quant.h"
 #include "quant/quant_mpeg4.h"  
42  #include "motion/motion.h"  #include "motion/motion.h"
43    #include "motion/gmc.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(ARCH_IS_X86_64)) && defined(_MSC_VER)
56  #       include <windows.h>  #       include <windows.h>
57  #else  #elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || 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(ARCH_IS_X86_64)) && 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_X86_64) || 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    
127    
128  /* detect cpu flags  */  /* detect cpu flags  */
129  static unsigned int  static unsigned int
130  detect_cpu_flags()  detect_cpu_flags(void)
131  {  {
132          /* enable native assembly optimizations by default */          /* enable native assembly optimizations by default */
133          unsigned int cpu_flags = XVID_CPU_ASM;          unsigned int cpu_flags = XVID_CPU_ASM;
134    
135  #if defined(ARCH_IS_IA32)  #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)
136          cpu_flags |= check_cpu_features();          cpu_flags |= check_cpu_features();
137          if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger))          if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger))
138                  cpu_flags &= ~XVID_CPU_SSE;                  cpu_flags &= ~XVID_CPU_SSE;
# 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)
 {  
         int cpu_flags;  
   
         /* Inform the client the API version */  
         init_param->api_version = API_VERSION;  
   
         /* Inform the client the core build - unused because we're still alpha */  
         init_param->core_build = 1000;  
   
         /* Do we have to force CPU features  ? */  
         if ((init_param->cpu_flags & XVID_CPU_FORCE)) {  
   
                 cpu_flags = init_param->cpu_flags;  
   
         } else {  
   
                 cpu_flags = detect_cpu_flags();  
         }  
   
         if ((init_param->cpu_flags & XVID_CPU_CHKONLY))  
170          {          {
171                  init_param->cpu_flags = cpu_flags;          unsigned int cpu_flags;
                 return XVID_ERR_OK;  
         }  
172    
173          init_param->cpu_flags = cpu_flags;          if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */
174                    return XVID_ERR_VERSION;
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 209  Line 184 
184          idct = idct_int32;          idct = idct_int32;
185    
186          /* Only needed on PPC Altivec archs */          /* Only needed on PPC Altivec archs */
187          sadInit = 0;          sadInit = NULL;
188    
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 231  Line 211 
211          transfer_8to16sub  = transfer_8to16sub_c;          transfer_8to16sub  = transfer_8to16sub_c;
212          transfer_8to16subro  = transfer_8to16subro_c;          transfer_8to16subro  = transfer_8to16subro_c;
213          transfer_8to16sub2 = transfer_8to16sub2_c;          transfer_8to16sub2 = transfer_8to16sub2_c;
214            transfer_8to16sub2ro = transfer_8to16sub2ro_c;
215          transfer_16to8add  = transfer_16to8add_c;          transfer_16to8add  = transfer_16to8add_c;
216          transfer8x8_copy   = transfer8x8_copy_c;          transfer8x8_copy   = transfer8x8_copy_c;
217            transfer8x4_copy   = transfer8x4_copy_c;
218    
219          /* Interlacing functions */          /* Interlacing functions */
220          MBFieldTest = MBFieldTest_c;          MBFieldTest = MBFieldTest_c;
# Line 242  Line 224 
224          interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_c;          interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_c;
225          interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;          interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;
226    
227            interpolate8x4_halfpel_h  = interpolate8x4_halfpel_h_c;
228            interpolate8x4_halfpel_v  = interpolate8x4_halfpel_v_c;
229            interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_c;
230    
231            interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_c;
232            interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_c;
233            interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_c;
234            interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_c;
235    
236          interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c;          interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c;
237          interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c;          interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c;
238          interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c;          interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c;
# Line 256  Line 247 
247          interpolate8x8_avg2 = interpolate8x8_avg2_c;          interpolate8x8_avg2 = interpolate8x8_avg2_c;
248          interpolate8x8_avg4 = interpolate8x8_avg4_c;          interpolate8x8_avg4 = interpolate8x8_avg4_c;
249    
250          /* reduced resoltuion */          /* postprocessing */
251          copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C;          image_brightness = image_brightness_c;
         add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C;  
         vfilter_31 = xvid_VFilter_31_C;  
         hfilter_31 = xvid_HFilter_31_C;  
         filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C;  
         filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C;  
252    
253          /* Initialize internal colorspace transformation tables */          /* Initialize internal colorspace transformation tables */
254          colorspace_init();          colorspace_init();
# Line 275  Line 261 
261          bgra_to_yv12    = bgra_to_yv12_c;          bgra_to_yv12    = bgra_to_yv12_c;
262          abgr_to_yv12    = abgr_to_yv12_c;          abgr_to_yv12    = abgr_to_yv12_c;
263          rgba_to_yv12    = rgba_to_yv12_c;          rgba_to_yv12    = rgba_to_yv12_c;
264            argb_to_yv12    = argb_to_yv12_c;
265          yuyv_to_yv12    = yuyv_to_yv12_c;          yuyv_to_yv12    = yuyv_to_yv12_c;
266          uyvy_to_yv12    = uyvy_to_yv12_c;          uyvy_to_yv12    = uyvy_to_yv12_c;
267    
# Line 284  Line 271 
271          bgrai_to_yv12   = bgrai_to_yv12_c;          bgrai_to_yv12   = bgrai_to_yv12_c;
272          abgri_to_yv12   = abgri_to_yv12_c;          abgri_to_yv12   = abgri_to_yv12_c;
273          rgbai_to_yv12   = rgbai_to_yv12_c;          rgbai_to_yv12   = rgbai_to_yv12_c;
274            argbi_to_yv12   = argbi_to_yv12_c;
275          yuyvi_to_yv12   = yuyvi_to_yv12_c;          yuyvi_to_yv12   = yuyvi_to_yv12_c;
276          uyvyi_to_yv12   = uyvyi_to_yv12_c;          uyvyi_to_yv12   = uyvyi_to_yv12_c;
277    
   
278          /* All colorspace transformation functions YV12->User format */          /* All colorspace transformation functions YV12->User format */
279          yv12_to_rgb555  = yv12_to_rgb555_c;          yv12_to_rgb555  = yv12_to_rgb555_c;
280          yv12_to_rgb565  = yv12_to_rgb565_c;          yv12_to_rgb565  = yv12_to_rgb565_c;
# Line 295  Line 282 
282          yv12_to_bgra    = yv12_to_bgra_c;          yv12_to_bgra    = yv12_to_bgra_c;
283          yv12_to_abgr    = yv12_to_abgr_c;          yv12_to_abgr    = yv12_to_abgr_c;
284          yv12_to_rgba    = yv12_to_rgba_c;          yv12_to_rgba    = yv12_to_rgba_c;
285            yv12_to_argb    = yv12_to_argb_c;
286          yv12_to_yuyv    = yv12_to_yuyv_c;          yv12_to_yuyv    = yv12_to_yuyv_c;
287          yv12_to_uyvy    = yv12_to_uyvy_c;          yv12_to_uyvy    = yv12_to_uyvy_c;
288    
# Line 304  Line 292 
292          yv12_to_bgrai   = yv12_to_bgrai_c;          yv12_to_bgrai   = yv12_to_bgrai_c;
293          yv12_to_abgri   = yv12_to_abgri_c;          yv12_to_abgri   = yv12_to_abgri_c;
294          yv12_to_rgbai   = yv12_to_rgbai_c;          yv12_to_rgbai   = yv12_to_rgbai_c;
295            yv12_to_argbi   = yv12_to_argbi_c;
296          yv12_to_yuyvi   = yv12_to_yuyvi_c;          yv12_to_yuyvi   = yv12_to_yuyvi_c;
297          yv12_to_uyvyi   = yv12_to_uyvyi_c;          yv12_to_uyvyi   = yv12_to_uyvyi_c;
298    
# Line 315  Line 304 
304          sad8bi   = sad8bi_c;          sad8bi   = sad8bi_c;
305          dev16    = dev16_c;          dev16    = dev16_c;
306          sad16v   = sad16v_c;          sad16v   = sad16v_c;
307            sse8_16bit = sse8_16bit_c;
308            sse8_8bit  = sse8_8bit_c;
309    
310  //      Halfpel8_Refine = Halfpel8_Refine_c;          init_GMC(cpu_flags);
311    
312  #if defined(ARCH_IS_IA32)  #if defined(ARCH_IS_IA32)
313    
         if ((cpu_flags & XVID_CPU_ASM))  
         {  
                 vfilter_31 = xvid_VFilter_31_x86;  
                 hfilter_31 = xvid_HFilter_31_x86;  
         }  
   
314          if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) ||          if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) ||
315                  (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) ||                  (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) ||
316                  (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2))                  (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2))
# Line 337  Line 322 
322          if ((cpu_flags & XVID_CPU_MMX)) {          if ((cpu_flags & XVID_CPU_MMX)) {
323    
324                  /* Forward and Inverse Discrete Cosine Transformation functions */                  /* Forward and Inverse Discrete Cosine Transformation functions */
325                  fdct = fdct_mmx;                  fdct = fdct_mmx_skal;
326                  idct = idct_mmx;                  idct = idct_mmx;
327    
328                    /* Qpel stuff */
329                    xvid_QP_Funcs = &xvid_QP_Funcs_mmx;
330                    xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx;
331    
332                  /* Quantization related functions */                  /* Quantization related functions */
333                  quant_intra   = quant_intra_mmx;                  quant_h263_intra   = quant_h263_intra_mmx;
334                  dequant_intra = dequant_intra_mmx;                  quant_h263_inter   = quant_h263_inter_mmx;
335                  quant_inter   = quant_inter_mmx;                  dequant_h263_intra = dequant_h263_intra_mmx;
336                  dequant_inter = dequant_inter_mmx;                  dequant_h263_inter = dequant_h263_inter_mmx;
337    
338                  quant4_intra   = quant4_intra_mmx;                  quant_mpeg_intra   = quant_mpeg_intra_mmx;
339                  dequant4_intra = dequant4_intra_mmx;                  quant_mpeg_inter   = quant_mpeg_inter_mmx;
340                  quant4_inter   = quant4_inter_mmx;                  dequant_mpeg_intra = dequant_mpeg_intra_mmx;
341                  dequant4_inter = dequant4_inter_mmx;                  dequant_mpeg_inter = dequant_mpeg_inter_mmx;
342    
343                  /* Block related functions */                  /* Block related functions */
344                  transfer_8to16copy = transfer_8to16copy_mmx;                  transfer_8to16copy = transfer_8to16copy_mmx;
# Line 359  Line 348 
348                  transfer_8to16sub2 = transfer_8to16sub2_mmx;                  transfer_8to16sub2 = transfer_8to16sub2_mmx;
349                  transfer_16to8add  = transfer_16to8add_mmx;                  transfer_16to8add  = transfer_16to8add_mmx;
350                  transfer8x8_copy   = transfer8x8_copy_mmx;                  transfer8x8_copy   = transfer8x8_copy_mmx;
351                    transfer8x4_copy   = transfer8x4_copy_mmx;
352    
353                  /* Interlacing Functions */                  /* Interlacing Functions */
354                  MBFieldTest = MBFieldTest_mmx;                  MBFieldTest = MBFieldTest_mmx;
# Line 368  Line 358 
358                  interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_mmx;                  interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_mmx;
359                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;
360    
361                    interpolate8x4_halfpel_h  = interpolate8x4_halfpel_h_mmx;
362                    interpolate8x4_halfpel_v  = interpolate8x4_halfpel_v_mmx;
363                    interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_mmx;
364    
365                    interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_mmx;
366                    interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_mmx;
367                    interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_mmx;
368                    interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_mmx;
369    
370                  interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx;                  interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx;
371                  interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx;                  interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx;
372    
373                  interpolate8x8_avg2 = interpolate8x8_avg2_mmx;                  interpolate8x8_avg2 = interpolate8x8_avg2_mmx;
374                  interpolate8x8_avg4 = interpolate8x8_avg4_mmx;                  interpolate8x8_avg4 = interpolate8x8_avg4_mmx;
375    
376                  /* reduced resolution */                  /* postprocessing */
377                  copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx;                  image_brightness = image_brightness_mmx;
                 add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx;  
                 hfilter_31 = xvid_HFilter_31_mmx;  
                 filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx;  
                 filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx;  
378    
379                  /* image input xxx_to_yv12 related functions */                  /* image input xxx_to_yv12 related functions */
380                  yv12_to_yv12  = yv12_to_yv12_mmx;                  yv12_to_yv12  = yv12_to_yv12_mmx;
# Line 405  Line 400 
400                  sad8bi  = sad8bi_mmx;                  sad8bi  = sad8bi_mmx;
401                  dev16    = dev16_mmx;                  dev16    = dev16_mmx;
402                  sad16v   = sad16v_mmx;                  sad16v   = sad16v_mmx;
403                    sse8_16bit = sse8_16bit_mmx;
404                    sse8_8bit  = sse8_8bit_mmx;
405          }          }
406    
407          /* 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 420 
420    
421          if ((cpu_flags & XVID_CPU_MMXEXT)) {          if ((cpu_flags & XVID_CPU_MMXEXT)) {
422    
423                  /* Inverse DCT */                  /* DCT */
424                    fdct = fdct_xmm_skal;
425                  idct = idct_xmm;                  idct = idct_xmm;
426    
427                  /* Interpolation */                  /* Interpolation */
# Line 431  Line 429 
429                  interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_xmm;                  interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_xmm;
430                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm;                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm;
431    
432                  /* reduced resolution */                  interpolate8x4_halfpel_h  = interpolate8x4_halfpel_h_xmm;
433                  copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm;                  interpolate8x4_halfpel_v  = interpolate8x4_halfpel_v_xmm;
434                  add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm;                  interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_xmm;
435    
436                    interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_xmm;
437                    interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm;
438                    interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm;
439                    interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm;
440    
441                  /* Quantization */                  /* Quantization */
442                  quant4_intra = quant4_intra_xmm;                  quant_mpeg_inter = quant_mpeg_inter_xmm;
                 quant4_inter = quant4_inter_xmm;  
443    
444                  dequant_intra = dequant_intra_xmm;                  dequant_h263_intra = dequant_h263_intra_xmm;
445                  dequant_inter = dequant_inter_xmm;                  dequant_h263_inter = dequant_h263_inter_xmm;
446    
447                  /* Buffer transfer */                  /* Buffer transfer */
448                  transfer_8to16sub2 = transfer_8to16sub2_xmm;                  transfer_8to16sub2 = transfer_8to16sub2_xmm;
449                    transfer_8to16sub2ro = transfer_8to16sub2ro_xmm;
450    
451                  /* Colorspace transformation */                  /* Colorspace transformation */
452                  yv12_to_yv12  = yv12_to_yv12_xmm;                  yv12_to_yv12  = yv12_to_yv12_xmm;
# Line 465  Line 468 
468                  interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;                  interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;
469                  interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;                  interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;
470                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;                  interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;
471    
472                    interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dn;
473                    interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dn;
474                    interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dn;
475          }          }
476    
477          if ((cpu_flags & XVID_CPU_3DNOWEXT)) {          if ((cpu_flags & XVID_CPU_3DNOWEXT)) {
478    
                 /* Inverse DCT */  
                 idct =  idct_3dne;  
   
479                  /* Buffer transfer */                  /* Buffer transfer */
480                  transfer_8to16copy =  transfer_8to16copy_3dne;                  transfer_8to16copy =  transfer_8to16copy_3dne;
481                  transfer_16to8copy = transfer_16to8copy_3dne;                  transfer_16to8copy = transfer_16to8copy_3dne;
482                  transfer_8to16sub =  transfer_8to16sub_3dne;                  transfer_8to16sub =  transfer_8to16sub_3dne;
483                  transfer_8to16subro =  transfer_8to16subro_3dne;                  transfer_8to16subro =  transfer_8to16subro_3dne;
                 transfer_8to16sub2 =  transfer_8to16sub2_3dne;  
484                  transfer_16to8add = transfer_16to8add_3dne;                  transfer_16to8add = transfer_16to8add_3dne;
485                  transfer8x8_copy = transfer8x8_copy_3dne;                  transfer8x8_copy = transfer8x8_copy_3dne;
486                    transfer8x4_copy = transfer8x4_copy_3dne;
487    
488                    if ((cpu_flags & XVID_CPU_MMXEXT)) {
489                            /* Inverse DCT */
490                            idct =  idct_3dne;
491    
492                            /* Buffer transfer */
493                            transfer_8to16sub2 =  transfer_8to16sub2_3dne;
494    
495                            /* Interpolation */
496                            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne;
497                            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne;
498                            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;
499    
500                            interpolate8x4_halfpel_h = interpolate8x4_halfpel_h_3dne;
501                            interpolate8x4_halfpel_v = interpolate8x4_halfpel_v_3dne;
502                            interpolate8x4_halfpel_hv = interpolate8x4_halfpel_hv_3dne;
503    
504                  /* Quantization */                  /* Quantization */
505                  dequant4_intra = dequant4_intra_3dne;                          quant_h263_intra = quant_h263_intra_3dne;               /* cmov only */
506                  dequant4_inter = dequant4_inter_3dne;                          quant_h263_inter = quant_h263_inter_3dne;
507                  quant_intra = quant_intra_3dne;                          dequant_mpeg_intra = dequant_mpeg_intra_3dne;   /* cmov only */
508                  quant_inter = quant_inter_3dne;                          dequant_mpeg_inter = dequant_mpeg_inter_3dne;
509                  dequant_intra = dequant_intra_3dne;                          dequant_h263_intra = dequant_h263_intra_3dne;
510                  dequant_inter = dequant_inter_3dne;                          dequant_h263_inter = dequant_h263_inter_3dne;
511    
512                  /* ME functions */                  /* ME functions */
513                  calc_cbp = calc_cbp_3dne;                  calc_cbp = calc_cbp_3dne;
514    
515                  sad16 = sad16_3dne;                  sad16 = sad16_3dne;
516                  sad8 = sad8_3dne;                  sad8 = sad8_3dne;
517                  sad16bi = sad16bi_3dne;                  sad16bi = sad16bi_3dne;
518                  sad8bi = sad8bi_3dne;                  sad8bi = sad8bi_3dne;
519                  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;  
520          }          }
521            }
522    
523          if ((cpu_flags & XVID_CPU_SSE2)) {          if ((cpu_flags & XVID_CPU_SSE2)) {
524    
525                  calc_cbp = calc_cbp_sse2;                  calc_cbp = calc_cbp_sse2;
526    
527                  /* Quantization */                  /* Quantization */
528                  quant_intra   = quant_intra_sse2;                  quant_h263_intra   = quant_h263_intra_sse2;
529                  dequant_intra = dequant_intra_sse2;                  quant_h263_inter   = quant_h263_inter_sse2;
530                  quant_inter   = quant_inter_sse2;                  dequant_h263_intra = dequant_h263_intra_sse2;
531                  dequant_inter = dequant_inter_sse2;                  dequant_h263_inter = dequant_h263_inter_sse2;
532    
533  #if defined(EXPERIMENTAL_SSE2_CODE)                  /* SAD operators */
                 /* ME; slower than xmm */  
534                  sad16    = sad16_sse2;                  sad16    = sad16_sse2;
535                  dev16    = dev16_sse2;                  dev16    = dev16_sse2;
536  #endif  
537                  /* Forward and Inverse DCT */                  /* DCT operators */
538                  idct  = idct_sse2;                  fdct = fdct_sse2_skal;
539                  fdct = fdct_sse2;      /* idct = idct_sse2_skal; */   /* Is now IEEE1180 and Walken compliant. Disabled until fully tested. */
540    
541                    /* postprocessing */
542                    image_brightness = image_brightness_sse2;
543          }          }
544  #endif  #endif /* ARCH_IS_IA32 */
545    
546  #if defined(ARCH_IS_IA64)  #if defined(ARCH_IS_IA64)
547          if ((cpu_flags & XVID_CPU_ASM)) { //use assembler routines?          if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */
548            idct_ia64_init();            idct_ia64_init();
549            fdct = fdct_ia64;            fdct = fdct_ia64;
550            idct = idct_ia64;   //not yet working, crashes            idct = idct_ia64;   /*not yet working, crashes */
551            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;            interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
552            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;            interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
553            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;            interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
# Line 537  Line 555 
555            sad16bi = sad16bi_ia64;            sad16bi = sad16bi_ia64;
556            sad8 = sad8_ia64;            sad8 = sad8_ia64;
557            dev16 = dev16_ia64;            dev16 = dev16_ia64;
558  //        Halfpel8_Refine = Halfpel8_Refine_ia64;  /*        Halfpel8_Refine = Halfpel8_Refine_ia64; */
559            quant_intra = quant_intra_ia64;            quant_h263_intra = quant_h263_intra_ia64;
560            dequant_intra = dequant_intra_ia64;            quant_h263_inter = quant_h263_inter_ia64;
561            quant_inter = quant_inter_ia64;            dequant_h263_intra = dequant_h263_intra_ia64;
562            dequant_inter = dequant_inter_ia64;            dequant_h263_inter = dequant_h263_inter_ia64;
563            transfer_8to16copy = transfer_8to16copy_ia64;            transfer_8to16copy = transfer_8to16copy_ia64;
564            transfer_16to8copy = transfer_16to8copy_ia64;            transfer_16to8copy = transfer_16to8copy_ia64;
565            transfer_8to16sub = transfer_8to16sub_ia64;            transfer_8to16sub = transfer_8to16sub_ia64;
566            transfer_8to16sub2 = transfer_8to16sub2_ia64;            transfer_8to16sub2 = transfer_8to16sub2_ia64;
567            transfer_16to8add = transfer_16to8add_ia64;            transfer_16to8add = transfer_16to8add_ia64;
568            transfer8x8_copy = transfer8x8_copy_ia64;            transfer8x8_copy = transfer8x8_copy_ia64;
           DEBUG("Using IA-64 assembler routines.\n");  
569          }          }
570  #endif  #endif
571    
572  #if defined(ARCH_IS_PPC)  #if defined(ARCH_IS_PPC)
573          if ((cpu_flags & XVID_CPU_ASM))          if ((cpu_flags & XVID_CPU_ALTIVEC)) {
574          {            /* sad operators */
575                  calc_cbp = calc_cbp_ppc;                    sad16 = sad16_altivec_c;
576          }                    sad16bi = sad16bi_altivec_c;
577                      sad8 = sad8_altivec_c;
578          if ((cpu_flags & XVID_CPU_ALTIVEC))                    dev16 = dev16_altivec_c;
579          {  
580                  calc_cbp = calc_cbp_altivec;            sse8_16bit = sse8_16bit_altivec_c;
581                  fdct = fdct_altivec;  
582                  idct = idct_altivec;            /* mem transfer */
583                  sadInit = sadInit_altivec;            transfer_8to16copy = transfer_8to16copy_altivec_c;
584                  sad16 = sad16_altivec;            transfer_16to8copy = transfer_16to8copy_altivec_c;
585                  sad8 = sad8_altivec;            transfer_8to16sub = transfer_8to16sub_altivec_c;
586                  dev16 = dev16_altivec;            transfer_8to16subro = transfer_8to16subro_altivec_c;
587          }            transfer_8to16sub2 = transfer_8to16sub2_altivec_c;
588  #endif            transfer_16to8add = transfer_16to8add_altivec_c;
589              transfer8x8_copy = transfer8x8_copy_altivec_c;
         return XVID_ERR_OK;  
 }  
590    
591              /* Inverse DCT */
592              idct = idct_altivec_c;
593    
594              /* Interpolation */
595              interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c;
596              interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c;
597              interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c;
598    
599              interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c;
600              interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c;
601    
602                      interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c;
603                      interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c;
604                      interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c;
605                      interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c;
606    
607              /* Colorspace conversion */
608              bgra_to_yv12 = bgra_to_yv12_altivec_c;
609              abgr_to_yv12 = abgr_to_yv12_altivec_c;
610              rgba_to_yv12 = rgba_to_yv12_altivec_c;
611              argb_to_yv12 = argb_to_yv12_altivec_c;
612    
613  static int            yuyv_to_yv12 = yuyv_to_yv12_altivec_c;
614  xvid_init_convert(XVID_INIT_CONVERTINFO* convert)            uyvy_to_yv12 = uyvy_to_yv12_altivec_c;
 {  
         // 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;  
         }  
615    
616              yv12_to_yuyv = yv12_to_yuyv_altivec_c;
617              yv12_to_uyvy = yv12_to_uyvy_altivec_c;
618    
619          emms();            /* Quantization */
620          return XVID_ERR_OK;            quant_h263_intra = quant_h263_intra_altivec_c;
621              quant_h263_inter = quant_h263_inter_altivec_c;
622              dequant_h263_intra = dequant_h263_intra_altivec_c;
623              dequant_h263_inter = dequant_h263_inter_altivec_c;
624    
625                      dequant_mpeg_intra = dequant_mpeg_intra_altivec_c;
626                      dequant_mpeg_inter = dequant_mpeg_inter_altivec_c;
627    
628                      /* Qpel stuff */
629                      xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C;
630                      xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C;
631  }  }
632    #endif
633    
634    #if defined(ARCH_IS_X86_64)
635            /* For now, only XVID_CPU_ASM is looked for, so user can still
636             * disable asm usage the usual way. When Intel EMT64 cpus will
637             * be out, maybe we'll have to check more precisely what cpu
638             * features there really are. */
639            if (cpu_flags & XVID_CPU_ASM) {
640                    /* SIMD state flusher */
641                    emms = emms_mmx;
642    
643                    /* DCT operators */
644                    fdct = fdct_skal_x86_64;
645                    idct = idct_x86_64;
646    
647                    /* SAD operators */
648                    sad16      = sad16_x86_64;
649                    sad8       = sad8_x86_64;
650                    sad16bi    = sad16bi_x86_64;
651                    sad8bi     = sad8bi_x86_64;
652                    dev16      = dev16_x86_64;
653                    sad16v     = sad16v_x86_64;
654                    sse8_16bit = sse8_16bit_x86_64;
655                    sse8_8bit  = sse8_8bit_x86_64;
656    
657                    /* Interpolation operators */
658                    interpolate8x8_halfpel_h  = interpolate8x8_halfpel_h_x86_64;
659                    interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_x86_64;
660                    interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_x86_64;
661    
662                    interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_x86_64;
663                    interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_x86_64;
664                    interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_x86_64;
665                    interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_x86_64;
666    
667  void fill8(uint8_t * block, int size, int value)                  interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_x86_64;
668  {                  interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_x86_64;
         int i;  
         for (i = 0; i < size; i++)  
                 block[i] = value;  
 }  
669    
670  void fill16(int16_t * block, int size, int value)                  interpolate8x8_avg2 = interpolate8x8_avg2_x86_64;
671  {                  interpolate8x8_avg4 = interpolate8x8_avg4_x86_64;
         int i;  
         for (i = 0; i < size; i++)  
                 block[i] = value;  
 }  
672    
673  #define RANDOM(min,max) min + (rand() % (max-min))                  /* Quantization related functions */
674                    quant_h263_intra   = quant_h263_intra_x86_64;
675                    quant_h263_inter   = quant_h263_inter_x86_64;
676                    dequant_h263_intra = dequant_h263_intra_x86_64;
677                    dequant_h263_inter = dequant_h263_inter_x86_64;
678                    /*quant_mpeg_intra   = quant_mpeg_intra_x86_64; fix me! */
679                    quant_mpeg_inter   = quant_mpeg_inter_x86_64;
680                    dequant_mpeg_intra   = dequant_mpeg_intra_x86_64;
681                    dequant_mpeg_inter   = dequant_mpeg_inter_x86_64;
682    
683  void random8(uint8_t * block, int size, int min, int max)                  /* Block related functions */
684  {                  transfer_8to16copy  = transfer_8to16copy_x86_64;
685          int i;                  transfer_16to8copy  = transfer_16to8copy_x86_64;
686          for (i = 0; i < size; i++)                  transfer_8to16sub   = transfer_8to16sub_x86_64;
687                  block[i] = RANDOM(min,max);                  transfer_8to16subro = transfer_8to16subro_x86_64;
688  }                  transfer_8to16sub2  = transfer_8to16sub2_x86_64;
689                    transfer_8to16sub2ro= transfer_8to16sub2ro_x86_64;
690                    transfer_16to8add   = transfer_16to8add_x86_64;
691                    transfer8x8_copy    = transfer8x8_copy_x86_64;
692    
693                    /* Qpel stuff */
694                    xvid_QP_Funcs = &xvid_QP_Funcs_x86_64;
695                    xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_x86_64;
696    
697  void random16(int16_t * block, int size, int min, int max)                  /* Interlacing Functions */
698  {                  MBFieldTest = MBFieldTest_x86_64;
         int i;  
         for (i = 0; i < size; i++)  
                 block[i] = RANDOM(min,max);  
699  }  }
700    #endif
701    
702  int compare16(const int16_t * blockA, const int16_t * blockB, int size)  #if defined(_DEBUG)
703  {      xvid_debug = init->debug;
704          int i;  #endif
         for (i = 0; i < size; i++)  
                 if (blockA[i] != blockB[i])  
                         return 1;  
   
         return 0;  
 }  
705    
706  int diff16(const int16_t * blockA, const int16_t * blockB, int size)      return(0);
 {  
         int i, diff = 0;  
         for (i = 0; i < size; i++)  
                 diff += ABS(blockA[i]-blockB[i]);  
         return diff;  
707  }  }
708    
709    
710  #define XVID_TEST_RANDOM        0x00000001      /* random input data */  static int
711  #define XVID_TEST_VERBOSE       0x00000002      /* verbose error output */  xvid_gbl_info(xvid_gbl_info_t * info)
   
   
 #define TEST_FORWARD    0x00000001      /* intra */  
 #define TEST_FDCT  (TEST_FORWARD)  
 #define TEST_IDCT  (0)  
   
 static int test_transform(void * funcA, void * funcB, const char * nameB,  
                                    int test, int flags)  
 {  
         int i;  
         int64_t timeSTART;  
         int64_t timeA = 0;  
         int64_t timeB = 0;  
         DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE);  
         DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE);  
         int min, max;  
         int count = 0;  
   
         int tmp;  
         int min_error = 0x10000*64;  
         int max_error = 0;  
   
   
         if ((test & TEST_FORWARD))      /* forward */  
712          {          {
713                  min = -256;          if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */
714                  max = 255;                  return XVID_ERR_VERSION;
         }else{          /* inverse */  
                 min = -2048;  
                 max = 2047;  
         }  
715    
716          for (i = 0; i < 64*64; i++)          info->actual_version = XVID_VERSION;
717          {          info->build = "xvid-1.2.0-dev";
718                  if ((flags & XVID_TEST_RANDOM))          info->cpu_flags = detect_cpu_flags();
719                  {    info->num_threads = 0;
                         random16(arrayA, 64, min, max);  
                 }else{  
                         fill16(arrayA, 64, i);  
                 }  
                 memcpy(arrayB, arrayA, 64*sizeof(int16_t));  
720    
721                  if ((test & TEST_FORWARD))  #if defined(WIN32)
722                  {                  {
723                          timeSTART = read_counter();      DWORD dwProcessAffinityMask, dwSystemAffinityMask;
724                          ((fdctFunc*)funcA)(arrayA);      if (GetProcessAffinityMask(GetCurrentProcess(), &dwProcessAffinityMask, &dwSystemAffinityMask)) {
725                          timeA += read_counter() - timeSTART;        int i;
726          for(i=0; i<32; i++) {
727                          timeSTART = read_counter();          if ((dwProcessAffinityMask & (1<<i)))
728                          ((fdctFunc*)funcB)(arrayB);            info->num_threads++;
                         timeB += read_counter() - timeSTART;  
729                  }                  }
                 else  
                 {  
                         timeSTART = read_counter();  
                         ((idctFunc*)funcA)(arrayA);  
                         timeA += read_counter() - timeSTART;  
   
                         timeSTART = read_counter();  
                         ((idctFunc*)funcB)(arrayB);  
                         timeB += read_counter() - timeSTART;  
730                  }                  }
   
                 tmp = diff16(arrayA, arrayB, 64) / 64;  
                 if (tmp > max_error)  
                         max_error = tmp;  
                 if (tmp < min_error)  
                         min_error = tmp;  
   
                 count++;  
731          }          }
732    #endif
         /* 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);  
733    
734          return 0;          return 0;
735  }  }
736    
737    
738  #define TEST_QUANT      0x00000001      /* forward quantization */  static int
739  #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 */  
740                          {                          {
741                                  timeSTART = read_counter();          int width;
742                                  ((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q);          int height;
743                                  timeA += read_counter() - timeSTART;          int width2;
744            int height2;
745            IMAGE img;
746    
747                                  timeSTART = read_counter();          if (XVID_VERSION_MAJOR(convert->version) != 1)   /* v1.x.x */
748                                  ((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;  
749    
750                                  timeSTART = read_counter();  #if 0
751                                  retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q);          const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP);
752                                  timeB += read_counter() - timeSTART;  #endif
753                          }          width = convert->width;
754            height = convert->height;
755            width2 = convert->width/2;
756            height2 = convert->height/2;
757    
758                          /* 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))  
759                          {                          {
760                                  errors++;                  case XVID_CSP_YV12 :
761                                  if ((flags & XVID_TEST_VERBOSE))                          img.y = convert->input.plane[0];
762                                          printf("%s error: q=%i, i=%i\n", nameB, q, i);                          img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height;
763                          }                          img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2;
764                            image_output(&img, width, height, width,
765                                                    (uint8_t**)convert->output.plane, convert->output.stride,
766                                                    convert->output.csp, convert->interlacing);
767                            break;
768    
769                          count++;                  default :
770                  }                          return XVID_ERR_FORMAT;
771          }          }
772    
         printf("%s:\t%i", nameB, (int)(timeB / count));  
         if (errors>0)  
                 printf("\t(%i errors out of %i)", errors, count);  
         printf("\n");  
773    
774            emms();
775          return 0;          return 0;
776  }  }
777    
778    /*****************************************************************************
779     * XviD Global Entry point
780  int xvid_init_test(int flags)   *
781  {   * Well this function initialize all internal function pointers according
782          int cpu_flags;   * to the CPU features forced by the library client or autodetected (depending
783     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
784          srand(time(0));   * image colorspace transformation tables.
785     *
786          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;  
 }  
787    
788    
789  int  int
790  xvid_init(void *handle,  xvid_global(void *handle,
791                    int opt,                    int opt,
792                    void *param1,                    void *param1,
793                    void *param2)                    void *param2)
794  {  {
795          switch(opt)          switch(opt)
796          {          {
797                  case XVID_INIT_INIT :                  case XVID_GBL_INIT :
798                          return xvid_init_init((XVID_INIT_PARAM*)param1);                          return xvid_gbl_init((xvid_gbl_init_t*)param1);
799    
800                  case XVID_INIT_CONVERT :          case XVID_GBL_INFO :
801                          return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1);              return xvid_gbl_info((xvid_gbl_info_t*)param1);
802    
803                  case XVID_INIT_TEST :                  case XVID_GBL_CONVERT :
804                          return xvid_init_test((int)param1);                          return xvid_gbl_convert((xvid_gbl_convert_t*)param1);
805    
806                  default :                  default :
807                          return XVID_ERR_FAIL;                          return XVID_ERR_FAIL;
# Line 969  Line 825 
825                          void *param2)                          void *param2)
826  {  {
827          switch (opt) {          switch (opt) {
         case XVID_DEC_DECODE:  
                 return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2);  
   
828          case XVID_DEC_CREATE:          case XVID_DEC_CREATE:
829                  return decoder_create((XVID_DEC_PARAM *) param1);                  return decoder_create((xvid_dec_create_t *) param1);
830    
831          case XVID_DEC_DESTROY:          case XVID_DEC_DESTROY:
832                  return decoder_destroy((DECODER *) handle);                  return decoder_destroy((DECODER *) handle);
833    
834            case XVID_DEC_DECODE:
835                    return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2);
836    
837          default:          default:
838                  return XVID_ERR_FAIL;                  return XVID_ERR_FAIL;
839          }          }
# Line 1003  Line 859 
859          switch (opt) {          switch (opt) {
860          case XVID_ENC_ENCODE:          case XVID_ENC_ENCODE:
861    
862                  if (((Encoder *) handle)->mbParam.max_bframes >= 0)                  return enc_encode((Encoder *) handle,
863                  return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1,                                                            (xvid_enc_frame_t *) param1,
864                                                            (XVID_ENC_STATS *) param2);                                                            (xvid_enc_stats_t *) param2);
                 else  
                 return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1,  
                                                           (XVID_ENC_STATS *) param2);  
865    
866          case XVID_ENC_CREATE:          case XVID_ENC_CREATE:
867                  return encoder_create((XVID_ENC_PARAM *) param1);                  return enc_create((xvid_enc_create_t *) param1);
868    
869          case XVID_ENC_DESTROY:          case XVID_ENC_DESTROY:
870                  return encoder_destroy((Encoder *) handle);                  return enc_destroy((Encoder *) handle);
871    
872          default:          default:
873                  return XVID_ERR_FAIL;                  return XVID_ERR_FAIL;

Legend:
Removed from v.1.42  
changed lines
  Added in v.1.70

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