| 1 | /**************************************************************************** |
| 2 | * |
| 3 | * fttrigon.c |
| 4 | * |
| 5 | * FreeType trigonometric functions (body). |
| 6 | * |
| 7 | * Copyright (C) 2001-2019 by |
| 8 | * David Turner, Robert Wilhelm, and Werner Lemberg. |
| 9 | * |
| 10 | * This file is part of the FreeType project, and may only be used, |
| 11 | * modified, and distributed under the terms of the FreeType project |
| 12 | * license, LICENSE.TXT. By continuing to use, modify, or distribute |
| 13 | * this file you indicate that you have read the license and |
| 14 | * understand and accept it fully. |
| 15 | * |
| 16 | */ |
| 17 | |
| 18 | /************************************************************************** |
| 19 | * |
| 20 | * This is a fixed-point CORDIC implementation of trigonometric |
| 21 | * functions as well as transformations between Cartesian and polar |
| 22 | * coordinates. The angles are represented as 16.16 fixed-point values |
| 23 | * in degrees, i.e., the angular resolution is 2^-16 degrees. Note that |
| 24 | * only vectors longer than 2^16*180/pi (or at least 22 bits) on a |
| 25 | * discrete Cartesian grid can have the same or better angular |
| 26 | * resolution. Therefore, to maintain this precision, some functions |
| 27 | * require an interim upscaling of the vectors, whereas others operate |
| 28 | * with 24-bit long vectors directly. |
| 29 | * |
| 30 | */ |
| 31 | |
| 32 | #include <ft2build.h> |
| 33 | #include FT_INTERNAL_OBJECTS_H |
| 34 | #include FT_INTERNAL_CALC_H |
| 35 | #include FT_TRIGONOMETRY_H |
| 36 | |
| 37 | |
| 38 | /* the Cordic shrink factor 0.858785336480436 * 2^32 */ |
| 39 | #define FT_TRIG_SCALE 0xDBD95B16UL |
| 40 | |
| 41 | /* the highest bit in overflow-safe vector components, */ |
| 42 | /* MSB of 0.858785336480436 * sqrt(0.5) * 2^30 */ |
| 43 | #define FT_TRIG_SAFE_MSB 29 |
| 44 | |
| 45 | /* this table was generated for FT_PI = 180L << 16, i.e. degrees */ |
| 46 | #define FT_TRIG_MAX_ITERS 23 |
| 47 | |
| 48 | static const FT_Angle |
| 49 | ft_trig_arctan_table[] = |
| 50 | { |
| 51 | 1740967L, 919879L, 466945L, 234379L, 117304L, 58666L, 29335L, |
| 52 | 14668L, 7334L, 3667L, 1833L, 917L, 458L, 229L, 115L, |
| 53 | 57L, 29L, 14L, 7L, 4L, 2L, 1L |
| 54 | }; |
| 55 | |
| 56 | |
| 57 | #ifdef FT_LONG64 |
| 58 | |
| 59 | /* multiply a given value by the CORDIC shrink factor */ |
| 60 | static FT_Fixed |
| 61 | ft_trig_downscale( FT_Fixed val ) |
| 62 | { |
| 63 | FT_Int s = 1; |
| 64 | |
| 65 | |
| 66 | if ( val < 0 ) |
| 67 | { |
| 68 | val = -val; |
| 69 | s = -1; |
| 70 | } |
| 71 | |
| 72 | /* 0x40000000 comes from regression analysis between true */ |
| 73 | /* and CORDIC hypotenuse, so it minimizes the error */ |
| 74 | val = (FT_Fixed)( |
| 75 | ( (FT_UInt64)val * FT_TRIG_SCALE + 0x40000000UL ) >> 32 ); |
| 76 | |
| 77 | return s < 0 ? -val : val; |
| 78 | } |
| 79 | |
| 80 | #else /* !FT_LONG64 */ |
| 81 | |
| 82 | /* multiply a given value by the CORDIC shrink factor */ |
| 83 | static FT_Fixed |
| 84 | ft_trig_downscale( FT_Fixed val ) |
| 85 | { |
| 86 | FT_Int s = 1; |
| 87 | FT_UInt32 lo1, hi1, lo2, hi2, lo, hi, i1, i2; |
| 88 | |
| 89 | |
| 90 | if ( val < 0 ) |
| 91 | { |
| 92 | val = -val; |
| 93 | s = -1; |
| 94 | } |
| 95 | |
| 96 | lo1 = (FT_UInt32)val & 0x0000FFFFU; |
| 97 | hi1 = (FT_UInt32)val >> 16; |
| 98 | lo2 = FT_TRIG_SCALE & 0x0000FFFFU; |
| 99 | hi2 = FT_TRIG_SCALE >> 16; |
| 100 | |
| 101 | lo = lo1 * lo2; |
| 102 | i1 = lo1 * hi2; |
| 103 | i2 = lo2 * hi1; |
| 104 | hi = hi1 * hi2; |
| 105 | |
| 106 | /* Check carry overflow of i1 + i2 */ |
| 107 | i1 += i2; |
| 108 | hi += (FT_UInt32)( i1 < i2 ) << 16; |
| 109 | |
| 110 | hi += i1 >> 16; |
| 111 | i1 = i1 << 16; |
| 112 | |
| 113 | /* Check carry overflow of i1 + lo */ |
| 114 | lo += i1; |
| 115 | hi += ( lo < i1 ); |
| 116 | |
| 117 | /* 0x40000000 comes from regression analysis between true */ |
| 118 | /* and CORDIC hypotenuse, so it minimizes the error */ |
| 119 | |
| 120 | /* Check carry overflow of lo + 0x40000000 */ |
| 121 | lo += 0x40000000UL; |
| 122 | hi += ( lo < 0x40000000UL ); |
| 123 | |
| 124 | val = (FT_Fixed)hi; |
| 125 | |
| 126 | return s < 0 ? -val : val; |
| 127 | } |
| 128 | |
| 129 | #endif /* !FT_LONG64 */ |
| 130 | |
| 131 | |
| 132 | /* undefined and never called for zero vector */ |
| 133 | static FT_Int |
| 134 | ft_trig_prenorm( FT_Vector* vec ) |
| 135 | { |
| 136 | FT_Pos x, y; |
| 137 | FT_Int shift; |
| 138 | |
| 139 | |
| 140 | x = vec->x; |
| 141 | y = vec->y; |
| 142 | |
| 143 | shift = FT_MSB( (FT_UInt32)( FT_ABS( x ) | FT_ABS( y ) ) ); |
| 144 | |
| 145 | if ( shift <= FT_TRIG_SAFE_MSB ) |
| 146 | { |
| 147 | shift = FT_TRIG_SAFE_MSB - shift; |
| 148 | vec->x = (FT_Pos)( (FT_ULong)x << shift ); |
| 149 | vec->y = (FT_Pos)( (FT_ULong)y << shift ); |
| 150 | } |
| 151 | else |
| 152 | { |
| 153 | shift -= FT_TRIG_SAFE_MSB; |
| 154 | vec->x = x >> shift; |
| 155 | vec->y = y >> shift; |
| 156 | shift = -shift; |
| 157 | } |
| 158 | |
| 159 | return shift; |
| 160 | } |
| 161 | |
| 162 | |
| 163 | static void |
| 164 | ft_trig_pseudo_rotate( FT_Vector* vec, |
| 165 | FT_Angle theta ) |
| 166 | { |
| 167 | FT_Int i; |
| 168 | FT_Fixed x, y, xtemp, b; |
| 169 | const FT_Angle *arctanptr; |
| 170 | |
| 171 | |
| 172 | x = vec->x; |
| 173 | y = vec->y; |
| 174 | |
| 175 | /* Rotate inside [-PI/4,PI/4] sector */ |
| 176 | while ( theta < -FT_ANGLE_PI4 ) |
| 177 | { |
| 178 | xtemp = y; |
| 179 | y = -x; |
| 180 | x = xtemp; |
| 181 | theta += FT_ANGLE_PI2; |
| 182 | } |
| 183 | |
| 184 | while ( theta > FT_ANGLE_PI4 ) |
| 185 | { |
| 186 | xtemp = -y; |
| 187 | y = x; |
| 188 | x = xtemp; |
| 189 | theta -= FT_ANGLE_PI2; |
| 190 | } |
| 191 | |
| 192 | arctanptr = ft_trig_arctan_table; |
| 193 | |
| 194 | /* Pseudorotations, with right shifts */ |
| 195 | for ( i = 1, b = 1; i < FT_TRIG_MAX_ITERS; b <<= 1, i++ ) |
| 196 | { |
| 197 | if ( theta < 0 ) |
| 198 | { |
| 199 | xtemp = x + ( ( y + b ) >> i ); |
| 200 | y = y - ( ( x + b ) >> i ); |
| 201 | x = xtemp; |
| 202 | theta += *arctanptr++; |
| 203 | } |
| 204 | else |
| 205 | { |
| 206 | xtemp = x - ( ( y + b ) >> i ); |
| 207 | y = y + ( ( x + b ) >> i ); |
| 208 | x = xtemp; |
| 209 | theta -= *arctanptr++; |
| 210 | } |
| 211 | } |
| 212 | |
| 213 | vec->x = x; |
| 214 | vec->y = y; |
| 215 | } |
| 216 | |
| 217 | |
| 218 | static void |
| 219 | ft_trig_pseudo_polarize( FT_Vector* vec ) |
| 220 | { |
| 221 | FT_Angle theta; |
| 222 | FT_Int i; |
| 223 | FT_Fixed x, y, xtemp, b; |
| 224 | const FT_Angle *arctanptr; |
| 225 | |
| 226 | |
| 227 | x = vec->x; |
| 228 | y = vec->y; |
| 229 | |
| 230 | /* Get the vector into [-PI/4,PI/4] sector */ |
| 231 | if ( y > x ) |
| 232 | { |
| 233 | if ( y > -x ) |
| 234 | { |
| 235 | theta = FT_ANGLE_PI2; |
| 236 | xtemp = y; |
| 237 | y = -x; |
| 238 | x = xtemp; |
| 239 | } |
| 240 | else |
| 241 | { |
| 242 | theta = y > 0 ? FT_ANGLE_PI : -FT_ANGLE_PI; |
| 243 | x = -x; |
| 244 | y = -y; |
| 245 | } |
| 246 | } |
| 247 | else |
| 248 | { |
| 249 | if ( y < -x ) |
| 250 | { |
| 251 | theta = -FT_ANGLE_PI2; |
| 252 | xtemp = -y; |
| 253 | y = x; |
| 254 | x = xtemp; |
| 255 | } |
| 256 | else |
| 257 | { |
| 258 | theta = 0; |
| 259 | } |
| 260 | } |
| 261 | |
| 262 | arctanptr = ft_trig_arctan_table; |
| 263 | |
| 264 | /* Pseudorotations, with right shifts */ |
| 265 | for ( i = 1, b = 1; i < FT_TRIG_MAX_ITERS; b <<= 1, i++ ) |
| 266 | { |
| 267 | if ( y > 0 ) |
| 268 | { |
| 269 | xtemp = x + ( ( y + b ) >> i ); |
| 270 | y = y - ( ( x + b ) >> i ); |
| 271 | x = xtemp; |
| 272 | theta += *arctanptr++; |
| 273 | } |
| 274 | else |
| 275 | { |
| 276 | xtemp = x - ( ( y + b ) >> i ); |
| 277 | y = y + ( ( x + b ) >> i ); |
| 278 | x = xtemp; |
| 279 | theta -= *arctanptr++; |
| 280 | } |
| 281 | } |
| 282 | |
| 283 | /* round theta to acknowledge its error that mostly comes */ |
| 284 | /* from accumulated rounding errors in the arctan table */ |
| 285 | if ( theta >= 0 ) |
| 286 | theta = FT_PAD_ROUND( theta, 16 ); |
| 287 | else |
| 288 | theta = -FT_PAD_ROUND( -theta, 16 ); |
| 289 | |
| 290 | vec->x = x; |
| 291 | vec->y = theta; |
| 292 | } |
| 293 | |
| 294 | |
| 295 | /* documentation is in fttrigon.h */ |
| 296 | |
| 297 | FT_EXPORT_DEF( FT_Fixed ) |
| 298 | FT_Cos( FT_Angle angle ) |
| 299 | { |
| 300 | FT_Vector v; |
| 301 | |
| 302 | |
| 303 | FT_Vector_Unit( &v, angle ); |
| 304 | |
| 305 | return v.x; |
| 306 | } |
| 307 | |
| 308 | |
| 309 | /* documentation is in fttrigon.h */ |
| 310 | |
| 311 | FT_EXPORT_DEF( FT_Fixed ) |
| 312 | FT_Sin( FT_Angle angle ) |
| 313 | { |
| 314 | FT_Vector v; |
| 315 | |
| 316 | |
| 317 | FT_Vector_Unit( &v, angle ); |
| 318 | |
| 319 | return v.y; |
| 320 | } |
| 321 | |
| 322 | |
| 323 | /* documentation is in fttrigon.h */ |
| 324 | |
| 325 | FT_EXPORT_DEF( FT_Fixed ) |
| 326 | FT_Tan( FT_Angle angle ) |
| 327 | { |
| 328 | FT_Vector v = { 1 << 24, 0 }; |
| 329 | |
| 330 | |
| 331 | ft_trig_pseudo_rotate( &v, angle ); |
| 332 | |
| 333 | return FT_DivFix( v.y, v.x ); |
| 334 | } |
| 335 | |
| 336 | |
| 337 | /* documentation is in fttrigon.h */ |
| 338 | |
| 339 | FT_EXPORT_DEF( FT_Angle ) |
| 340 | FT_Atan2( FT_Fixed dx, |
| 341 | FT_Fixed dy ) |
| 342 | { |
| 343 | FT_Vector v; |
| 344 | |
| 345 | |
| 346 | if ( dx == 0 && dy == 0 ) |
| 347 | return 0; |
| 348 | |
| 349 | v.x = dx; |
| 350 | v.y = dy; |
| 351 | ft_trig_prenorm( &v ); |
| 352 | ft_trig_pseudo_polarize( &v ); |
| 353 | |
| 354 | return v.y; |
| 355 | } |
| 356 | |
| 357 | |
| 358 | /* documentation is in fttrigon.h */ |
| 359 | |
| 360 | FT_EXPORT_DEF( void ) |
| 361 | FT_Vector_Unit( FT_Vector* vec, |
| 362 | FT_Angle angle ) |
| 363 | { |
| 364 | if ( !vec ) |
| 365 | return; |
| 366 | |
| 367 | vec->x = FT_TRIG_SCALE >> 8; |
| 368 | vec->y = 0; |
| 369 | ft_trig_pseudo_rotate( vec, angle ); |
| 370 | vec->x = ( vec->x + 0x80L ) >> 8; |
| 371 | vec->y = ( vec->y + 0x80L ) >> 8; |
| 372 | } |
| 373 | |
| 374 | |
| 375 | /* documentation is in fttrigon.h */ |
| 376 | |
| 377 | FT_EXPORT_DEF( void ) |
| 378 | FT_Vector_Rotate( FT_Vector* vec, |
| 379 | FT_Angle angle ) |
| 380 | { |
| 381 | FT_Int shift; |
| 382 | FT_Vector v; |
| 383 | |
| 384 | |
| 385 | if ( !vec || !angle ) |
| 386 | return; |
| 387 | |
| 388 | v = *vec; |
| 389 | |
| 390 | if ( v.x == 0 && v.y == 0 ) |
| 391 | return; |
| 392 | |
| 393 | shift = ft_trig_prenorm( &v ); |
| 394 | ft_trig_pseudo_rotate( &v, angle ); |
| 395 | v.x = ft_trig_downscale( v.x ); |
| 396 | v.y = ft_trig_downscale( v.y ); |
| 397 | |
| 398 | if ( shift > 0 ) |
| 399 | { |
| 400 | FT_Int32 half = (FT_Int32)1L << ( shift - 1 ); |
| 401 | |
| 402 | |
| 403 | vec->x = ( v.x + half - ( v.x < 0 ) ) >> shift; |
| 404 | vec->y = ( v.y + half - ( v.y < 0 ) ) >> shift; |
| 405 | } |
| 406 | else |
| 407 | { |
| 408 | shift = -shift; |
| 409 | vec->x = (FT_Pos)( (FT_ULong)v.x << shift ); |
| 410 | vec->y = (FT_Pos)( (FT_ULong)v.y << shift ); |
| 411 | } |
| 412 | } |
| 413 | |
| 414 | |
| 415 | /* documentation is in fttrigon.h */ |
| 416 | |
| 417 | FT_EXPORT_DEF( FT_Fixed ) |
| 418 | FT_Vector_Length( FT_Vector* vec ) |
| 419 | { |
| 420 | FT_Int shift; |
| 421 | FT_Vector v; |
| 422 | |
| 423 | |
| 424 | if ( !vec ) |
| 425 | return 0; |
| 426 | |
| 427 | v = *vec; |
| 428 | |
| 429 | /* handle trivial cases */ |
| 430 | if ( v.x == 0 ) |
| 431 | { |
| 432 | return FT_ABS( v.y ); |
| 433 | } |
| 434 | else if ( v.y == 0 ) |
| 435 | { |
| 436 | return FT_ABS( v.x ); |
| 437 | } |
| 438 | |
| 439 | /* general case */ |
| 440 | shift = ft_trig_prenorm( &v ); |
| 441 | ft_trig_pseudo_polarize( &v ); |
| 442 | |
| 443 | v.x = ft_trig_downscale( v.x ); |
| 444 | |
| 445 | if ( shift > 0 ) |
| 446 | return ( v.x + ( 1L << ( shift - 1 ) ) ) >> shift; |
| 447 | |
| 448 | return (FT_Fixed)( (FT_UInt32)v.x << -shift ); |
| 449 | } |
| 450 | |
| 451 | |
| 452 | /* documentation is in fttrigon.h */ |
| 453 | |
| 454 | FT_EXPORT_DEF( void ) |
| 455 | FT_Vector_Polarize( FT_Vector* vec, |
| 456 | FT_Fixed *length, |
| 457 | FT_Angle *angle ) |
| 458 | { |
| 459 | FT_Int shift; |
| 460 | FT_Vector v; |
| 461 | |
| 462 | |
| 463 | if ( !vec || !length || !angle ) |
| 464 | return; |
| 465 | |
| 466 | v = *vec; |
| 467 | |
| 468 | if ( v.x == 0 && v.y == 0 ) |
| 469 | return; |
| 470 | |
| 471 | shift = ft_trig_prenorm( &v ); |
| 472 | ft_trig_pseudo_polarize( &v ); |
| 473 | |
| 474 | v.x = ft_trig_downscale( v.x ); |
| 475 | |
| 476 | *length = shift >= 0 ? ( v.x >> shift ) |
| 477 | : (FT_Fixed)( (FT_UInt32)v.x << -shift ); |
| 478 | *angle = v.y; |
| 479 | } |
| 480 | |
| 481 | |
| 482 | /* documentation is in fttrigon.h */ |
| 483 | |
| 484 | FT_EXPORT_DEF( void ) |
| 485 | FT_Vector_From_Polar( FT_Vector* vec, |
| 486 | FT_Fixed length, |
| 487 | FT_Angle angle ) |
| 488 | { |
| 489 | if ( !vec ) |
| 490 | return; |
| 491 | |
| 492 | vec->x = length; |
| 493 | vec->y = 0; |
| 494 | |
| 495 | FT_Vector_Rotate( vec, angle ); |
| 496 | } |
| 497 | |
| 498 | |
| 499 | /* documentation is in fttrigon.h */ |
| 500 | |
| 501 | FT_EXPORT_DEF( FT_Angle ) |
| 502 | FT_Angle_Diff( FT_Angle angle1, |
| 503 | FT_Angle angle2 ) |
| 504 | { |
| 505 | FT_Angle delta = angle2 - angle1; |
| 506 | |
| 507 | |
| 508 | while ( delta <= -FT_ANGLE_PI ) |
| 509 | delta += FT_ANGLE_2PI; |
| 510 | |
| 511 | while ( delta > FT_ANGLE_PI ) |
| 512 | delta -= FT_ANGLE_2PI; |
| 513 | |
| 514 | return delta; |
| 515 | } |
| 516 | |
| 517 | |
| 518 | /* END */ |
| 519 | |