MatrixMatrixMult
org.ejml.alg.dense.mult

Class MatrixMatrixMult



  • public class MatrixMatrixMultextends Object

    This class contains various types of matrix matrix multiplication operations for RowD1Matrix64F.

    Two algorithms that are equivalent can often have very different runtime performance. This is because of how modern computers uses fast memory caches to speed up reading/writing to data. Depending on the order in which variables are processed different algorithms can run much faster than others, even if the number of operations is the same.

    Algorithms that are labeled as 'reorder' are designed to avoid caching jumping issues, some times at the cost of increasing the number of operations. This is important for large matrices. The straight forward implementation seems to be faster for small matrices.

    Algorithms that are labeled as 'aux' use an auxiliary array of length n. This array is used to create a copy of an out of sequence column vector that is referenced several times. This reduces the number of cache misses. If the 'aux' parameter passed in is null then the array is declared internally.

    Typically the straight forward implementation runs about 30% faster on smaller matrices and about 5 times slower on larger matrices. This is all computer architecture and matrix shape/size specific.

    ******** IMPORTANT **********
    This class was auto generated using GeneratorMatrixMatrixMult If this code needs to be modified, please modify GeneratorMatrixMatrixMult instead and regenerate the code by running that.

SCaVis 2.2 © jWork.ORG