summaryrefslogtreecommitdiffstats
path: root/depedencies/include/glm/gtc/random.inl
diff options
context:
space:
mode:
Diffstat (limited to 'depedencies/include/glm/gtc/random.inl')
-rw-r--r--depedencies/include/glm/gtc/random.inl350
1 files changed, 350 insertions, 0 deletions
diff --git a/depedencies/include/glm/gtc/random.inl b/depedencies/include/glm/gtc/random.inl
new file mode 100644
index 0000000..ad5926e
--- /dev/null
+++ b/depedencies/include/glm/gtc/random.inl
@@ -0,0 +1,350 @@
+/// @ref gtc_random
+/// @file glm/gtc/random.inl
+
+#include "../geometric.hpp"
+#include "../exponential.hpp"
+#include <cstdlib>
+#include <ctime>
+#include <cassert>
+
+namespace glm{
+namespace detail
+{
+ template <typename T, precision P, template <class, precision> class vecType>
+ struct compute_rand
+ {
+ GLM_FUNC_QUALIFIER static vecType<T, P> call();
+ };
+
+ template <precision P>
+ struct compute_rand<uint8, P, tvec1>
+ {
+ GLM_FUNC_QUALIFIER static tvec1<uint8, P> call()
+ {
+ return tvec1<uint8, P>(
+ std::rand() % std::numeric_limits<uint8>::max());
+ }
+ };
+
+ template <precision P>
+ struct compute_rand<uint8, P, tvec2>
+ {
+ GLM_FUNC_QUALIFIER static tvec2<uint8, P> call()
+ {
+ return tvec2<uint8, P>(
+ std::rand() % std::numeric_limits<uint8>::max(),
+ std::rand() % std::numeric_limits<uint8>::max());
+ }
+ };
+
+ template <precision P>
+ struct compute_rand<uint8, P, tvec3>
+ {
+ GLM_FUNC_QUALIFIER static tvec3<uint8, P> call()
+ {
+ return tvec3<uint8, P>(
+ std::rand() % std::numeric_limits<uint8>::max(),
+ std::rand() % std::numeric_limits<uint8>::max(),
+ std::rand() % std::numeric_limits<uint8>::max());
+ }
+ };
+
+ template <precision P>
+ struct compute_rand<uint8, P, tvec4>
+ {
+ GLM_FUNC_QUALIFIER static tvec4<uint8, P> call()
+ {
+ return tvec4<uint8, P>(
+ std::rand() % std::numeric_limits<uint8>::max(),
+ std::rand() % std::numeric_limits<uint8>::max(),
+ std::rand() % std::numeric_limits<uint8>::max(),
+ std::rand() % std::numeric_limits<uint8>::max());
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_rand<uint16, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint16, P> call()
+ {
+ return
+ (vecType<uint16, P>(compute_rand<uint8, P, vecType>::call()) << static_cast<uint16>(8)) |
+ (vecType<uint16, P>(compute_rand<uint8, P, vecType>::call()) << static_cast<uint16>(0));
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_rand<uint32, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint32, P> call()
+ {
+ return
+ (vecType<uint32, P>(compute_rand<uint16, P, vecType>::call()) << static_cast<uint32>(16)) |
+ (vecType<uint32, P>(compute_rand<uint16, P, vecType>::call()) << static_cast<uint32>(0));
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_rand<uint64, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint64, P> call()
+ {
+ return
+ (vecType<uint64, P>(compute_rand<uint32, P, vecType>::call()) << static_cast<uint64>(32)) |
+ (vecType<uint64, P>(compute_rand<uint32, P, vecType>::call()) << static_cast<uint64>(0));
+ }
+ };
+
+ template <typename T, precision P, template <class, precision> class vecType>
+ struct compute_linearRand
+ {
+ GLM_FUNC_QUALIFIER static vecType<T, P> call(vecType<T, P> const & Min, vecType<T, P> const & Max);
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<int8, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<int8, P> call(vecType<int8, P> const & Min, vecType<int8, P> const & Max)
+ {
+ return (vecType<int8, P>(compute_rand<uint8, P, vecType>::call() % vecType<uint8, P>(Max + static_cast<int8>(1) - Min))) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<uint8, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint8, P> call(vecType<uint8, P> const & Min, vecType<uint8, P> const & Max)
+ {
+ return (compute_rand<uint8, P, vecType>::call() % (Max + static_cast<uint8>(1) - Min)) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<int16, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<int16, P> call(vecType<int16, P> const & Min, vecType<int16, P> const & Max)
+ {
+ return (vecType<int16, P>(compute_rand<uint16, P, vecType>::call() % vecType<uint16, P>(Max + static_cast<int16>(1) - Min))) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<uint16, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint16, P> call(vecType<uint16, P> const & Min, vecType<uint16, P> const & Max)
+ {
+ return (compute_rand<uint16, P, vecType>::call() % (Max + static_cast<uint16>(1) - Min)) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<int32, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<int32, P> call(vecType<int32, P> const & Min, vecType<int32, P> const & Max)
+ {
+ return (vecType<int32, P>(compute_rand<uint32, P, vecType>::call() % vecType<uint32, P>(Max + static_cast<int32>(1) - Min))) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<uint32, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint32, P> call(vecType<uint32, P> const & Min, vecType<uint32, P> const & Max)
+ {
+ return (compute_rand<uint32, P, vecType>::call() % (Max + static_cast<uint32>(1) - Min)) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<int64, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<int64, P> call(vecType<int64, P> const & Min, vecType<int64, P> const & Max)
+ {
+ return (vecType<int64, P>(compute_rand<uint64, P, vecType>::call() % vecType<uint64, P>(Max + static_cast<int64>(1) - Min))) + Min;
+ }
+ };
+
+ template <precision P, template <class, precision> class vecType>
+ struct compute_linearRand<uint64, P, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<uint64, P> call(vecType<uint64, P> const & Min, vecType<uint64, P> const & Max)
+ {
+ return (compute_rand<uint64, P, vecType>::call() % (Max + static_cast<uint64>(1) - Min)) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<float, lowp, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<float, lowp> call(vecType<float, lowp> const & Min, vecType<float, lowp> const & Max)
+ {
+ return vecType<float, lowp>(compute_rand<uint8, lowp, vecType>::call()) / static_cast<float>(std::numeric_limits<uint8>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<float, mediump, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<float, mediump> call(vecType<float, mediump> const & Min, vecType<float, mediump> const & Max)
+ {
+ return vecType<float, mediump>(compute_rand<uint16, mediump, vecType>::call()) / static_cast<float>(std::numeric_limits<uint16>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<float, highp, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<float, highp> call(vecType<float, highp> const & Min, vecType<float, highp> const & Max)
+ {
+ return vecType<float, highp>(compute_rand<uint32, highp, vecType>::call()) / static_cast<float>(std::numeric_limits<uint32>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<double, lowp, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<double, lowp> call(vecType<double, lowp> const & Min, vecType<double, lowp> const & Max)
+ {
+ return vecType<double, lowp>(compute_rand<uint16, lowp, vecType>::call()) / static_cast<double>(std::numeric_limits<uint16>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<double, mediump, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<double, mediump> call(vecType<double, mediump> const & Min, vecType<double, mediump> const & Max)
+ {
+ return vecType<double, mediump>(compute_rand<uint32, mediump, vecType>::call()) / static_cast<double>(std::numeric_limits<uint32>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<double, highp, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<double, highp> call(vecType<double, highp> const & Min, vecType<double, highp> const & Max)
+ {
+ return vecType<double, highp>(compute_rand<uint64, highp, vecType>::call()) / static_cast<double>(std::numeric_limits<uint64>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<long double, lowp, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<long double, lowp> call(vecType<long double, lowp> const & Min, vecType<long double, lowp> const & Max)
+ {
+ return vecType<long double, lowp>(compute_rand<uint32, lowp, vecType>::call()) / static_cast<long double>(std::numeric_limits<uint32>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<long double, mediump, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<long double, mediump> call(vecType<long double, mediump> const & Min, vecType<long double, mediump> const & Max)
+ {
+ return vecType<long double, mediump>(compute_rand<uint64, mediump, vecType>::call()) / static_cast<long double>(std::numeric_limits<uint64>::max()) * (Max - Min) + Min;
+ }
+ };
+
+ template <template <class, precision> class vecType>
+ struct compute_linearRand<long double, highp, vecType>
+ {
+ GLM_FUNC_QUALIFIER static vecType<long double, highp> call(vecType<long double, highp> const & Min, vecType<long double, highp> const & Max)
+ {
+ return vecType<long double, highp>(compute_rand<uint64, highp, vecType>::call()) / static_cast<long double>(std::numeric_limits<uint64>::max()) * (Max - Min) + Min;
+ }
+ };
+}//namespace detail
+
+ template <typename genType>
+ GLM_FUNC_QUALIFIER genType linearRand(genType Min, genType Max)
+ {
+ return detail::compute_linearRand<genType, highp, tvec1>::call(
+ tvec1<genType, highp>(Min),
+ tvec1<genType, highp>(Max)).x;
+ }
+
+ template <typename T, precision P, template <typename, precision> class vecType>
+ GLM_FUNC_QUALIFIER vecType<T, P> linearRand(vecType<T, P> const & Min, vecType<T, P> const & Max)
+ {
+ return detail::compute_linearRand<T, P, vecType>::call(Min, Max);
+ }
+
+ template <typename genType>
+ GLM_FUNC_QUALIFIER genType gaussRand(genType Mean, genType Deviation)
+ {
+ genType w, x1, x2;
+
+ do
+ {
+ x1 = linearRand(genType(-1), genType(1));
+ x2 = linearRand(genType(-1), genType(1));
+
+ w = x1 * x1 + x2 * x2;
+ } while(w > genType(1));
+
+ return x2 * Deviation * Deviation * sqrt((genType(-2) * log(w)) / w) + Mean;
+ }
+
+ template <typename T, precision P, template <typename, precision> class vecType>
+ GLM_FUNC_QUALIFIER vecType<T, P> gaussRand(vecType<T, P> const & Mean, vecType<T, P> const & Deviation)
+ {
+ return detail::functor2<T, P, vecType>::call(gaussRand, Mean, Deviation);
+ }
+
+ template <typename T>
+ GLM_FUNC_QUALIFIER tvec2<T, defaultp> diskRand(T Radius)
+ {
+ tvec2<T, defaultp> Result(T(0));
+ T LenRadius(T(0));
+
+ do
+ {
+ Result = linearRand(
+ tvec2<T, defaultp>(-Radius),
+ tvec2<T, defaultp>(Radius));
+ LenRadius = length(Result);
+ }
+ while(LenRadius > Radius);
+
+ return Result;
+ }
+
+ template <typename T>
+ GLM_FUNC_QUALIFIER tvec3<T, defaultp> ballRand(T Radius)
+ {
+ tvec3<T, defaultp> Result(T(0));
+ T LenRadius(T(0));
+
+ do
+ {
+ Result = linearRand(
+ tvec3<T, defaultp>(-Radius),
+ tvec3<T, defaultp>(Radius));
+ LenRadius = length(Result);
+ }
+ while(LenRadius > Radius);
+
+ return Result;
+ }
+
+ template <typename T>
+ GLM_FUNC_QUALIFIER tvec2<T, defaultp> circularRand(T Radius)
+ {
+ T a = linearRand(T(0), T(6.283185307179586476925286766559f));
+ return tvec2<T, defaultp>(cos(a), sin(a)) * Radius;
+ }
+
+ template <typename T>
+ GLM_FUNC_QUALIFIER tvec3<T, defaultp> sphericalRand(T Radius)
+ {
+ T z = linearRand(T(-1), T(1));
+ T a = linearRand(T(0), T(6.283185307179586476925286766559f));
+
+ T r = sqrt(T(1) - z * z);
+
+ T x = r * cos(a);
+ T y = r * sin(a);
+
+ return tvec3<T, defaultp>(x, y, z) * Radius;
+ }
+}//namespace glm