1/*
2 * Mesa 3-D graphics library
3 * Version:  6.1
4 *
5 * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
20 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 *
24 * Authors:
25 *    Gareth Hughes
26 */
27
28#include "main/glheader.h"
29#include "main/context.h"
30#include "main/macros.h"
31#include "main/imports.h"
32
33#include "m_matrix.h"
34#include "m_xform.h"
35
36#include "m_debug.h"
37#include "m_debug_util.h"
38
39#ifdef __UNIXOS2__
40/* The linker doesn't like empty files */
41static char dummy;
42#endif
43
44#ifdef DEBUG_MATH  /* This code only used for debugging */
45
46static clip_func *clip_tab[2] = {
47   _mesa_clip_tab,
48   _mesa_clip_np_tab
49};
50static char *cnames[2] = {
51   "_mesa_clip_tab",
52   "_mesa_clip_np_tab"
53};
54#ifdef RUN_DEBUG_BENCHMARK
55static char *cstrings[2] = {
56   "clip, perspective divide",
57   "clip, no divide"
58};
59#endif
60
61
62/* =============================================================
63 * Reference cliptests
64 */
65
66static GLvector4f *ref_cliptest_points4( GLvector4f *clip_vec,
67					 GLvector4f *proj_vec,
68					 GLubyte clipMask[],
69					 GLubyte *orMask,
70					 GLubyte *andMask,
71					 GLboolean viewport_z_clip )
72{
73   const GLuint stride = clip_vec->stride;
74   const GLuint count = clip_vec->count;
75   const GLfloat *from = (GLfloat *)clip_vec->start;
76   GLuint c = 0;
77   GLfloat (*vProj)[4] = (GLfloat (*)[4])proj_vec->start;
78   GLubyte tmpAndMask = *andMask;
79   GLubyte tmpOrMask = *orMask;
80   GLuint i;
81   for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) ) {
82      const GLfloat cx = from[0];
83      const GLfloat cy = from[1];
84      const GLfloat cz = from[2];
85      const GLfloat cw = from[3];
86      GLubyte mask = 0;
87      if ( -cx + cw < 0 ) mask |= CLIP_RIGHT_BIT;
88      if (  cx + cw < 0 ) mask |= CLIP_LEFT_BIT;
89      if ( -cy + cw < 0 ) mask |= CLIP_TOP_BIT;
90      if (  cy + cw < 0 ) mask |= CLIP_BOTTOM_BIT;
91      if (viewport_z_clip) {
92	 if ( -cz + cw < 0 ) mask |= CLIP_FAR_BIT;
93	 if (  cz + cw < 0 ) mask |= CLIP_NEAR_BIT;
94      }
95      clipMask[i] = mask;
96      if ( mask ) {
97	 c++;
98	 tmpAndMask &= mask;
99	 tmpOrMask |= mask;
100	 vProj[i][0] = 0;
101	 vProj[i][1] = 0;
102	 vProj[i][2] = 0;
103	 vProj[i][3] = 1;
104      } else {
105	 GLfloat oow = 1.0F / cw;
106	 vProj[i][0] = cx * oow;
107	 vProj[i][1] = cy * oow;
108	 vProj[i][2] = cz * oow;
109	 vProj[i][3] = oow;
110      }
111   }
112
113   *orMask = tmpOrMask;
114   *andMask = (GLubyte) (c < count ? 0 : tmpAndMask);
115
116   proj_vec->flags |= VEC_SIZE_4;
117   proj_vec->size = 4;
118   proj_vec->count = clip_vec->count;
119   return proj_vec;
120}
121
122/* Keep these here for now, even though we don't use them...
123 */
124static GLvector4f *ref_cliptest_points3( GLvector4f *clip_vec,
125					 GLvector4f *proj_vec,
126					 GLubyte clipMask[],
127					 GLubyte *orMask,
128					 GLubyte *andMask,
129                                         GLboolean viewport_z_clip )
130{
131   const GLuint stride = clip_vec->stride;
132   const GLuint count = clip_vec->count;
133   const GLfloat *from = (GLfloat *)clip_vec->start;
134
135   GLubyte tmpOrMask = *orMask;
136   GLubyte tmpAndMask = *andMask;
137   GLuint i;
138   for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) ) {
139      const GLfloat cx = from[0], cy = from[1], cz = from[2];
140      GLubyte mask = 0;
141      if ( cx >  1.0 )		mask |= CLIP_RIGHT_BIT;
142      else if ( cx < -1.0 )	mask |= CLIP_LEFT_BIT;
143      if ( cy >  1.0 )		mask |= CLIP_TOP_BIT;
144      else if ( cy < -1.0 )	mask |= CLIP_BOTTOM_BIT;
145      if (viewport_z_clip) {
146         if ( cz >  1.0 )		mask |= CLIP_FAR_BIT;
147         else if ( cz < -1.0 )	mask |= CLIP_NEAR_BIT;
148      }
149      clipMask[i] = mask;
150      tmpOrMask |= mask;
151      tmpAndMask &= mask;
152   }
153
154   *orMask = tmpOrMask;
155   *andMask = tmpAndMask;
156   return clip_vec;
157}
158
159static GLvector4f * ref_cliptest_points2( GLvector4f *clip_vec,
160					  GLvector4f *proj_vec,
161					  GLubyte clipMask[],
162					  GLubyte *orMask,
163					  GLubyte *andMask,
164                                          GLboolean viewport_z_clip )
165{
166   const GLuint stride = clip_vec->stride;
167   const GLuint count = clip_vec->count;
168   const GLfloat *from = (GLfloat *)clip_vec->start;
169
170   GLubyte tmpOrMask = *orMask;
171   GLubyte tmpAndMask = *andMask;
172   GLuint i;
173
174   (void) viewport_z_clip;
175
176   for ( i = 0 ; i < count ; i++, STRIDE_F(from, stride) ) {
177      const GLfloat cx = from[0], cy = from[1];
178      GLubyte mask = 0;
179      if ( cx >  1.0 )		mask |= CLIP_RIGHT_BIT;
180      else if ( cx < -1.0 )	mask |= CLIP_LEFT_BIT;
181      if ( cy >  1.0 )		mask |= CLIP_TOP_BIT;
182      else if ( cy < -1.0 )	mask |= CLIP_BOTTOM_BIT;
183      clipMask[i] = mask;
184      tmpOrMask |= mask;
185      tmpAndMask &= mask;
186   }
187
188   *orMask = tmpOrMask;
189   *andMask = tmpAndMask;
190   return clip_vec;
191}
192
193static clip_func ref_cliptest[5] = {
194   0,
195   0,
196   ref_cliptest_points2,
197   ref_cliptest_points3,
198   ref_cliptest_points4
199};
200
201
202/* =============================================================
203 * Cliptest tests
204 */
205
206ALIGN16(static GLfloat, s[TEST_COUNT][4]);
207ALIGN16(static GLfloat, d[TEST_COUNT][4]);
208ALIGN16(static GLfloat, r[TEST_COUNT][4]);
209
210
211/**
212 * Check if X, Y or Z component of the coordinate is close to W, in terms
213 * of the clip test.
214 */
215static GLboolean
216xyz_close_to_w(const GLfloat c[4])
217{
218   float k = 0.0001;
219   return (fabs(c[0] - c[3]) < k ||
220           fabs(c[1] - c[3]) < k ||
221           fabs(c[2] - c[3]) < k ||
222           fabs(-c[0] - c[3]) < k ||
223           fabs(-c[1] - c[3]) < k ||
224           fabs(-c[2] - c[3]) < k);
225}
226
227
228
229static int test_cliptest_function( clip_func func, int np,
230				   int psize, long *cycles )
231{
232   GLvector4f source[1], dest[1], ref[1];
233   GLubyte dm[TEST_COUNT], dco, dca;
234   GLubyte rm[TEST_COUNT], rco, rca;
235   int i, j;
236#ifdef  RUN_DEBUG_BENCHMARK
237   int cycle_i;                /* the counter for the benchmarks we run */
238#endif
239   GLboolean viewport_z_clip = GL_TRUE;
240
241   (void) cycles;
242
243   if ( psize > 4 ) {
244      _mesa_problem( NULL, "test_cliptest_function called with psize > 4\n" );
245      return 0;
246   }
247
248   for ( i = 0 ; i < TEST_COUNT ; i++) {
249      ASSIGN_4V( d[i], 0.0, 0.0, 0.0, 1.0 );
250      ASSIGN_4V( s[i], 0.0, 0.0, 0.0, 1.0 );
251      for ( j = 0 ; j < psize ; j++ )
252         s[i][j] = rnd();
253   }
254
255   source->data = (GLfloat(*)[4])s;
256   source->start = (GLfloat *)s;
257   source->count = TEST_COUNT;
258   source->stride = sizeof(s[0]);
259   source->size = 4;
260   source->flags = 0;
261
262   dest->data = (GLfloat(*)[4])d;
263   dest->start = (GLfloat *)d;
264   dest->count = TEST_COUNT;
265   dest->stride = sizeof(float[4]);
266   dest->size = 0;
267   dest->flags = 0;
268
269   ref->data = (GLfloat(*)[4])r;
270   ref->start = (GLfloat *)r;
271   ref->count = TEST_COUNT;
272   ref->stride = sizeof(float[4]);
273   ref->size = 0;
274   ref->flags = 0;
275
276   dco = rco = 0;
277   dca = rca = CLIP_FRUSTUM_BITS;
278
279   ref_cliptest[psize]( source, ref, rm, &rco, &rca, viewport_z_clip );
280
281   if ( mesa_profile ) {
282      BEGIN_RACE( *cycles );
283      func( source, dest, dm, &dco, &dca, viewport_z_clip );
284      END_RACE( *cycles );
285   }
286   else {
287      func( source, dest, dm, &dco, &dca, viewport_z_clip );
288   }
289
290   if ( dco != rco ) {
291      printf( "\n-----------------------------\n" );
292      printf( "dco = 0x%02x   rco = 0x%02x\n", dco, rco );
293      return 0;
294   }
295   if ( dca != rca ) {
296      printf( "\n-----------------------------\n" );
297      printf( "dca = 0x%02x   rca = 0x%02x\n", dca, rca );
298      return 0;
299   }
300   for ( i = 0 ; i < TEST_COUNT ; i++ ) {
301      if ( dm[i] != rm[i] ) {
302         GLfloat *c = source->start;
303         STRIDE_F(c, source->stride * i);
304         if (psize == 4 && xyz_close_to_w(c)) {
305            /* The coordinate is very close to the clip plane.  The clipmask
306             * may vary depending on code path, but that's OK.
307             */
308            continue;
309         }
310	 printf( "\n-----------------------------\n" );
311	 printf( "mask[%d] = 0x%02x   ref mask[%d] = 0x%02x\n", i, dm[i], i,rm[i] );
312         printf(" coord = %f, %f, %f, %f\n",
313                c[0], c[1], c[2], c[3]);
314	 return 0;
315      }
316   }
317
318   /* Only verify output on projected points4 case.  FIXME: Do we need
319    * to test other cases?
320    */
321   if ( np || psize < 4 )
322      return 1;
323
324   for ( i = 0 ; i < TEST_COUNT ; i++ ) {
325      for ( j = 0 ; j < 4 ; j++ ) {
326         if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
327            printf( "\n-----------------------------\n" );
328            printf( "(i = %i, j = %i)  dm = 0x%02x   rm = 0x%02x\n",
329		    i, j, dm[i], rm[i] );
330            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
331		    d[i][0], r[i][0], r[i][0]-d[i][0],
332		    MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
333            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
334		    d[i][1], r[i][1], r[i][1]-d[i][1],
335		    MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
336            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
337		    d[i][2], r[i][2], r[i][2]-d[i][2],
338		    MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
339            printf( "%f \t %f \t [diff = %e - %i bit missed]\n",
340		    d[i][3], r[i][3], r[i][3]-d[i][3],
341		    MAX_PRECISION - significand_match( d[i][3], r[i][3] ) );
342            return 0;
343         }
344      }
345   }
346
347   return 1;
348}
349
350void _math_test_all_cliptest_functions( char *description )
351{
352   int np, psize;
353   long benchmark_tab[2][4];
354   static int first_time = 1;
355
356   if ( first_time ) {
357      first_time = 0;
358      mesa_profile = _mesa_getenv( "MESA_PROFILE" );
359   }
360
361#ifdef RUN_DEBUG_BENCHMARK
362   if ( mesa_profile ) {
363      if ( !counter_overhead ) {
364	 INIT_COUNTER();
365	 printf( "counter overhead: %ld cycles\n\n", counter_overhead );
366      }
367      printf( "cliptest results after hooking in %s functions:\n", description );
368   }
369#endif
370
371#ifdef RUN_DEBUG_BENCHMARK
372   if ( mesa_profile ) {
373      printf( "\n\t" );
374      for ( psize = 2 ; psize <= 4 ; psize++ ) {
375	 printf( " p%d\t", psize );
376      }
377      printf( "\n--------------------------------------------------------\n\t" );
378   }
379#endif
380
381   for ( np = 0 ; np < 2 ; np++ ) {
382      for ( psize = 2 ; psize <= 4 ; psize++ ) {
383	 clip_func func = clip_tab[np][psize];
384	 long *cycles = &(benchmark_tab[np][psize-1]);
385
386	 if ( test_cliptest_function( func, np, psize, cycles ) == 0 ) {
387	    char buf[100];
388	    sprintf( buf, "%s[%d] failed test (%s)",
389		     cnames[np], psize, description );
390	    _mesa_problem( NULL, "%s", buf );
391	 }
392#ifdef RUN_DEBUG_BENCHMARK
393	 if ( mesa_profile )
394	    printf( " %li\t", benchmark_tab[np][psize-1] );
395#endif
396      }
397#ifdef RUN_DEBUG_BENCHMARK
398      if ( mesa_profile )
399	 printf( " | [%s]\n\t", cstrings[np] );
400#endif
401   }
402#ifdef RUN_DEBUG_BENCHMARK
403   if ( mesa_profile )
404      printf( "\n" );
405#endif
406}
407
408
409#endif /* DEBUG_MATH */
410