【Java毕业设计】毕业设计的相关代码 Java版本 后续应该还有C++版本的.zip

package com.mcy.core.cannonAndFox; import com.google.common.util.concurrent.ThreadFactoryBuilder; import java.util.concurrent.*; /** * 用Cannon算法实现矩阵相乘 * @author manchaoyang 2019/1/15 */ public class MatrixMultiplicationAlgorithmCannon extends MatrixMultiplicationAlgorithm { private static int AVAILABLE_PROCESSORS = 20; private ThreadFactory cannonComputeThreadFactory = new ThreadFactoryBuilder() .setNameFormat("cannonAndFox-compute-matrix-pool-%d").build(); private ExecutorService cannonThreadPool = new ThreadPoolExecutor(AVAILABLE_PROCESSORS, AVAILABLE_PROCESSORS, 100L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue<Runnable>(1024), cannonComputeThreadFactory, new ThreadPoolExecutor.AbortPolicy()); /** * 按照一定的规则将矩阵分成的块循环整体左移 * * @param matrix * @param subMatrixLength * @param step */ private void loopLeftShift(long[][] matrix, int subMatrixLength, int step) { int rowMatrixCount = getMatrixCount(matrix.length, subMatrixLength); for (int i = 0; i < rowMatrixCount; i++) { loopLeftShiftByRow(matrix, subMatrixLength, i, step); } } /** * 按照一定的规则将矩阵分成的块循环整体上移 * * @param matrix * @param subMatrixLength * @param step */ private void loopUpShift(long[][] matrix, int subMatrixLength, int step) { int colMatrixCount = getMatrixCount(matrix.length, subMatrixLength); for (int j = 0; j < colMatrixCount; j++) { loopUpShiftByColumn(matrix, subMatrixLength, j, step); } } /** * 按照一定的规则将矩阵分成的块按某列循环移动 * * @param matrix * @param subMatrixLength * @param step */ private void loopLeftShiftByRow(long[][] matrix, int subMatrixLength, int rowNum, int step) { int colMatrixCount = getMatrixCount(matrix.length, subMatrixLength); long[][] subMatrix = extractSubMatrix(matrix, rowNum, 0, subMatrixLength); int nowIndex = 0; for (int k = 0; k < colMatrixCount; k++) { int nextIndex = (nowIndex + step) % colMatrixCount; subMatrix = copyMatrixToPointIndex(matrix, subMatrix, rowNum, nextIndex); nowIndex = nextIndex; } } /** * 按照一定的规则将矩阵分成的块按某行循环移动 * * @param matrix * @param subMatrixLength * @param step */ private void loopUpShiftByColumn(long[][] matrix, int subMatrixLength, int columnNum, int step) { int colMatrixCount = getMatrixCount(matrix.length, subMatrixLength); long[][] subMatrix = extractSubMatrix(matrix, 0, columnNum, subMatrixLength); int nowIndex = 0; for (int k = 0; k < colMatrixCount; k++) { int nextIndex = (nowIndex + step) % colMatrixCount; subMatrix = copyMatrixToPointIndex(matrix, subMatrix, nextIndex, columnNum); nowIndex = nextIndex; } } private int getMatrixCount(int matrixLength, int subMatrixLength) { return matrixLength / subMatrixLength; } /** * 按照子矩阵编号抽取子矩阵 * * @param matrix * @param i * @param j * @return */ private long[][] extractSubMatrix(long[][] matrix, int i, int j, int subMatrixLength) { long[][] subMatrix = new long[subMatrixLength][subMatrixLength]; int begRow = i * subMatrixLength; int begCol = j * subMatrixLength; for (int row = 0; row < subMatrixLength; row++) { for (int col = 0; col < subMatrixLength; col++) { subMatrix[row][col] = matrix[begRow + row][begCol + col]; } } return subMatrix; } /** * 将指定的子矩阵拷贝到相应的位置将原子矩阵元素返回 * * @param matrix * @param subMatrix * @param i * @param j * @return */ private long[][] copyMatrixToPointIndex(long[][] matrix, long[][] subMatrix, int i, int j) { int subMatrixLength = subMatrix.length; long[][] oldSubMatrix = new long[subMatrixLength][subMatrixLength]; int begRow = i * subMatrixLength; int begCol = j * subMatrixLength; for (int row = 0; row < subMatrixLength; row++) { for (int col = 0; col < subMatrixLength; col++) { oldSubMatrix[row][col] = matrix[begRow + row][begCol + col]; matrix[begRow + row][begCol + col] = subMatrix[row][col]; } } return oldSubMatrix; } private int getSubMatrixLength(int matrixLength) { int mod = (int) Math.sqrt(AVAILABLE_PROCESSORS); while (mod > 0 && matrixLength % mod != 0) { mod--; } return matrixLength / mod; } /** * 进行Cannon之前先进行数据对准 * * @param matrix1 * @param matrix2 */ private void dataAlignment(long[][] matrix1, long[][] matrix2, int subMatrixLength) { for (int i = 1; i < matrix1.length / subMatrixLength; i++) { loopLeftShiftByRow(matrix1, subMatrixLength, i, i); } for (int j = 1; j < matrix2[0].length / subMatrixLength; j++) { loopUpShiftByColumn(matrix2, subMatrixLength, j, j); } } @Override public long[][] computeByArray(final long[][] matrix1, final long[][] matrix2) throws InterruptedException { if (judgeMatrixLegal(matrix1, matrix2)) { return null; } long[][] result = new long[matrix1.length][matrix1[0].length]; int matrixLength = matrix1.length; final int subMatrixLength = getSubMatrixLength(matrixLength); //数据对准 dataAlignment(matrix1, matrix2, subMatrixLength); //计算点积 for (int m = 0; m < matrixLength / subMatrixLength; m++) { CountDownLatch countDownLatch = new CountDownLatch((matrixLength / subMatrixLength) * (matrixLength / subMatrixLength)); for (int i = 0; i < matrixLength / subMatrixLength; i++) { for (int j = 0; j < matrixLength / subMatrixLength; j++) { cannonThreadPool.execute(new ComputingUnitTask(matrix1, matrix2, result, i, j, m, countDownLatch, subMatrixLength)); } } countDownLatch.await(); } return result; } /** * 每个计算单元代表一个处理器线程 */ private class ComputingUnitTask implements Runnable { /** * 下面三个矩阵不产生写冲突 不需要加锁 */ private long[][] matrix1; private long[][] matrix2; private long[][] result; private int i; private int j; private int times; private CountDownLatch taskNumber; private int subMatrixLength; public ComputingUnitTask(long[][] matrix1, long[][] matrix2, long[][] result, int i, int j, int times, CountDownLatch taskNumber, int subMatrixLength) { this.matrix1 = matrix1; this.matrix2 = matrix2; this.result = result; this.i = i; this.j = j; this.taskNumber = taskNumber; this.subMatrixLength = subMatrixLength; this.times = times; } public void run() { int rowBeg = i * subMatrixLength; int columnBeg = j * subMatrixLength; int rowData = (i + times) % (matrix1.length / subMatrixLength) * subMatrixLength; int columnData = (j + times) % (matrix1.length / subMatrixLength) * subMatrixLength; for (int h = 0; h < subMatrixLength; h++) { for (int m = 0; m < subMatrixLength; m++) { for (int k = 0; k < subMatrixLength; k++) {