initial implemenation of templated classes :
Matrix class first.
This commit is contained in:
marauder2k7 2024-07-27 15:29:54 +01:00
parent c0dec83a21
commit dd25f1c58a
2 changed files with 653 additions and 0 deletions

View file

@ -209,3 +209,377 @@ EngineFieldTable::Field MatrixFEngineExport::getMatrixField()
typedef MatrixF ThisType;
return _FIELD_AS(F32, m, m, 16, "");
}
//------------------------------------
// Templatized matrix class to replace MATRIXF above
// row-major for now, since torque says it uses that
// but in future could cut down on transpose calls if
// we switch to column major.
//------------------------------------
template<typename DATA_TYPE, U32 rows, U32 cols>
const Matrix<DATA_TYPE, rows, cols> Matrix<DATA_TYPE, rows, cols>::Identity = []() {
Matrix<DATA_TYPE, rows, cols> identity(true);
return identity;
}();
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>::Matrix(const EulerF& e)
{
set(e);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::set(const EulerF& e)
{
// when the template refactor is done, euler will be able to be setup in different ways
AssertFatal(rows >= 3 && cols >= 3, "EulerF can only initialize 3x3 or more");
static_assert(std::is_same<DATA_TYPE, float>::value, "Can only initialize eulers with floats for now");
F32 cosPitch, sinPitch;
mSinCos(e.x, sinPitch, cosPitch);
F32 cosYaw, sinYaw;
mSinCos(e.y, sinYaw, cosYaw);
F32 cosRoll, sinRoll;
mSinCos(e.z, sinRoll, cosRoll);
enum {
AXIS_X = (1 << 0),
AXIS_Y = (1 << 1),
AXIS_Z = (1 << 2)
};
U32 axis = 0;
if (e.x != 0.0f) axis |= AXIS_X;
if (e.y != 0.0f) axis |= AXIS_Y;
if (e.z != 0.0f) axis |= AXIS_Z;
switch (axis) {
case 0:
(*this) = Matrix<DATA_TYPE, rows, cols>(true);
break;
case AXIS_X:
(*this)(0, 0) = 1.0f; (*this)(1, 0) = 0.0f; (*this)(2, 0) = 0.0f;
(*this)(0, 1) = 0.0f; (*this)(1, 1) = cosPitch; (*this)(2, 1) = -sinPitch;
(*this)(0, 2) = 0.0f; (*this)(1, 2) = sinPitch; (*this)(2, 2) = cosPitch;
break;
case AXIS_Y:
(*this)(0, 0) = cosYaw; (*this)(1, 0) = 0.0f; (*this)(2, 0) = sinYaw;
(*this)(0, 1) = 0.0f; (*this)(1, 1) = 1.0f; (*this)(2, 1) = 0.0f;
(*this)(0, 2) = -sinYaw; (*this)(1, 2) = 0.0f; (*this)(2, 2) = cosYaw;
break;
case AXIS_Z:
(*this)(0, 0) = cosRoll; (*this)(1, 0) = -sinRoll; (*this)(2, 0) = 0.0f;
(*this)(0, 1) = sinRoll; (*this)(1, 1) = cosRoll; (*this)(2, 1) = 0.0f;
(*this)(0, 2) = 0.0f; (*this)(1, 2) = 0.0f; (*this)(2, 2) = 0.0f;
break;
default:
F32 r1 = cosYaw * cosRoll;
F32 r2 = cosYaw * sinRoll;
F32 r3 = sinYaw * cosRoll;
F32 r4 = sinYaw * sinRoll;
// the matrix looks like this:
// r1 - (r4 * sin(x)) r2 + (r3 * sin(x)) -cos(x) * sin(y)
// -cos(x) * sin(z) cos(x) * cos(z) sin(x)
// r3 + (r2 * sin(x)) r4 - (r1 * sin(x)) cos(x) * cos(y)
//
// where:
// r1 = cos(y) * cos(z)
// r2 = cos(y) * sin(z)
// r3 = sin(y) * cos(z)
// r4 = sin(y) * sin(z)
// init the euler 3x3 rotation matrix.
(*this)(0, 0) = r1 - (r4 * sinPitch); (*this)(1, 0) = -cosPitch * sinRoll; (*this)(2, 0) = r3 + (r2 * sinPitch);
(*this)(0, 1) = r2 + (r3 * sinPitch); (*this)(1, 1) = cosPitch * cosRoll; (*this)(2, 1) = r4 - (r1 * sinPitch);
(*this)(0, 2) = -cosPitch * sinYaw; (*this)(1, 2) = sinPitch; (*this)(2, 2) = cosPitch * cosYaw;
break;
}
if (rows == 4) {
(*this)(3, 0) = 0.0f;
(*this)(3, 1) = 0.0f;
(*this)(3, 2) = 0.0f;
}
if (cols == 4) {
(*this)(0, 3) = 0.0f;
(*this)(1, 3) = 0.0f;
(*this)(2, 3) = 0.0f;
}
if (rows == 4 && cols == 4) {
(*this)(3, 3) = 1.0f;
}
return(*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>::Matrix(const EulerF& e, const Point3F p)
{
set(e, p);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::set(const EulerF& e, const Point3F p)
{
AssertFatal(rows >= 3 && cols >= 4, "Euler and Point can only initialize 3x4 or more");
// call set euler, this already sets the last row if it exists.
set(e);
// does this need to multiply with the result of the euler? or are we just setting position.
(*this)(0, 3) = p.x;
(*this)(1, 3) = p.y;
(*this)(2, 3) = p.z;
return (*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::inverse()
{
// TODO: insert return statement here
AssertFatal(rows == cols, "Can only perform inverse on square matrices.");
const U32 size = rows;
// Create augmented matrix [this | I]
Matrix<DATA_TYPE, size, 2 * size> augmentedMatrix;
Matrix<DATA_TYPE, size, size> resultMatrix;
for (U32 i = 0; i < size; i++) {
for (U32 j = 0; j < size; j++) {
augmentedMatrix(i, j) = (*this)(i, j);
augmentedMatrix(i, j + size) = (i == j) ? static_cast<DATA_TYPE>(1) : static_cast<DATA_TYPE>(0);
}
}
// Apply gauss-joran elimination
for (U32 i = 0; i < size; i++) {
U32 pivotRow = i;
for (U32 k = i + 1; k < size; k++) {
// use std::abs until the templated math functions are in place.
if (std::abs(augmentedMatrix(k, i)) > std::abs(augmentedMatrix(pivotRow, i))) {
pivotRow = k;
}
}
// Swap if needed.
if (i != pivotRow) {
for (U32 j = 0; j < 2 * size; j++) {
std::swap(augmentedMatrix(i, j), augmentedMatrix(pivotRow, j));
}
}
// Early out if pivot is 0, return a new empty matrix.
if (augmentedMatrix(i, i) == static_cast<DATA_TYPE>(0)) {
return Matrix<DATA_TYPE, rows, cols>();
}
DATA_TYPE pivotVal = augmentedMatrix(i, i);
// scale the pivot
for (U32 j = 0; j < 2 * size; ++j) {
augmentedMatrix(i, j) /= pivotVal;
}
// Eliminate the current column in all other rows
for (std::size_t k = 0; k < size; k++) {
if (k != i) {
DATA_TYPE factor = augmentedMatrix(k, i);
for (std::size_t j = 0; j < 2 * size; ++j) {
augmentedMatrix(k, j) -= factor * augmentedMatrix(i, j);
}
}
}
}
for (U32 i = 0; i < size; i++) {
for (U32 j = 0; j < size; j++) {
resultMatrix(i, j) = augmentedMatrix(i, j + size);
}
}
return resultMatrix;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
void Matrix<DATA_TYPE, rows, cols>::invert()
{
(*this) = inverse();
}
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::setCrossProduct(const Point3F& p)
{
AssertFatal(rows == 4 && cols == 4, "Cross product only supported on 4x4 for now");
(*this)(0, 0) = 0;
(*this)(0, 1) = -p.z;
(*this)(0, 2) = p.y;
(*this)(0, 3) = 0;
(*this)(1, 0) = p.z;
(*this)(1, 1) = 0;
(*this)(1, 2) = -p.x;
(*this)(1, 3) = 0;
(*this)(2, 0) = -p.y;
(*this)(2, 1) = p.x;
(*this)(2, 2) = 0;
(*this)(2, 3) = 0;
(*this)(3, 0) = 0;
(*this)(3, 1) = 0;
(*this)(3, 2) = 0;
(*this)(3, 3) = 1;
return (*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::setTensorProduct(const Point3F& p, const Point3F& q)
{
AssertFatal(rows == 4 && cols == 4, "Tensor product only supported on 4x4 for now");
(*this)(0, 0) = p.x * q.x;
(*this)(0, 1) = p.x * q.y;
(*this)(0, 2) = p.x * q.z;
(*this)(0, 3) = 0;
(*this)(1, 0) = p.y * q.x;
(*this)(1, 1) = p.y * q.y;
(*this)(1, 2) = p.y * q.z;
(*this)(1, 3) = 0;
(*this)(2, 0) = p.z * q.x;
(*this)(2, 1) = p.z * q.y;
(*this)(2, 2) = p.z * q.z;
(*this)(2, 3) = 0;
(*this)(3, 0) = 0;
(*this)(3, 1) = 0;
(*this)(3, 2) = 0;
(*this)(3, 3) = 1;
return (*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
bool Matrix<DATA_TYPE, rows, cols>::isAffine() const
{
if ((*this)(rows - 1, cols - 1) != 1.0f) {
return false;
}
for (U32 col = 0; col < cols - 1; ++col) {
if ((*this)(rows - 1, col) != 0.0f) {
return false;
}
}
Point3F one, two, three;
getColumn(0, &one);
getColumn(1, &two);
getColumn(2, &three);
// check columns
{
if (mDot(one, two) > 0.0001f ||
mDot(one, three) > 0.0001f ||
mDot(two, three) > 0.0001f)
return false;
if (mFabs(1.0f - one.lenSquared()) > 0.0001f ||
mFabs(1.0f - two.lenSquared()) > 0.0001f ||
mFabs(1.0f - three.lenSquared()) > 0.0001f)
return false;
}
getRow(0, &one);
getRow(1, &two);
getRow(2, &three);
// check rows
{
if (mDot(one, two) > 0.0001f ||
mDot(one, three) > 0.0001f ||
mDot(two, three) > 0.0001f)
return false;
if (mFabs(1.0f - one.lenSquared()) > 0.0001f ||
mFabs(1.0f - two.lenSquared()) > 0.0001f ||
mFabs(1.0f - three.lenSquared()) > 0.0001f)
return false;
}
return true;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
EulerF Matrix<DATA_TYPE, rows, cols>::toEuler() const
{
AssertFatal(rows >= 3 && cols >= 3, "Euler rotations require at least a 3x3 matrix.");
// Extract rotation matrix components
const DATA_TYPE m00 = (*this)(0, 0);
const DATA_TYPE m01 = (*this)(0, 1);
const DATA_TYPE m02 = (*this)(0, 2);
const DATA_TYPE m10 = (*this)(1, 0);
const DATA_TYPE m11 = (*this)(1, 1);
const DATA_TYPE m21 = (*this)(2, 1);
const DATA_TYPE m22 = (*this)(2, 2);
// like all others assume float for now.
EulerF r;
r.x = mAsin(mClampF(m21, -1.0, 1.0));
if (mCos(r.x) != 0.0f) {
r.y = mAtan2(-m02, m22); // yaw
r.z = mAtan2(-m10, m11); // roll
}
else {
r.y = 0.0f;
r.z = mAtan2(m01, m00); // this rolls when pitch is +90 degrees
}
return r;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
void Matrix<DATA_TYPE, rows, cols>::dumpMatrix(const char* caption) const
{
U32 size = (caption == NULL) ? 0 : dStrlen(caption);
FrameTemp<char> spacer(size + 1);
char* spacerRef = spacer;
// is_floating_point should return true for floats and doubles.
const char* formatSpec = std::is_floating_point_v<DATA_TYPE> ? " %-8.4f" : " %d";
dMemset(spacerRef, ' ', size);
// null terminate.
spacerRef[size] = '\0';
/*Con::printf("%s = | %-8.4f %-8.4f %-8.4f %-8.4f |", caption, m[idx(0, 0)], m[idx(0, 1)], m[idx(0, 2)], m[idx(0, 3)]);
Con::printf("%s | %-8.4f %-8.4f %-8.4f %-8.4f |", spacerRef, m[idx(1, 0)], m[idx(1, 1)], m[idx(1, 2)], m[idx(1, 3)]);
Con::printf("%s | %-8.4f %-8.4f %-8.4f %-8.4f |", spacerRef, m[idx(2, 0)], m[idx(2, 1)], m[idx(2, 2)], m[idx(2, 3)]);
Con::printf("%s | %-8.4f %-8.4f %-8.4f %-8.4f |", spacerRef, m[idx(3, 0)], m[idx(3, 1)], m[idx(3, 2)], m[idx(3, 3)]);*/
StringBuilder str;
str.format("%s = |", caption);
for (U32 i = 0; i < rows; i++) {
if (i > 0) {
str.append(spacerRef);
}
for (U32 j = 0; j < cols; j++) {
str.format(formatSpec, (*this)(i, j));
}
str.append(" |\n");
}
Con::printf("%s", str.end().c_str());
}

View file

@ -620,4 +620,283 @@ inline void mTransformPlane(const MatrixF& mat, const Point3F& scale, const Plan
m_matF_x_scale_x_planeF(mat, &scale.x, &plane.x, &result->x);
}
//------------------------------------
// Templatized matrix class to replace MATRIXF above
// row-major for now, since torque says it uses that
// but in future could cut down on transpose calls if
// we switch to column major.
//------------------------------------
template<typename DATA_TYPE, U32 rows, U32 cols>
class Matrix {
friend class MatrixTemplateExport;
private:
DATA_TYPE data[rows * cols];
public:
static_assert(rows >= 2 && cols >= 2, "Matrix must have at least 2 rows and 2 cols.");
// ------ Setters and initializers ------
explicit Matrix(bool identity = false) {
std::fill(data, data + (rows * cols), DATA_TYPE(0));
if (identity) {
for (U32 i = 0; i < rows; i++) {
for (U32 j = 0; j < cols; j++) {
// others already get filled with 0
if (j == i)
(*this)(i, j) = static_cast<DATA_TYPE>(1);
}
}
}
}
explicit Matrix(const EulerF& e);
/// Make this an identity matrix.
Matrix<DATA_TYPE, rows, cols>& identity();
Matrix<DATA_TYPE, rows, cols>& set(const EulerF& e);
Matrix(const EulerF& e, const Point3F p);
Matrix<DATA_TYPE, rows, cols>& set(const EulerF& e, const Point3F p);
Matrix<DATA_TYPE, rows, cols>& inverse();
Matrix<DATA_TYPE, rows, cols>& transpose();
void invert();
Matrix<DATA_TYPE, rows, cols>& setCrossProduct(const Point3F& p);
Matrix<DATA_TYPE, rows, cols>& setTensorProduct(const Point3F& p, const Point3F& q);
/// M * Matrix(p) -> M
Matrix<DATA_TYPE, rows, cols>& scale(const Point3F& s);
Matrix<DATA_TYPE, rows, cols>& scale(DATA_TYPE s) { return scale(Point3F(s, s, s)); }
// ------ Getters ------
bool isAffine() const;
Point3F getScale() const;
EulerF toEuler() const;
Point3F getPosition() const;
void getColumn(S32 col, Point4F* cptr) const;
Point4F getColumn4F(S32 col) const { Point4F ret; getColumn(col, &ret); return ret; }
void getColumn(S32 col, Point3F* cptr) const;
Point3F getColumn3F(S32 col) const { Point3F ret; getColumn(col, &ret); return ret; }
void getRow(S32 row, Point4F* cptr) const;
Point4F getRow4F(S32 row) const { Point4F ret; getRow(row, &ret); return ret; }
void getRow(S32 row, Point3F* cptr) const;
Point3F getRow3F(S32 row) const { Point3F ret; getRow(row, &ret); return ret; }
DATA_TYPE* getData() {
return data;
}
const DATA_TYPE* getData() const {
return data;
}
void dumpMatrix(const char* caption = NULL) const;
// Static identity matrix
static const Matrix Identity;
// ------ Operators ------
operator DATA_TYPE* () { return (data); }
operator const DATA_TYPE* () const { return (DATA_TYPE*)(data); }
DATA_TYPE& operator()(U32 row, U32 col) {
if (row >= rows || col >= cols)
AssertFatal(false, "Matrix indices out of range");
return data[col * rows + row];
}
const DATA_TYPE& operator()(U32 row, U32 col) const {
if (row >= rows || col >= cols)
AssertFatal(false, "Matrix indices out of range");
return data[col * rows + row];
}
};
//--------------------------------------------
// INLINE FUNCTIONS
//--------------------------------------------
template<typename DATA_TYPE, U32 rows, U32 cols>
inline Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::transpose()
{
// square matrices can just swap, non square requires a temp mat.
if (rows == cols) {
for (U32 i = 0; i < rows; i++) {
for (U32 j = 0; j < cols; j++) {
std::swap((*this)(j, i), (*this)(i, j));
}
}
}
else {
Matrix<DATA_TYPE, rows, cols> result;
for (U32 i = 0; i < rows; i++) {
for (U32 j = 0; j < cols; j++) {
result(j, i) = (*this)(i, j);
}
}
std::copy(std::begin(result.data), std::end(result.data), std::begin(data));
}
return (*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::identity()
{
for (U32 i = 0; i < rows; i++) {
for (U32 j = 0; j < cols; j++) {
if (j == i)
(*this)(i, j) = static_cast<DATA_TYPE>(1);
else
(*this)(i, j) = static_cast<DATA_TYPE>(0);
}
}
return (*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline Matrix<DATA_TYPE, rows, cols>& Matrix<DATA_TYPE, rows, cols>::scale(const Point3F& s)
{
// torques scale applies directly, does not create another matrix to multiply with the translation matrix.
AssertFatal(rows >= 3 && cols >= 3, "Scale can only be applied 3x3 or more");
for (U32 i = 0; i < 3; i++) {
for (U32 j = 0; j < 3; j++) {
DATA_TYPE scale = (i == 0) ? s.x : (i == 1) ? s.y : s.z;
(*this)(i, j) *= scale;
}
}
return (*this);
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline Point3F Matrix<DATA_TYPE, rows, cols>::getScale() const
{
// this function assumes the matrix has scale applied through the scale(const Point3F& s) function.
// for now assume float since we have point3F.
AssertFatal(rows >= 3 && cols >= 3, "Scale can only be applied 3x3 or more");
Point3F scale;
scale.x = mSqrt((*this)(0, 0) * (*this)(0, 0) + (*this)(1, 0) * (*this)(1, 0) + (*this)(2, 0) * (*this)(2, 0));
scale.y = mSqrt((*this)(0, 1) * (*this)(0, 1) + (*this)(1, 1) * (*this)(1, 1) + (*this)(2, 1) * (*this)(2, 1));
scale.z = mSqrt((*this)(0, 2) * (*this)(0, 2) + (*this)(1, 2) * (*this)(1, 2) + (*this)(2, 2) * (*this)(2, 2));
return scale;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline Point3F Matrix<DATA_TYPE, rows, cols>::getPosition() const
{
Point3F pos;
getColumn(3, &pos);
return pos;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline void Matrix<DATA_TYPE, rows, cols>::getColumn(S32 col, Point4F* cptr) const
{
if (rows >= 2)
{
cptr->x = (*this)(0, col);
cptr->y = (*this)(1, col);
}
if (rows >= 3)
cptr->z = (*this)(2, col);
else
cptr->z = 0.0f;
if (rows >= 4)
cptr->w = (*this)(3, col);
else
cptr->w = 0.0f;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline void Matrix<DATA_TYPE, rows, cols>::getColumn(S32 col, Point3F* cptr) const
{
if (rows >= 2)
{
cptr->x = (*this)(0, col);
cptr->y = (*this)(1, col);
}
if (rows >= 3)
cptr->z = (*this)(2, col);
else
cptr->z = 0.0f;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline void Matrix<DATA_TYPE, rows, cols>::getRow(S32 row, Point4F* cptr) const
{
if (cols >= 2)
{
cptr->x = (*this)(row, 0);
cptr->y = (*this)(row, 1);
}
if (cols >= 3)
cptr->z = (*this)(row, 2);
else
cptr->z = 0.0f;
if (cols >= 4)
cptr->w = (*this)(row, 3);
else
cptr->w = 0.0f;
}
template<typename DATA_TYPE, U32 rows, U32 cols>
inline void Matrix<DATA_TYPE, rows, cols>::getRow(S32 row, Point3F* cptr) const
{
if (cols >= 2)
{
cptr->x = (*this)(row, 0);
cptr->y = (*this)(row, 1);
}
if (cols >= 3)
cptr->z = (*this)(row, 2);
else
cptr->z = 0.0f;
}
//--------------------------------------------
// INLINE FUNCTIONS END
//--------------------------------------------
typedef Matrix<F32, 4, 4> Matrix4F;
class MatrixTemplateExport
{
public:
template <typename T, U32 rows, U32 cols>
static EngineFieldTable::Field getMatrixField();
};
template<typename T, U32 rows, U32 cols>
inline EngineFieldTable::Field MatrixTemplateExport::getMatrixField()
{
typedef Matrix<T, rows, cols> ThisType;
return _FIELD_AS(T, data, data, rows * cols, "");
}
#endif //_MMATRIX_H_