summaryrefslogtreecommitdiff
path: root/Build/source/libs/xpdf/xpdf-src/splash/SplashMath.h
blob: 45f4e53e6bbb6c80839960ce6741279e2852161c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
//========================================================================
//
// SplashMath.h
//
// Copyright 2003-2013 Glyph & Cog, LLC
//
//========================================================================

#ifndef SPLASHMATH_H
#define SPLASHMATH_H

#include <aconf.h>

#if USE_FIXEDPONT
#  include "FixedPoint.h"
#else
#  include <math.h>
#  if (defined(__GNUC__) && defined(__SSE2__)) || \
      (defined(_WIN32) && (_M_IX86_FP == 2 || defined(_M_X64)))
#    include <emmintrin.h>
#  endif
#endif
#include "SplashTypes.h"

static inline SplashCoord splashAbs(SplashCoord x) {
#if USE_FIXEDPOINT
  return FixedPoint::abs(x);
#else
  return fabs(x);
#endif
}

// floor() and (int)() are implemented separately, which results
// in changing the FPCW multiple times - so we optimize it with
// some inline assembly or SSE intrinsics.
static inline int splashFloor(SplashCoord x) {
#if USE_FIXEDPOINT

  //--- fixed point

  return FixedPoint::floor(x);

#elif (defined(__GNUC__) && defined(__SSE2__)) || \
      (defined(_WIN32) && (_M_IX86_FP == 2 || defined(_M_X64)))

  //--- SSE2 intrinsics
  // NB: 64-bit x86 guarantees availability of SSE2.

  __m128d m1, m2;
  int i, s;

  m1 = _mm_set_sd(x);
  i = _mm_cvttsd_si32(m1);
  m2 = _mm_cvtsi32_sd(m1, i);
  s = _mm_ucomigt_sd(m2, m1);
  return i - s;

#elif defined(__GNUC__) && defined(__i386__) && !defined(__APPLE__)

  //--- x87 inline assembly (gcc/clang)
  // (this code fails on OSX for reasons I don't understand)

  Gushort oldCW, newCW, t;
  int result;

  __asm__ volatile("fnstcw %0\n"
		   "movw   %0, %3\n"
		   "andw   $0xf3ff, %3\n"
		   "orw    $0x0400, %3\n"
		   "movw   %3, %1\n"       // round down
		   "fldcw  %1\n"
		   "fistl %2\n"
		   "fldcw  %0\n"
		   : "=m" (oldCW), "=m" (newCW), "=m" (result), "=r" (t)
		   : "t" (x));
  return result;

#elif defined(_WIN32) && defined(_M_IX86)

  //--- x87 inline assembly (VC)

  Gushort oldCW, newCW;
  int result;

  __asm fld QWORD PTR x
  __asm fnstcw WORD PTR oldCW
  __asm mov ax, WORD PTR oldCW
  __asm and ax, 0xf3ff
  __asm or ax, 0x0400
  __asm mov WORD PTR newCW, ax     // round down
  __asm fldcw WORD PTR newCW
  __asm fistp DWORD PTR result
  __asm fldcw WORD PTR oldCW
  return result;

#else

  //--- all others

  return (int)floor(x);

#endif
}

// ceil() and (int)() are implemented separately, which results
// in changing the FPCW multiple times - so we optimize it with
// some inline assembly or SSE intrinsics.
static inline int splashCeil(SplashCoord x) {
#if USE_FIXEDPOINT

  //--- fixed point

  return FixedPoint::ceil(x);

#elif (defined(__GNUC__) && defined(__SSE2__)) || \
      (defined(_WIN32) && (_M_IX86_FP == 2 || defined(_M_X64)))

  //--- SSE2 intrinsics
  // NB: 64-bit x86 guarantees availability of SSE2.

  __m128d m1, m2;
  int i, s;

  m1 = _mm_set_sd(x);
  i = _mm_cvttsd_si32(m1);
  m2 = _mm_cvtsi32_sd(m1, i);
  s = _mm_ucomilt_sd(m2, m1);
  return i + s;

#elif defined(__GNUC__) && defined(__i386__) && !defined(__APPLE__)

  //--- x87 inline assembly (gcc/clang)
  // (this code fails on OSX for reasons I don't understand)

  Gushort oldCW, newCW, t;
  int result;

  __asm__ volatile("fnstcw %0\n"
		   "movw   %0, %3\n"
		   "andw   $0xf3ff, %3\n"
		   "orw    $0x0800, %3\n"
		   "movw   %3, %1\n"       // round up
		   "fldcw  %1\n"
		   "fistl %2\n"
		   "fldcw  %0\n"
		   : "=m" (oldCW), "=m" (newCW), "=m" (result), "=r" (t)
		   : "t" (x));
  return result;

#elif defined(_WIN32) && defined(_M_IX86)

  //--- x87 inline assembly (VC)

  // ceil() and (int)() are implemented separately, which results
  // in changing the FPCW multiple times - so we optimize it with
  // some inline assembly
  Gushort oldCW, newCW;
  int result;

  __asm fld QWORD PTR x
  __asm fnstcw WORD PTR oldCW
  __asm mov ax, WORD PTR oldCW
  __asm and ax, 0xf3ff
  __asm or ax, 0x0800
  __asm mov WORD PTR newCW, ax     // round up
  __asm fldcw WORD PTR newCW
  __asm fistp DWORD PTR result
  __asm fldcw WORD PTR oldCW
  return result;

#else

  //--- all others

  return (int)ceil(x);

#endif
}

static inline int splashRound(SplashCoord x) {
#if USE_FIXEDPOINT

  //--- fixed point

  return FixedPoint::round(x);

#else

  //--- all others

  return splashFloor(x + 0.5);

#endif
}

static inline SplashCoord splashAvg(SplashCoord x, SplashCoord y) {
#if USE_FIXEDPOINT
  return FixedPoint::avg(x, y);
#else
  return 0.5 * (x + y);
#endif
}

static inline SplashCoord splashSqrt(SplashCoord x) {
#if USE_FIXEDPOINT
  return FixedPoint::sqrt(x);
#else
  return sqrt(x);
#endif
}

static inline SplashCoord splashPow(SplashCoord x, SplashCoord y) {
#if USE_FIXEDPOINT
  return FixedPoint::pow(x, y);
#else
  return pow(x, y);
#endif
}

static inline SplashCoord splashDist(SplashCoord x0, SplashCoord y0,
				     SplashCoord x1, SplashCoord y1) {
  SplashCoord dx, dy;
  dx = x1 - x0;
  dy = y1 - y0;
#if USE_FIXEDPOINT
  // this handles the situation where dx*dx or dy*dy is too large to
  // fit in the 16.16 fixed point format
  SplashCoord dxa, dya, d;
  dxa = splashAbs(dx);
  dya = splashAbs(dy);
  if (dxa == 0 && dya == 0) {
    return 0;
  } else if (dxa > dya) {
    d = dya / dxa;
    return dxa * FixedPoint::sqrt(d*d + 1);
  } else {
    d = dxa / dya;
    return dya * FixedPoint::sqrt(d*d + 1);
  }
#else
  return sqrt(dx * dx + dy * dy);
#endif
}

static inline GBool splashCheckDet(SplashCoord m11, SplashCoord m12,
				   SplashCoord m21, SplashCoord m22,
				   SplashCoord epsilon) {
#if USE_FIXEDPOINT
  return FixedPoint::checkDet(m11, m12, m21, m22, epsilon);
#else
  return fabs(m11 * m22 - m12 * m21) >= epsilon;
#endif
}

// Perform stroke adjustment on a SplashCoord range [xMin, xMax),
// resulting in an int range [*xMinI, *xMaxI).
//
// There are several options:
//
// 1. Round both edge coordinates.
//    Pro: adjacent strokes/fills line up without any gaps or
//         overlaps
//    Con: lines with the same original floating point width can
//         end up with different integer widths, e.g.:
//               xMin  = 10.1   xMax  = 11.3   (width = 1.2)
//           --> xMinI = 10     xMaxI = 11     (width = 1)
//         but
//               xMin  = 10.4   xMax  = 11.6   (width = 1.2)
//           --> xMinI = 10     xMaxI = 12     (width = 2)
//
// 2. Round the min coordinate; add the ceiling of the width.
//    Pro: lines with the same original floating point width will
//         always end up with the same integer width
//    Con: adjacent strokes/fills can have overlaps (which is
//         problematic with transparency)
//    (This could use floor on the min coordinate, instead of
//    rounding, with similar results.)
//    (If the width is rounded instead of using ceiling, the results
//    Are similar, except that adjacent strokes/fills can have gaps
//    as well as overlaps.)
//
// 3. Use floor on the min coordinate and ceiling on the max
//    coordinate.
//    Pro: lines always end up at least as wide as the original
//         floating point width
//    Con: adjacent strokes/fills can have overlaps, and lines with
//         the same original floating point width can end up with
//         different integer widths; the integer width can be more
//         than one pixel wider than the original width, e.g.:
//               xMin  = 10.9   xMax  = 12.1   (width = 1.2)
//           --> xMinI = 10     xMaxI = 13     (width = 3)
//         but
//               xMin  = 10.1   xMax  = 11.3   (width = 1.2)
//           --> xMinI = 10     xMaxI = 12     (width = 2)
//
// 4. Use a hybrid approach, choosing between two of the above
//    options, based on width.  E.g., use #2 if width <= 4, and use #1
//    if width > 4.
//
// If w >= 0 and strokeAdjMode is splashStrokeAdjustCAD then a special
// mode for projecting line caps is enabled, with w being the
// transformed line width.

static inline void splashStrokeAdjust(SplashCoord xMin, SplashCoord xMax,
				      int *xMinI, int *xMaxI,
				      SplashStrokeAdjustMode strokeAdjMode,
				      SplashCoord w = -1) {
  int x0, x1;

  // make sure the coords fit in 32-bit ints
#if USE_FIXEDPOINT
  if (xMin < -32767) {
    xMin = -32767;
  } else if (xMin > 32767) {
    xMin = 32767;
  }
  if (xMax < -32767) {
    xMax = -32767;
  } else if (xMax > 32767) {
    xMax = 32767;
  }
#else
  if (xMin < -1e9) {
    xMin = -1e9;
  } else if (xMin > 1e9) {
    xMin = 1e9;
  }
  if (xMax < -1e9) {
    xMax = -1e9;
  } else if (xMax > 1e9) {
    xMax = 1e9;
  }
#endif

  // this will never be called with strokeAdjMode == splashStrokeAdjustOff
  if (strokeAdjMode == splashStrokeAdjustCAD) {
    x0 = splashRound(xMin);
    if (w >= 0) {
      x1 = splashRound(xMax - w) + splashRound(w);
    } else {
      x1 = x0 + splashRound(xMax - xMin);
    }
  } else {
    // NB: enable exactly one of these.
#if 1 // 1. Round both edge coordinates.
    x0 = splashRound(xMin);
    x1 = splashRound(xMax);
#endif
#if 0 // 2. Round the min coordinate; add the ceiling of the width.
    x0 = splashRound(xMin);
    x1 = x0 + splashCeil(xMax - xMin);
#endif
#if 0 // 3. Use floor on the min coord and ceiling on the max coord.
    x0 = splashFloor(xMin);
    x1 = splashCeil(xMax);
#endif
#if 0 // 4. Hybrid.
    SplashCoord w = xMax - xMin;
    x0 = splashRound(xMin);
    if (w > 4) {
      x1 = splashRound(xMax);
    } else {
      x1 = x0 + splashRound(w);
    }
#endif
  }
  if (x0 == x1) {
    if (xMin + xMax < 2 * x0) {
      --x0;
    } else {
      ++x1;
    }
  }
  *xMinI = x0;
  *xMaxI = x1;
}

#endif