redgpu/bullet: Bullet Physics SDK: real-time collision detection and multi-physics simulation for VR, games, visual effects, robotics, machine learning etc.

src/LinearMath/btMatrixX.h
533 строки11 KB
Erwin Coumans
fix slow full btMatrixX operator* (was not used)
09 окт 2019, 05:31
09 окт 2019, 05:313a4159c
/*
Bullet Continuous Collision Detection and Physics Library
Copyright (c) 2003-2013 Erwin Coumans  http://bulletphysics.org

This software is provided 'as-is', without any express or implied warranty.
In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose, 
including commercial applications, and to alter it and redistribute it freely, 
subject to the following restrictions:

1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
///original version written by Erwin Coumans, October 2013

#ifndef BT_MATRIX_X_H
#define BT_MATRIX_X_H

#include "LinearMath/btQuickprof.h"
#include "LinearMath/btAlignedObjectArray.h"
#include <stdio.h>

//#define BT_DEBUG_OSTREAM
#ifdef BT_DEBUG_OSTREAM
#include <iostream>
#include <iomanip>  // std::setw
#endif              //BT_DEBUG_OSTREAM

class btIntSortPredicate
{
public:
	bool operator()(const int& a, const int& b) const
	{
		return a < b;
	}
};

template <typename T>
struct btVectorX
{
	btAlignedObjectArray<T> m_storage;

	btVectorX()
	{
	}
	btVectorX(int numRows)
	{
		m_storage.resize(numRows);
	}

	void resize(int rows)
	{
		m_storage.resize(rows);
	}
	int cols() const
	{
		return 1;
	}
	int rows() const
	{
		return m_storage.size();
	}
	int size() const
	{
		return rows();
	}

	T nrm2() const
	{
		T norm = T(0);

		int nn = rows();

		{
			if (nn == 1)
			{
				norm = btFabs((*this)[0]);
			}
			else
			{
				T scale = 0.0;
				T ssq = 1.0;

				/* The following loop is equivalent to this call to the LAPACK
				 auxiliary routine:   CALL SLASSQ( N, X, INCX, SCALE, SSQ ) */

				for (int ix = 0; ix < nn; ix++)
				{
					if ((*this)[ix] != 0.0)
					{
						T absxi = btFabs((*this)[ix]);
						if (scale < absxi)
						{
							T temp;
							temp = scale / absxi;
							ssq = ssq * (temp * temp) + BT_ONE;
							scale = absxi;
						}
						else
						{
							T temp;
							temp = absxi / scale;
							ssq += temp * temp;
						}
					}
				}
				norm = scale * sqrt(ssq);
			}
		}
		return norm;
	}
	void setZero()
	{
		if (m_storage.size())
		{
			//	for (int i=0;i<m_storage.size();i++)
			//		m_storage[i]=0;
			//memset(&m_storage[0],0,sizeof(T)*m_storage.size());
			btSetZero(&m_storage[0], m_storage.size());
		}
	}
	const T& operator[](int index) const
	{
		return m_storage[index];
	}

	T& operator[](int index)
	{
		return m_storage[index];
	}

	T* getBufferPointerWritable()
	{
		return m_storage.size() ? &m_storage[0] : 0;
	}

	const T* getBufferPointer() const
	{
		return m_storage.size() ? &m_storage[0] : 0;
	}
};
/*
 template <typename T>
 void setElem(btMatrixX<T>& mat, int row, int col, T val)
 {
 mat.setElem(row,col,val);
 }
 */

template <typename T>
struct btMatrixX
{
	int m_rows;
	int m_cols;
	int m_operations;
	int m_resizeOperations;
	int m_setElemOperations;

	btAlignedObjectArray<T> m_storage;
	mutable btAlignedObjectArray<btAlignedObjectArray<int> > m_rowNonZeroElements1;

	T* getBufferPointerWritable()
	{
		return m_storage.size() ? &m_storage[0] : 0;
	}

	const T* getBufferPointer() const
	{
		return m_storage.size() ? &m_storage[0] : 0;
	}
	btMatrixX()
		: m_rows(0),
		  m_cols(0),
		  m_operations(0),
		  m_resizeOperations(0),
		  m_setElemOperations(0)
	{
	}
	btMatrixX(int rows, int cols)
		: m_rows(rows),
		  m_cols(cols),
		  m_operations(0),
		  m_resizeOperations(0),
		  m_setElemOperations(0)
	{
		resize(rows, cols);
	}
	void resize(int rows, int cols)
	{
		m_resizeOperations++;
		m_rows = rows;
		m_cols = cols;
		{
			BT_PROFILE("m_storage.resize");
			m_storage.resize(rows * cols);
		}
	}
	int cols() const
	{
		return m_cols;
	}
	int rows() const
	{
		return m_rows;
	}
	///we don't want this read/write operator(), because we cannot keep track of non-zero elements, use setElem instead
	/*T& operator() (int row,int col)
	{
		return m_storage[col*m_rows+row];
	}
	*/

	void addElem(int row, int col, T val)
	{
		if (val)
		{
			if (m_storage[col + row * m_cols] == 0.f)
			{
				setElem(row, col, val);
			}
			else
			{
				m_storage[row * m_cols + col] += val;
			}
		}
	}

	void setElem(int row, int col, T val)
	{
		m_setElemOperations++;
		m_storage[row * m_cols + col] = val;
	}

	void mulElem(int row, int col, T val)
	{
		m_setElemOperations++;
		//mul doesn't change sparsity info

		m_storage[row * m_cols + col] *= val;
	}

	void copyLowerToUpperTriangle()
	{
		int count = 0;
		for (int row = 0; row < rows(); row++)
		{
			for (int col = 0; col < row; col++)
			{
				setElem(col, row, (*this)(row, col));
				count++;
			}
		}
		//printf("copyLowerToUpperTriangle copied %d elements out of %dx%d=%d\n", count,rows(),cols(),cols()*rows());
	}

	const T& operator()(int row, int col) const
	{
		return m_storage[col + row * m_cols];
	}

	void setZero()
	{
		{
			BT_PROFILE("storage=0");
			if (m_storage.size())
			{
				btSetZero(&m_storage[0], m_storage.size());
			}
			//memset(&m_storage[0],0,sizeof(T)*m_storage.size());
			//for (int i=0;i<m_storage.size();i++)
			//			m_storage[i]=0;
		}
	}

	void setIdentity()
	{
		btAssert(rows() == cols());

		setZero();
		for (int row = 0; row < rows(); row++)
		{
			setElem(row, row, 1);
		}
	}

	void printMatrix(const char* msg) const
	{
		printf("%s ---------------------\n", msg);
		for (int i = 0; i < rows(); i++)
		{
			printf("\n");
			for (int j = 0; j < cols(); j++)
			{
				printf("%2.1f\t", (*this)(i, j));
			}
		}
		printf("\n---------------------\n");
	}

	void rowComputeNonZeroElements() const
	{
		m_rowNonZeroElements1.resize(rows());
		for (int i = 0; i < rows(); i++)
		{
			m_rowNonZeroElements1[i].resize(0);
			for (int j = 0; j < cols(); j++)
			{
				if ((*this)(i, j) != 0.f)
				{
					m_rowNonZeroElements1[i].push_back(j);
				}
			}
		}
	}
	btMatrixX transpose() const
	{
		//transpose is optimized for sparse matrices
		btMatrixX tr(m_cols, m_rows);
		tr.setZero();
		for (int i = 0; i < m_cols; i++)
			for (int j = 0; j < m_rows; j++)
			{
				T v = (*this)(j, i);
				if (v)
				{
					tr.setElem(i, j, v);
				}
			}
		return tr;
	}

	btMatrixX operator*(const btMatrixX& other)
	{
		//btMatrixX*btMatrixX implementation, brute force
		btAssert(cols() == other.rows());

		btMatrixX res(rows(), other.cols());
		res.setZero();
		//		BT_PROFILE("btMatrixX mul");
		for (int i = 0; i < rows(); ++i)
		{
			{
				for (int j = 0; j < other.cols(); ++j)
				{
					T dotProd = 0;
					{
						{
							int r = rows();
							int c = cols();

							for (int k = 0; k < cols(); k++)
							{
								T w = (*this)(i, k);
								if (other(k, j) != 0.f)
								{
									dotProd += w * other(k, j);
								}
							}
						}
					}
					if (dotProd)
						res.setElem(i, j, dotProd);
				}
			}
		}
		return res;
	}

	// this assumes the 4th and 8th rows of B and C are zero.
	void multiplyAdd2_p8r(const btScalar* B, const btScalar* C, int numRows, int numRowsOther, int row, int col)
	{
		const btScalar* bb = B;
		for (int i = 0; i < numRows; i++)
		{
			const btScalar* cc = C;
			for (int j = 0; j < numRowsOther; j++)
			{
				btScalar sum;
				sum = bb[0] * cc[0];
				sum += bb[1] * cc[1];
				sum += bb[2] * cc[2];
				sum += bb[4] * cc[4];
				sum += bb[5] * cc[5];
				sum += bb[6] * cc[6];
				addElem(row + i, col + j, sum);
				cc += 8;
			}
			bb += 8;
		}
	}

	void multiply2_p8r(const btScalar* B, const btScalar* C, int numRows, int numRowsOther, int row, int col)
	{
		btAssert(numRows > 0 && numRowsOther > 0 && B && C);
		const btScalar* bb = B;
		for (int i = 0; i < numRows; i++)
		{
			const btScalar* cc = C;
			for (int j = 0; j < numRowsOther; j++)
			{
				btScalar sum;
				sum = bb[0] * cc[0];
				sum += bb[1] * cc[1];
				sum += bb[2] * cc[2];
				sum += bb[4] * cc[4];
				sum += bb[5] * cc[5];
				sum += bb[6] * cc[6];
				setElem(row + i, col + j, sum);
				cc += 8;
			}
			bb += 8;
		}
	}

	void setSubMatrix(int rowstart, int colstart, int rowend, int colend, const T value)
	{
		int numRows = rowend + 1 - rowstart;
		int numCols = colend + 1 - colstart;

		for (int row = 0; row < numRows; row++)
		{
			for (int col = 0; col < numCols; col++)
			{
				setElem(rowstart + row, colstart + col, value);
			}
		}
	}

	void setSubMatrix(int rowstart, int colstart, int rowend, int colend, const btMatrixX& block)
	{
		btAssert(rowend + 1 - rowstart == block.rows());
		btAssert(colend + 1 - colstart == block.cols());
		for (int row = 0; row < block.rows(); row++)
		{
			for (int col = 0; col < block.cols(); col++)
			{
				setElem(rowstart + row, colstart + col, block(row, col));
			}
		}
	}
	void setSubMatrix(int rowstart, int colstart, int rowend, int colend, const btVectorX<T>& block)
	{
		btAssert(rowend + 1 - rowstart == block.rows());
		btAssert(colend + 1 - colstart == block.cols());
		for (int row = 0; row < block.rows(); row++)
		{
			for (int col = 0; col < block.cols(); col++)
			{
				setElem(rowstart + row, colstart + col, block[row]);
			}
		}
	}

	btMatrixX negative()
	{
		btMatrixX neg(rows(), cols());
		for (int i = 0; i < rows(); i++)
			for (int j = 0; j < cols(); j++)
			{
				T v = (*this)(i, j);
				neg.setElem(i, j, -v);
			}
		return neg;
	}
};

typedef btMatrixX<float> btMatrixXf;
typedef btVectorX<float> btVectorXf;

typedef btMatrixX<double> btMatrixXd;
typedef btVectorX<double> btVectorXd;

#ifdef BT_DEBUG_OSTREAM
template <typename T>
std::ostream& operator<<(std::ostream& os, const btMatrixX<T>& mat)
{
	os << " [";
	//printf("%s ---------------------\n",msg);
	for (int i = 0; i < mat.rows(); i++)
	{
		for (int j = 0; j < mat.cols(); j++)
		{
			os << std::setw(12) << mat(i, j);
		}
		if (i != mat.rows() - 1)
			os << std::endl
			   << "  ";
	}
	os << " ]";
	//printf("\n---------------------\n");

	return os;
}
template <typename T>
std::ostream& operator<<(std::ostream& os, const btVectorX<T>& mat)
{
	os << " [";
	//printf("%s ---------------------\n",msg);
	for (int i = 0; i < mat.rows(); i++)
	{
		os << std::setw(12) << mat[i];
		if (i != mat.rows() - 1)
			os << std::endl
			   << "  ";
	}
	os << " ]";
	//printf("\n---------------------\n");

	return os;
}

#endif  //BT_DEBUG_OSTREAM

inline void setElem(btMatrixXd& mat, int row, int col, double val)
{
	mat.setElem(row, col, val);
}

inline void setElem(btMatrixXf& mat, int row, int col, float val)
{
	mat.setElem(row, col, val);
}

#ifdef BT_USE_DOUBLE_PRECISION
#define btVectorXu btVectorXd
#define btMatrixXu btMatrixXd
#else
#define btVectorXu btVectorXf
#define btMatrixXu btMatrixXf
#endif  //BT_USE_DOUBLE_PRECISION

#endif  //BT_MATRIX_H_H