glm/test/gtx/gtx_dual_quaternion.cpp
2014-12-10 02:52:30 +01:00

220 lines
7.4 KiB
C++

///////////////////////////////////////////////////////////////////////////////////
/// OpenGL Mathematics (glm.g-truc.net)
///
/// Copyright (c) 2005 - 2014 G-Truc Creation (www.g-truc.net)
/// Permission is hereby granted, free of charge, to any person obtaining a copy
/// of this software and associated documentation files (the "Software"), to deal
/// in the Software without restriction, including without limitation the rights
/// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
/// copies of the Software, and to permit persons to whom the Software is
/// furnished to do so, subject to the following conditions:
///
/// The above copyright notice and this permission notice shall be included in
/// all copies or substantial portions of the Software.
///
/// Restrictions:
/// By making use of the Software for military purposes, you choose to make
/// a Bunny unhappy.
///
/// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
/// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
/// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
/// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
/// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
/// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
/// THE SOFTWARE.
///
/// @file test/gtx/gtx_dual_quaternion.cpp
/// @date 2013-02-10 / 2014-11-25
/// @author Christophe Riccio
///////////////////////////////////////////////////////////////////////////////////
#include <glm/gtx/dual_quaternion.hpp>
#include <glm/gtc/matrix_transform.hpp>
#include <glm/gtc/epsilon.hpp>
#include <glm/gtx/euler_angles.hpp>
#include <glm/vector_relational.hpp>
#if GLM_HAS_TRIVIAL_QUERIES
# include <type_traits>
#endif
int myrand()
{
static int holdrand = 1;
return (((holdrand = holdrand * 214013L + 2531011L) >> 16) & 0x7fff);
}
float myfrand() // returns values from -1 to 1 inclusive
{
return float(double(myrand()) / double( 0x7ffff )) * 2.0f - 1.0f;
}
int test_dquat_type()
{
glm::dvec3 vA;
glm::dquat dqA,dqB;
glm::ddualquat C(dqA,dqB);
glm::ddualquat B(dqA);
glm::ddualquat D(dqA,vA);
return 0;
}
int test_scalars()
{
float const Epsilon = 0.0001f;
int Error(0);
glm::quat src_q1 = glm::quat(1.0f,2.0f,3.0f,4.0f);
glm::quat src_q2 = glm::quat(5.0f,6.0f,7.0f,8.0f);
glm::dualquat src1(src_q1,src_q2);
{
glm::dualquat dst1 = src1 * 2.0f;
glm::dualquat dst2 = 2.0f * src1;
glm::dualquat dst3 = src1;
dst3 *= 2.0f;
glm::dualquat dstCmp(src_q1 * 2.0f,src_q2 * 2.0f);
Error += glm::all(glm::epsilonEqual(dst1.real,dstCmp.real, Epsilon)) && glm::all(glm::epsilonEqual(dst1.dual,dstCmp.dual, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(dst2.real,dstCmp.real, Epsilon)) && glm::all(glm::epsilonEqual(dst2.dual,dstCmp.dual, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(dst3.real,dstCmp.real, Epsilon)) && glm::all(glm::epsilonEqual(dst3.dual,dstCmp.dual, Epsilon)) ? 0 : 1;
}
{
glm::dualquat dst1 = src1 / 2.0f;
glm::dualquat dst2 = src1;
dst2 /= 2.0f;
glm::dualquat dstCmp(src_q1 / 2.0f,src_q2 / 2.0f);
Error += glm::all(glm::epsilonEqual(dst1.real,dstCmp.real, Epsilon)) && glm::all(glm::epsilonEqual(dst1.dual,dstCmp.dual, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(dst2.real,dstCmp.real, Epsilon)) && glm::all(glm::epsilonEqual(dst2.dual,dstCmp.dual, Epsilon)) ? 0 : 1;
}
return Error;
}
int test_inverse()
{
int Error(0);
float const Epsilon = 0.0001f;
glm::dualquat dqid;
glm::mat4x4 mid(1.0f);
for (int j = 0; j < 100; ++j)
{
glm::mat4x4 rot = glm::yawPitchRoll(myfrand() * 360.0f, myfrand() * 360.0f, myfrand() * 360.0f);
glm::vec3 vt = glm::vec3(myfrand() * 10.0f, myfrand() * 10.0f, myfrand() * 10.0f);
glm::mat4x4 m = glm::translate(mid, vt) * rot;
glm::quat qr = glm::quat_cast(m);
glm::dualquat dq(qr);
glm::dualquat invdq = glm::inverse(dq);
glm::dualquat r1 = invdq * dq;
glm::dualquat r2 = dq * invdq;
Error += glm::all(glm::epsilonEqual(r1.real, dqid.real, Epsilon)) && glm::all(glm::epsilonEqual(r1.dual, dqid.dual, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(r2.real, dqid.real, Epsilon)) && glm::all(glm::epsilonEqual(r2.dual, dqid.dual, Epsilon)) ? 0 : 1;
// testing commutative property
glm::dualquat r ( glm::quat( myfrand() * glm::pi<float>() * 2.0f, myfrand(), myfrand(), myfrand() ),
glm::vec3(myfrand() * 10.0f, myfrand() * 10.0f, myfrand() * 10.0f) );
glm::dualquat riq = (r * invdq) * dq;
glm::dualquat rqi = (r * dq) * invdq;
Error += glm::all(glm::epsilonEqual(riq.real, rqi.real, Epsilon)) && glm::all(glm::epsilonEqual(riq.dual, rqi.dual, Epsilon)) ? 0 : 1;
}
return Error;
}
int test_mul()
{
int Error(0);
float const Epsilon = 0.0001f;
glm::mat4x4 mid(1.0f);
for (int j = 0; j < 100; ++j)
{
// generate random rotations and translations and compare transformed by matrix and dualquats random points
glm::vec3 vt1 = glm::vec3(myfrand() * 10.0f, myfrand() * 10.0f, myfrand() * 10.0f);
glm::vec3 vt2 = glm::vec3(myfrand() * 10.0f, myfrand() * 10.0f, myfrand() * 10.0f);
glm::mat4x4 rot1 = glm::yawPitchRoll(myfrand() * 360.0f, myfrand() * 360.0f, myfrand() * 360.0f);
glm::mat4x4 rot2 = glm::yawPitchRoll(myfrand() * 360.0f, myfrand() * 360.0f, myfrand() * 360.0f);
glm::mat4x4 m1 = glm::translate(mid, vt1) * rot1;
glm::mat4x4 m2 = glm::translate(mid, vt2) * rot2;
glm::mat4x4 m3 = m2 * m1;
glm::mat4x4 m4 = m1 * m2;
glm::quat qrot1 = glm::quat_cast(rot1);
glm::quat qrot2 = glm::quat_cast(rot2);
glm::dualquat dq1 = glm::dualquat(qrot1,vt1);
glm::dualquat dq2 = glm::dualquat(qrot2,vt2);
glm::dualquat dq3 = dq2 * dq1;
glm::dualquat dq4 = dq1 * dq2;
for (int i = 0; i < 100; ++i)
{
glm::vec4 src_pt = glm::vec4(myfrand() * 4.0f, myfrand() * 5.0f, myfrand() * 3.0f,1.0f);
// test both multiplication orders
glm::vec4 dst_pt_m3 = m3 * src_pt;
glm::vec4 dst_pt_dq3 = dq3 * src_pt;
glm::vec4 dst_pt_m3_i = glm::inverse(m3) * src_pt;
glm::vec4 dst_pt_dq3_i = src_pt * dq3;
glm::vec4 dst_pt_m4 = m4 * src_pt;
glm::vec4 dst_pt_dq4 = dq4 * src_pt;
glm::vec4 dst_pt_m4_i = glm::inverse(m4) * src_pt;
glm::vec4 dst_pt_dq4_i = src_pt * dq4;
Error += glm::all(glm::epsilonEqual(dst_pt_m3, dst_pt_dq3, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(dst_pt_m4, dst_pt_dq4, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(dst_pt_m3_i, dst_pt_dq3_i, Epsilon)) ? 0 : 1;
Error += glm::all(glm::epsilonEqual(dst_pt_m4_i, dst_pt_dq4_i, Epsilon)) ? 0 : 1;
}
}
return Error;
}
int test_dual_quat_ctr()
{
int Error(0);
# if GLM_HAS_TRIVIAL_QUERIES
// Error += std::is_trivially_default_constructible<glm::dualquat>::value ? 0 : 1;
// Error += std::is_trivially_default_constructible<glm::ddualquat>::value ? 0 : 1;
// Error += std::is_trivially_copy_assignable<glm::dualquat>::value ? 0 : 1;
// Error += std::is_trivially_copy_assignable<glm::ddualquat>::value ? 0 : 1;
Error += std::is_trivially_copyable<glm::dualquat>::value ? 0 : 1;
Error += std::is_trivially_copyable<glm::ddualquat>::value ? 0 : 1;
Error += std::is_copy_constructible<glm::dualquat>::value ? 0 : 1;
Error += std::is_copy_constructible<glm::ddualquat>::value ? 0 : 1;
# endif
return Error;
}
int main()
{
int Error(0);
Error += test_dual_quat_ctr();
Error += test_dquat_type();
Error += test_scalars();
Error += test_inverse();
Error += test_mul();
return Error;
}