|
- #pragma once
- #include <simd/vector.hpp>
- #include <simd/sse_mathfun_extension.h>
- #include <common.hpp>
- #include <math.hpp>
-
-
- namespace rack {
- namespace simd {
-
-
- // Functions based on instructions
-
- /** `~a & b` */
- inline float_4 andnot(float_4 a, float_4 b) {
- return float_4(_mm_andnot_ps(a.v, b.v));
- }
-
- /** Returns an integer with each bit corresponding to the most significant bit of each element.
- For example, `movemask(float_4::mask())` returns 0xf.
- */
- inline int movemask(float_4 a) {
- return _mm_movemask_ps(a.v);
- }
-
- /** Returns an integer with each bit corresponding to the most significant bit of each element.
- For example, `movemask(int32_4::mask())` returns 0xf.
- */
- inline int movemask(int32_4 a) {
- return _mm_movemask_ps(_mm_castsi128_ps(a.v));
- }
-
- /** Returns the approximate reciprocal square root.
- Much faster than `1/sqrt(x)`.
- */
- inline float_4 rsqrt(float_4 x) {
- return float_4(_mm_rsqrt_ps(x.v));
- }
-
- /** Returns the approximate reciprocal.
- Much faster than `1/x`.
- */
- inline float_4 rcp(float_4 x) {
- return float_4(_mm_rcp_ps(x.v));
- }
-
-
- // Nonstandard convenience functions
-
- inline float ifelse(bool cond, float a, float b) {
- return cond ? a : b;
- }
-
- /** Given a mask, returns a if mask is 0xffffffff per element, b if mask is 0x00000000 */
- inline float_4 ifelse(float_4 mask, float_4 a, float_4 b) {
- return (a & mask) | andnot(mask, b);
- }
-
- /** Returns a vector where element N is all 1's if the N'th bit of `a` is 1, or all 0's if the N'th bit of `a` is 0.
- */
- template <typename T>
- T movemaskInverse(int a);
-
- template <>
- inline int32_4 movemaskInverse<int32_4>(int a) {
- // Pick out N'th bit of `a` and check if it's 1.
- int32_4 mask1234 = int32_4(1, 2, 4, 8);
- return (mask1234 & int32_4(a)) == mask1234;
- }
-
- template <>
- inline float_4 movemaskInverse<float_4>(int a) {
- return float_4::cast(movemaskInverse<int32_4>(a));
- }
-
-
- // Standard math functions from std::
-
- /* Import std:: math functions into the simd namespace so you can use `sin(T)` etc in templated functions and get both the scalar and vector versions.
-
- Example:
-
- template <typename T>
- T sin_plus_cos(T x) {
- return simd::sin(x) + simd::cos(x);
- }
- */
-
- using std::fmax;
-
- inline float_4 fmax(float_4 x, float_4 b) {
- return float_4(_mm_max_ps(x.v, b.v));
- }
-
- using std::fmin;
-
- inline float_4 fmin(float_4 x, float_4 b) {
- return float_4(_mm_min_ps(x.v, b.v));
- }
-
- using std::sqrt;
-
- inline float_4 sqrt(float_4 x) {
- return float_4(_mm_sqrt_ps(x.v));
- }
-
- using std::log;
-
- inline float_4 log(float_4 x) {
- return float_4(sse_mathfun_log_ps(x.v));
- }
-
- using std::log10;
-
- inline float_4 log10(float_4 x) {
- return float_4(sse_mathfun_log_ps(x.v)) / std::log(10.f);
- }
-
- using std::log2;
-
- inline float_4 log2(float_4 x) {
- return float_4(sse_mathfun_log_ps(x.v)) / std::log(2.f);
- }
-
- using std::exp;
-
- inline float_4 exp(float_4 x) {
- return float_4(sse_mathfun_exp_ps(x.v));
- }
-
- using std::sin;
-
- inline float_4 sin(float_4 x) {
- return float_4(sse_mathfun_sin_ps(x.v));
- }
-
- using std::cos;
-
- inline float_4 cos(float_4 x) {
- return float_4(sse_mathfun_cos_ps(x.v));
- }
-
- using std::tan;
-
- inline float_4 tan(float_4 x) {
- return float_4(sse_mathfun_tan_ps(x.v));
- }
-
- using std::atan;
-
- inline float_4 atan(float_4 x) {
- return float_4(sse_mathfun_atan_ps(x.v));
- }
-
- using std::atan2;
-
- inline float_4 atan2(float_4 x, float_4 y) {
- return float_4(sse_mathfun_atan2_ps(x.v, y.v));
- }
-
- using std::trunc;
-
- inline float_4 trunc(float_4 a) {
- return float_4(_mm_cvtepi32_ps(_mm_cvttps_epi32(a.v)));
- }
-
- using std::floor;
-
- inline float_4 floor(float_4 a) {
- float_4 b = trunc(a);
- b -= (b > a) & 1.f;
- return b;
- }
-
- using std::ceil;
-
- inline float_4 ceil(float_4 a) {
- float_4 b = trunc(a);
- b += (b < a) & 1.f;
- return b;
- }
-
- using std::round;
-
- inline float_4 round(float_4 a) {
- a += ifelse(a < 0, -0.5f, 0.5f);
- float_4 b = trunc(a);
- return b;
- }
-
- using std::fmod;
-
- inline float_4 fmod(float_4 a, float_4 b) {
- return a - trunc(a / b) * b;
- }
-
- using std::hypot;
-
- inline float_4 hypot(float_4 a, float_4 b) {
- return sqrt(a * a + b * b);
- }
-
- using std::fabs;
-
- inline float_4 fabs(float_4 a) {
- // Sign bit
- int32_4 mask = ~0x80000000;
- return a & float_4::cast(mask);
- }
-
- using std::abs;
-
- inline float_4 abs(float_4 a) {
- return fabs(a);
- }
-
- inline float_4 abs(std::complex<float_4> a) {
- return hypot(a.real(), a.imag());
- }
-
- using std::arg;
-
- inline float_4 arg(std::complex<float_4> a) {
- return atan2(a.imag(), a.real());
- }
-
- using std::pow;
-
- inline float_4 pow(float_4 a, float_4 b) {
- return exp(b * log(a));
- }
-
- inline float_4 pow(float a, float_4 b) {
- return exp(b * std::log(a));
- }
-
- template <typename T>
- T pow(T a, int b) {
- // Optimal with `-O3 -funsafe-math-optimizations` when b is known at compile-time
- T p = 1;
- for (int i = 1; i <= b; i *= 2) {
- if (i & b)
- p *= a;
- a *= a;
- }
- return p;
- }
-
- // From math.hpp
-
- using math::clamp;
-
- inline float_4 clamp(float_4 x, float_4 a, float_4 b) {
- return fmin(fmax(x, a), b);
- }
-
- using math::rescale;
-
- inline float_4 rescale(float_4 x, float_4 xMin, float_4 xMax, float_4 yMin, float_4 yMax) {
- return yMin + (x - xMin) / (xMax - xMin) * (yMax - yMin);
- }
-
- using math::crossfade;
-
- inline float_4 crossfade(float_4 a, float_4 b, float_4 p) {
- return a + (b - a) * p;
- }
-
- using math::sgn;
-
- inline float_4 sgn(float_4 x) {
- float_4 signbit = x & -0.f;
- float_4 nonzero = (x != 0.f);
- return signbit | (nonzero & 1.f);
- }
-
-
- } // namespace simd
- } // namespace rack
|