Package org.apache.sysds.runtime.matrix.data

Class LibMatrixMult

java.lang.Object
org.apache.sysds.runtime.matrix.data.LibMatrixMult
public class LibMatrixMult extends Object
MB: Library for matrix multiplications including MM, MV, VV for all combinations of dense, sparse, ultrasparse representations and special operations such as transpose-self matrix multiplication.

In general all implementations use internally dense outputs for direct access, but change the final result to sparse if necessary. The only exceptions are ultra-sparse matrix mult, wsloss and wsigmoid.

  • Field Details

  • Method Details

    • matrixMult

      public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2)
      Performs a matrix multiplication All variants use a IKJ access pattern, and internally use dense output. After the actual computation, we recompute nnz and check for sparse/dense representation.
      Parameters:
      m1 - first matrix
      m2 - second matrix
      Returns:
      ret Matrix Block

    • matrixMult

      public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, int k)
      Performs a matrix multiplication All variants use a IKJ access pattern, and internally use dense output. After the actual computation, we recompute nnz and check for sparse/dense representation.
      Parameters:
      m1 - first matrix
      m2 - second matrix
      k - maximum parallelism
      Returns:
      ret Matrix Block

    • matrixMult

      public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret)
      Performs a matrix multiplication and stores the result in the output matrix. All variants use a IKJ access pattern, and internally use dense output. After the actual computation, we recompute nnz and check for sparse/dense representation.
      Parameters:
      m1 - first matrix
      m2 - second matrix
      ret - result matrix
      Returns:
      ret Matrix Block

    • matrixMult

      public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, boolean fixedRet)
      This method allows one to disabling exam sparsity. This feature is useful if matrixMult is used as an intermediate operation (for example: LibMatrixDNN). It makes sense for LibMatrixDNN because the output is internally consumed by another dense instruction, which makes repeated conversion to sparse wasteful. This should be used in rare cases and if you are unsure, use the method 'matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret)' instead.
      Parameters:
      m1 - first matrix
      m2 - second matrix
      ret - result matrix
      fixedRet - if true, output representation is fixed and nnzs not recomputed
      Returns:
      ret Matrix Block

    • matrixMult

      public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, int k)
      Performs a multi-threaded matrix multiplication and stores the result in the output matrix. The parameter k (k>=1) determines the max parallelism k' with k'=min(k, vcores, m1.rlen).
      Parameters:
      m1 - first matrix
      m2 - second matrix
      ret - result matrix
      k - maximum parallelism
      Returns:
      ret Matrix Block

    • matrixMultNonNative

      public static MatrixBlock matrixMultNonNative(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, int k)

    • matrixMult

      public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, boolean fixedRet, int k)
      Performs a matrix multiplication and stores the result in the output matrix. All variants use a IKJ access pattern, and internally use dense output. After the actual computation, we recompute nnz and check for sparse/dense representation. This method allows one to disabling exam sparsity. This feature is useful if matrixMult is used as an intermediate operation (for example: LibMatrixDNN). It makes sense for LibMatrixDNN because the output is internally consumed by another dense instruction, which makes repeated conversion to sparse wasteful. This should be used in rare cases and if you are unsure, use the method 'matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret)' instead. The parameter k (k>=1) determines the max parallelism k' with k'=min(k, vcores, m1.rlen).
      Parameters:
      m1 - first matrix
      m2 - second matrix
      ret - result matrix
      fixedRet - if true, output representation is fixed and nnzs not recomputed
      k - maximum parallelism
      Returns:
      ret Matrix Block

    • emptyMatrixMult

      public static MatrixBlock emptyMatrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret)

    • matrixMultChain

      public static void matrixMultChain(MatrixBlock mX, MatrixBlock mV, MatrixBlock mW, MatrixBlock ret, MapMultChain.ChainType ct)
      Performs a matrix multiplication chain operation of type t(X)%*%(X%*%v) or t(X)%*%(w*(X%*%v)). All variants use a IKJ access pattern, and internally use dense output. After the actual computation, we recompute nnz and check for sparse/dense representation.
      Parameters:
      mX - X matrix
      mV - v matrix
      mW - w matrix
      ret - result matrix
      ct - chain type

    • matrixMultChain

      public static void matrixMultChain(MatrixBlock mX, MatrixBlock mV, MatrixBlock mW, MatrixBlock ret, MapMultChain.ChainType ct, int k)
      Performs a parallel matrix multiplication chain operation of type t(X)%*%(X%*%v) or t(X)%*%(w*(X%*%v)). The parameter k (k>=1) determines the max parallelism k' with k'=min(k, vcores, m1.rlen). NOTE: This multi-threaded mmchain operation has additional memory requirements of k*ncol(X)*8bytes for partial aggregation. Current max memory: 256KB; otherwise redirect to sequential execution.
      Parameters:
      mX - X matrix
      mV - v matrix
      mW - w matrix
      ret - result matrix
      ct - chain type
      k - maximum parallelism

    • matrixMultTransposeSelf

      public static MatrixBlock matrixMultTransposeSelf(MatrixBlock m1, MatrixBlock ret, boolean leftTranspose)

    • matrixMultTransposeSelf

      public static void matrixMultTransposeSelf(MatrixBlock m1, MatrixBlock ret, boolean leftTranspose, boolean copyToLowerTriangle)

    • matrixMultTransposeSelf

      public static void matrixMultTransposeSelf(MatrixBlock m1, MatrixBlock ret, boolean leftTranspose, int k)
      TSMM with optional transposed left side or not (Transposed self matrix multiplication)
      Parameters:
      m1 - The matrix to do tsmm
      ret - The output matrix to allocate the result to
      leftTranspose - If the left side should be considered transposed
      k - the number of threads to use

    • matrixMultPermute

      public static void matrixMultPermute(MatrixBlock pm1, MatrixBlock m2, MatrixBlock ret1, MatrixBlock ret2)

    • matrixMultPermute

      public static void matrixMultPermute(MatrixBlock pm1, MatrixBlock m2, MatrixBlock ret1, MatrixBlock ret2, int k)

    • matrixMultWSLoss

      public static void matrixMultWSLoss(MatrixBlock mX, MatrixBlock mU, MatrixBlock mV, MatrixBlock mW, MatrixBlock ret, WeightedSquaredLoss.WeightsType wt)

    • matrixMultWSLoss

      public static void matrixMultWSLoss(MatrixBlock mX, MatrixBlock mU, MatrixBlock mV, MatrixBlock mW, MatrixBlock ret, WeightedSquaredLoss.WeightsType wt, int k)

    • matrixMultWSigmoid

      public static void matrixMultWSigmoid(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, MatrixBlock ret, WeightedSigmoid.WSigmoidType wt)

    • matrixMultWSigmoid

      public static void matrixMultWSigmoid(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, MatrixBlock ret, WeightedSigmoid.WSigmoidType wt, int k)

    • matrixMultWDivMM

      public static void matrixMultWDivMM(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, MatrixBlock mX, MatrixBlock ret, WeightedDivMM.WDivMMType wt)
      NOTE: This operation has limited NaN support, which is acceptable because all our sparse-safe operations have only limited NaN support. If this is not intended behavior, please disable the rewrite. In detail, this operator will produce for W/(U%*%t(V)) a zero intermediate for each zero in W (even if UVij is zero which would give 0/0=NaN) but INF/-INF for non-zero entries in V where the corresponding cell in (Y%*%X) is zero.
      Parameters:
      mW - matrix W
      mU - matrix U
      mV - matrix V
      mX - matrix X
      ret - result type
      wt - weighted divide matrix multiplication type

    • matrixMultWDivMM

      public static void matrixMultWDivMM(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, MatrixBlock mX, MatrixBlock ret, WeightedDivMM.WDivMMType wt, int k)
      NOTE: This operation has limited NaN support, which is acceptable because all our sparse-safe operations have only limited NaN support. If this is not intended behavior, please disable the rewrite. In detail, this operator will produce for W/(U%*%t(V)) a zero intermediate for each zero in W (even if UVij is zero which would give 0/0=NaN) but INF/-INF for non-zero entries in V where the corresponding cell in (Y%*%X) is zero.
      Parameters:
      mW - matrix W
      mU - matrix U
      mV - matrix V
      mX - matrix X
      ret - result matrix
      wt - weighted divide matrix multiplication type
      k - maximum parallelism

    • matrixMultWCeMM

      public static void matrixMultWCeMM(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, double eps, MatrixBlock ret, WeightedCrossEntropy.WCeMMType wt)

    • matrixMultWCeMM

      public static void matrixMultWCeMM(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, double eps, MatrixBlock ret, WeightedCrossEntropy.WCeMMType wt, int k)

    • matrixMultWuMM

      public static void matrixMultWuMM(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, MatrixBlock ret, WeightedUnaryMM.WUMMType wt, ValueFunction fn)

    • matrixMultWuMM

      public static void matrixMultWuMM(MatrixBlock mW, MatrixBlock mU, MatrixBlock mV, MatrixBlock ret, WeightedUnaryMM.WUMMType wt, ValueFunction fn, int k)

    • matrixMultDenseDenseMM

      public static void matrixMultDenseDenseMM(DenseBlock a, DenseBlock b, DenseBlock c, boolean transA, boolean transB, int n, int cd, int rl, int ru, int cl, int cu)

    • matrixMultDenseDenseMMDedup

      public static void matrixMultDenseDenseMMDedup(DenseBlockFP64DEDUP a, DenseBlock b, DenseBlockFP64DEDUP c, int n, int cd, int rl, int ru, ConcurrentHashMap<double[],double[]> cache)

    • matrixMultDenseDenseMM

      public static void matrixMultDenseDenseMM(DenseBlock a, DenseBlock b, DenseBlock c, int n, int cd, int rl, int ru, int cl, int cu)

    • dotProduct

      public static double dotProduct(double[] a, double[] b, int ai, int bi, int len)

    • dotProduct

      public static double dotProduct(double[] a, double[] b, int[] aix, int ai, int bi, int len)

    • vectMultiplyAdd

      public static void vectMultiplyAdd(double aval, double[] b, double[] c, int bi, int ci, int len)

    • vectMultiplyAdd

      public static void vectMultiplyAdd(double aval, double[] b, double[] c, int[] bix, int bi, int ci, int len)

    • vectMultiplyWrite

      public static void vectMultiplyWrite(double aval, double[] b, double[] c, int bi, int ci, int len)

    • vectMultiplyInPlace

      public static void vectMultiplyInPlace(double aval, double[] c, int ci, int len)

    • vectMultiplyInPlace

      public static void vectMultiplyInPlace(double[] a, double[] c, int[] cix, int ai, int ci, int len)

    • vectMultiplyWrite

      public static void vectMultiplyWrite(double[] a, double[] b, double[] c, int ai, int bi, int ci, int len)

    • vectMultiplyAdd

      public static void vectMultiplyAdd(double[] a, double[] b, double[] c, int ai, int bi, int ci, int len)

    • vectMultiplyWrite

      public static void vectMultiplyWrite(double[] a, double[] b, double[] c, int[] bix, int ai, int bi, int ci, int len)

    • vectMult2Write

      public static double[] vectMult2Write(double[] a, double[] c, int ai, int len)

    • vectMult2WriteDedicated2

      public static double[] vectMult2WriteDedicated2(double[] a, double[] c, int ai, int len)

    • vectMultiply

      public static void vectMultiply(double[] a, double[] c, int ai, int ci, int len)

    • vectAdd

      public static void vectAdd(double[] a, double bval, double[] c, int ai, int ci, int len)

    • vectAdd

      public static void vectAdd(double[] a, double[] c, int ai, int ci, int len)

    • vectAdd

      public static void vectAdd(double[] a, double[] c, int[] aix, int ai, int ci, int alen)

    • vectAddInPlace

      public static void vectAddInPlace(double aval, double[] c, int ci, int len)

    • copyUpperToLowerTriangle

      public static long copyUpperToLowerTriangle(MatrixBlock ret)

    • copyUpperToLowerTriangleDense

      public static long copyUpperToLowerTriangleDense(MatrixBlock ret)
      Used for all version of TSMM where the result is known to be symmetric. Hence, we compute only the upper triangular matrix and copy this partial result down to lower triangular matrix once.
      Parameters:
      ret - matrix
      Returns:
      number of non zeros

    • copyUpperToLowerTriangleSparse

      public static long copyUpperToLowerTriangleSparse(MatrixBlock ret)

    • prepMatrixMultTransposeSelfInput

      public static MatrixBlock prepMatrixMultTransposeSelfInput(MatrixBlock m1, boolean leftTranspose, boolean par)

    • isSkinnyRightHandSide

      public static boolean isSkinnyRightHandSide(long m1rlen, long m1clen, long m2rlen, long m2clen, boolean inclCacheSize)

    • satisfiesMultiThreadingConstraints

      public static boolean satisfiesMultiThreadingConstraints(MatrixBlock m1, int k)

    • satisfiesMultiThreadingConstraints

      public static boolean satisfiesMultiThreadingConstraints(MatrixBlock m1, boolean checkMem, boolean checkFLOPs, long FPfactor, int k)

    • satisfiesMultiThreadingConstraints

      public static boolean satisfiesMultiThreadingConstraints(MatrixBlock m1, MatrixBlock m2, boolean checkMem, boolean checkFLOPs, long FPfactor, int k)

    • isUltraSparseMatrixMult

      public static boolean isUltraSparseMatrixMult(MatrixBlock m1, MatrixBlock m2, boolean m1Perm)

    • isSparseOutputMatrixMult

      public static boolean isSparseOutputMatrixMult(MatrixBlock m1, MatrixBlock m2)

    • isSparseOutputTSMM

      public static boolean isSparseOutputTSMM(MatrixBlock m1)

    • isSparseOutputTSMM

      public static boolean isSparseOutputTSMM(MatrixBlock m1, boolean ultraSparse)

    • isOuterProductTSMM

      public static boolean isOuterProductTSMM(int rlen, int clen, boolean left)