| /** @file mathop.h | |
| ** @brief Math operations | |
| ** @author Andrea Vedaldi | |
| **/ | |
| /* AUTORIGHTS | |
| Copyright (C) 2007-09 Andrea Vedaldi and Brian Fulkerson | |
| This file is part of VLFeat, available in the terms of the GNU | |
| General Public License version 2. | |
| */ | |
| /** @brief Logarithm of 2 (math constant)*/ | |
| /** @brief Pi (math constant) */ | |
| /** @brief IEEE single precision epsilon (math constant) | |
| ** | |
| ** <code>1.0F + VL_EPSILON_F</code> is the smallest representable | |
| ** single precision number greater than @c 1.0F. Numerically, | |
| ** ::VL_EPSILON_F is equal to @f$ 2^{-23} @f$. | |
| ** | |
| **/ | |
| /** @brief IEEE double precision epsilon (math constant) | |
| ** | |
| ** <code>1.0 + VL_EPSILON_D</code> is the smallest representable | |
| ** double precision number greater than @c 1.0. Numerically, | |
| ** ::VL_EPSILON_D is equal to @f$ 2^{-52} @f$. | |
| **/ | |
| /* | |
| For the code below: An ANSI C compiler takes the two expressions, | |
| LONG_VAR and CHAR_VAR, and implicitly casts them to the type of the | |
| first member of the union. Refer to K&R Second Edition Page 148, | |
| last paragraph. | |
| */ | |
| /** @brief IEEE single precision quiet NaN constant */ | |
| static union { vl_uint32 raw ; float value ; } | |
| const vl_nan_f = | |
| { 0x7FC00000UL } ; | |
| /** @brief IEEE single precision infinity constant */ | |
| static union { vl_uint32 raw ; float value ; } | |
| const vl_infinity_f = | |
| { 0x7F800000UL } ; | |
| /** @brief IEEE double precision quiet NaN constant */ | |
| static union { vl_uint64 raw ; double value ; } | |
| const vl_nan_d = | |
| { 0x7FF8000000000000ui64 } ; | |
| { 0x7FF8000000000000ULL } ; | |
| /** @brief IEEE double precision infinity constant */ | |
| static union { vl_uint64 raw ; double value ; } | |
| const vl_infinity_d = | |
| { 0x7FF0000000000000ui64 } ; | |
| { 0x7FF0000000000000ULL } ; | |
| /** @brief IEEE single precision NaN (not signaling) */ | |
| /** @brief IEEE single precision positive infinity (not signaling) */ | |
| /** @brief IEEE double precision NaN (not signaling) */ | |
| /** @brief IEEE double precision positive infinity (not signaling) */ | |
| /** ------------------------------------------------------------------ | |
| ** @{ | |
| ** @brief Fast <code>mod(x, 2 * VL_PI)</code> | |
| ** | |
| ** @param x input value. | |
| ** | |
| ** The function is optimized for small absolute values of @a x. | |
| ** | |
| ** The result is guaranteed to be not smaller than 0. However, due to | |
| ** finite numerical precision and rounding errors, the result can be | |
| ** equal to 2 * VL_PI (for instance, if @c x is a very small negative | |
| ** number). | |
| ** | |
| ** @return <code>mod(x, 2 * VL_PI)</code> | |
| **/ | |
| VL_INLINE | |
| float vl_mod_2pi_f (float x) | |
| { | |
| while (x > 2 * VL_PI) x -= (float) (2 * VL_PI) ; | |
| while (x < 0.0F ) x += (float) (2 * VL_PI); | |
| return x ; | |
| } | |
| VL_INLINE | |
| double vl_mod_2pi_d (double x) | |
| { | |
| while (x > 2 * VL_PI) x -= 2 * VL_PI ; | |
| while (x < 0.0 ) x += 2 * VL_PI ; | |
| return x ; | |
| } | |
| /** @} */ | |
| /** ------------------------------------------------------------------ | |
| ** @{ | |
| ** @brief Fast <code>(int) floor(x)</code> | |
| ** | |
| ** @param x argument. | |
| ** @return @c (int) floor(x) | |
| **/ | |
| VL_INLINE | |
| int | |
| vl_floor_f (float x) | |
| { | |
| int xi = (int) x ; | |
| if (x >= 0 || (float) xi == x) return xi ; | |
| else return xi - 1 ; | |
| } | |
| VL_INLINE | |
| int | |
| vl_floor_d (double x) | |
| { | |
| int xi = (int) x ; | |
| if (x >= 0 || (double) xi == x) return xi ; | |
| else return xi - 1 ; | |
| } | |
| /* @} */ | |
| /** ------------------------------------------------------------------ | |
| ** @{ | |
| ** @brief Fast @c abs(x) | |
| ** @param x argument. | |
| ** @return @c abs(x) | |
| **/ | |
| VL_INLINE float | |
| vl_abs_f (float x) | |
| { | |
| return __builtin_fabsf (x) ; | |
| return fabsf(x) ; | |
| } | |
| VL_INLINE double | |
| vl_abs_d (double x) | |
| { | |
| return __builtin_fabs (x) ; | |
| return fabs(x) ; | |
| } | |
| /** @} */ | |
| /** ------------------------------------------------------------------ | |
| ** @{ | |
| ** @brief Fast @c atan2 approximation | |
| ** @param y argument. | |
| ** @param x argument. | |
| ** | |
| ** The function computes a relatively rough but fast approximation of | |
| ** @c atan2(y,x). | |
| ** | |
| ** @par Algorithm | |
| ** | |
| ** The algorithm approximates the function @f$ f(r)=atan((1-r)/(1+r)) | |
| ** @f$, @f$ r \in [-1,1] @f$ with a third order polynomial @f$ | |
| ** f(r)=c_0 + c_1 r + c_2 r^2 + c_3 r^3 @f$. To fit the polynomial | |
| ** we impose the constraints | |
| ** | |
| ** @f{eqnarray*} | |
| ** f(+1) &=& c_0 + c_1 + c_2 + c_3 = atan(0) = 0,\\ | |
| ** f(-1) &=& c_0 - c_1 + c_2 - c_3 = atan(\infty) = \pi/2,\\ | |
| ** f(0) &=& c_0 = atan(1) = \pi/4. | |
| ** @f} | |
| ** | |
| ** The last degree of freedom is fixed by minimizing the @f$ | |
| ** l^{\infty} @f$ error, which yields | |
| ** | |
| ** @f[ | |
| ** c_0=\pi/4, \quad | |
| ** c_1=-0.9675, \quad | |
| ** c_2=0, \quad | |
| ** c_3=0.1821, | |
| ** @f] | |
| ** | |
| ** with maximum error of 0.0061 radians at 0.35 degrees. | |
| ** | |
| ** @return Approximation of @c atan2(y,x). | |
| **/ | |
| VL_INLINE | |
| float | |
| vl_fast_atan2_f (float y, float x) | |
| { | |
| float angle, r ; | |
| float const c3 = 0.1821F ; | |
| float const c1 = 0.9675F ; | |
| float abs_y = vl_abs_f (y) + VL_EPSILON_F ; | |
| if (x >= 0) { | |
| r = (x - abs_y) / (x + abs_y) ; | |
| angle = (float) (VL_PI / 4) ; | |
| } else { | |
| r = (x + abs_y) / (abs_y - x) ; | |
| angle = (float) (3 * VL_PI / 4) ; | |
| } | |
| angle += (c3*r*r - c1) * r ; | |
| return (y < 0) ? - angle : angle ; | |
| } | |
| VL_INLINE | |
| double | |
| vl_fast_atan2_d (double y, double x) | |
| { | |
| double angle, r ; | |
| double const c3 = 0.1821 ; | |
| double const c1 = 0.9675 ; | |
| double abs_y = vl_abs_d (y) + VL_EPSILON_D ; | |
| if (x >= 0) { | |
| r = (x - abs_y) / (x + abs_y) ; | |
| angle = VL_PI / 4 ; | |
| } else { | |
| r = (x + abs_y) / (abs_y - x) ; | |
| angle = 3 * VL_PI / 4 ; | |
| } | |
| angle += (c3*r*r - c1) * r ; | |
| return (y < 0) ? - angle : angle ; | |
| } | |
| /** @} */ | |
| /** ------------------------------------------------------------------ | |
| ** @{ | |
| ** @brief Fast @c resqrt approximation | |
| ** @param x argument. | |
| ** | |
| ** The function quickly computes an approximation of @f$ x^{-1/2} | |
| ** @f$. | |
| ** | |
| ** @par Algorithm | |
| ** | |
| ** The goal is to compute @f$ y = x^{-1/2} @f$, which we do by | |
| ** finding the solution of @f$ 0 = f(y) = y^{-2} - x@f$ by two Newton | |
| ** steps. Each Newton iteration is given by | |
| ** | |
| ** @f[ | |
| ** y \leftarrow | |
| ** y - \frac{f(y)}{\dot f(y)} = | |
| ** y + \frac{1}{2} (y-xy^3) = | |
| ** \frac{y}{2} \left( 3 - xy^2 \right) | |
| ** @f] | |
| ** | |
| ** which yields a simple polynomial update rule. | |
| ** | |
| ** The clever bit (attributed to either J. Carmack or G. Tarolli) is | |
| ** the way an initial guess @f$ y \approx x^{-1/2} @f$ is chosen. | |
| ** | |
| ** @see <a href="http://www.lomont.org/Math/Papers/2003/InvSqrt.pdf">Inverse Sqare Root</a>. | |
| ** | |
| ** @return Approximation of @c resqrt(x). | |
| **/ | |
| VL_INLINE | |
| float | |
| vl_fast_resqrt_f (float x) | |
| { | |
| /* 32-bit version */ | |
| union { | |
| float x ; | |
| vl_int32 i ; | |
| } u ; | |
| float xhalf = (float) 0.5 * x ; | |
| /* convert floating point value in RAW integer */ | |
| u.x = x ; | |
| /* gives initial guess y0 */ | |
| u.i = 0x5f3759df - (u.i >> 1); | |
| /*u.i = 0xdf59375f - (u.i>>1);*/ | |
| /* two Newton steps */ | |
| u.x = u.x * ( (float) 1.5 - xhalf*u.x*u.x) ; | |
| u.x = u.x * ( (float) 1.5 - xhalf*u.x*u.x) ; | |
| return u.x ; | |
| } | |
| VL_INLINE | |
| double | |
| vl_fast_resqrt_d (double x) | |
| { | |
| /* 64-bit version */ | |
| union { | |
| double x ; | |
| vl_int64 i ; | |
| } u ; | |
| double xhalf = (double) 0.5 * x ; | |
| /* convert floating point value in RAW integer */ | |
| u.x = x ; | |
| /* gives initial guess y0 */ | |
| u.i = 0x5fe6ec85e7de30dai64 - (u.i >> 1) ; | |
| u.i = 0x5fe6ec85e7de30daLL - (u.i >> 1) ; | |
| /* two Newton steps */ | |
| u.x = u.x * ( (double) 1.5 - xhalf*u.x*u.x) ; | |
| u.x = u.x * ( (double) 1.5 - xhalf*u.x*u.x) ; | |
| return u.x ; | |
| } | |
| /** @} */ | |
| /** ------------------------------------------------------------------ | |
| ** @{ | |
| ** @brief Fast @c sqrt approximation | |
| ** @param x argument. | |
| ** @return Approximation of @c sqrt(x). | |
| ** | |
| ** The function computes a fast approximation of @c sqrt(x). | |
| ** | |
| ** @par Floating-point algorithm | |
| ** | |
| ** For the floating point cases, the function uses ::vl_fast_resqrt_f | |
| ** (or ::vl_fast_resqrt_d) to compute <code>x * | |
| ** vl_fast_resqrt_f(x)</code>. | |
| ** | |
| ** @par Integer algorithm | |
| ** | |
| ** We seek for the largest integer @e y such that @f$ y^2 \leq x @f$. | |
| ** Write @f$ y = w + b_k 2^k + z @f$ where the binary expansion of the | |
| ** various variable is | |
| ** | |
| ** @f[ | |
| ** x = \sum_{i=0}^{n-1} 2^i a_i, \qquad | |
| ** w = \sum_{i=k+1}^{m-1} 2^i b_i, \qquad | |
| ** z = \sum_{i=0}^{k-1} 2^i b_i. | |
| ** @f] | |
| ** | |
| ** Assume @e w known. Expanding the square and using the fact that | |
| ** @f$ b_k^2=b_k @f$, we obtain the following constraint for @f$ b_k | |
| ** @f$ and @e z: | |
| ** | |
| ** @f[ | |
| ** x - w^2 \geq 2^k ( 2 w + 2^k ) b_k + z (z + 2wz + 2^{k+1}z b_k) | |
| ** @f] | |
| ** | |
| ** A necessary condition for @f$ b_k = 1 @f$ is that this equation is | |
| ** satisfied for @f$ z = 0 @f$ (as the second term is always | |
| ** non-negative). In fact, this condition is also sufficient, since | |
| ** we are looking for the @e largest solution @e y. | |
| ** | |
| ** This yields the following iterative algorithm. First, note that if | |
| ** @e x is stored in @e n bits, where @e n is even, then the integer | |
| ** square root does not require more than @f$ m = n / 2 @f$ bit to be | |
| ** stored. Thus initially, @f$ w = 0 @f$ and @f$ k = m - 1 = n/2 - 1 | |
| ** @f$. Then, at each iteration the equation is tested, determining | |
| ** @f$ b_{m-1}, b_{m-2}, ... @f$ in this order. | |
| **/ | |
| VL_INLINE float | |
| vl_fast_sqrt_f (float x) | |
| { | |
| return (x < 1e-8) ? 0 : x * vl_fast_resqrt_f (x) ; | |
| } | |
| VL_INLINE double | |
| vl_fast_sqrt_d (float x) | |
| { | |
| return (x < 1e-8) ? 0 : x * vl_fast_resqrt_d (x) ; | |
| } | |
| VL_INLINE vl_uint32 vl_fast_sqrt_ui32 (vl_uint32 x) ; | |
| VL_INLINE vl_uint16 vl_fast_sqrt_ui16 (vl_uint16 x) ; | |
| VL_INLINE vl_uint8 vl_fast_sqrt_ui8 (vl_uint8 x) ; | |
| /** @} */ | |
| VL_FAST_SQRT_UI(vl_uint32, ui32) | |
| VL_FAST_SQRT_UI(vl_uint16, ui16) | |
| VL_FAST_SQRT_UI(vl_uint8, ui8 ) | |
| /** @} */ | |
| /* ---------------------------------------------------------------- */ | |
| typedef float (*VlFloatVectorComparisonFunction)(vl_size dimension, float const * X, float const * Y) ; | |
| typedef double (*VlDoubleVectorComparisonFunction)(vl_size dimension, double const * X, double const * Y) ; | |
| typedef void VlFloatVectorComparisonFunction ; | |
| typedef void VlDoubleVectorComparisonFunction ; | |
| /** @brief Vector comparison types */ | |
| enum _VlVectorComparisonType { | |
| VlDistanceL1, /**< l1 distance */ | |
| VlDistanceL2, /**< l2 distance */ | |
| VlDistanceChi2, /**< Chi2 distance */ | |
| VlKernelL1, /**< l1 kernel */ | |
| VlKernelL2, /**< l2 kernel */ | |
| VlKernelChi2 /**< Chi2 kernel */ | |
| } ; | |
| /** @brief Vector comparison types */ | |
| typedef enum _VlVectorComparisonType VlVectorComparisonType ; | |
| VL_EXPORT VlFloatVectorComparisonFunction | |
| vl_get_vector_comparison_function_f (VlVectorComparisonType type) ; | |
| VL_EXPORT VlDoubleVectorComparisonFunction | |
| vl_get_vector_comparison_function_d (VlVectorComparisonType type) ; | |
| VL_EXPORT void | |
| vl_eval_vector_comparison_on_all_pairs_f (float * result, vl_size dimension, | |
| float const * X, vl_size numDataX, | |
| float const * Y, vl_size numDataY, | |
| VlFloatVectorComparisonFunction function) ; | |
| VL_EXPORT void | |
| vl_eval_vector_comparison_on_all_pairs_d (double * result, vl_size dimension, | |
| double const * X, vl_size numDataX, | |
| double const * Y, vl_size numDataY, | |
| VlDoubleVectorComparisonFunction) ; | |
| /* VL_MATHOP_H */ | |