Why is B = numpy.dot(A,x) so much slower looping through doing B[i,:,:] = numpy.dot(A[i,:,:],x) )?

by

numpy.dot only delegates to a BLAS matrix multiply when the inputs each have dimension at most 2:

#if defined(HAVE_CBLAS)
    if (PyArray_NDIM(ap1) <= 2 && PyArray_NDIM(ap2) <= 2 &&
            (NPY_DOUBLE == typenum || NPY_CDOUBLE == typenum ||
             NPY_FLOAT == typenum || NPY_CFLOAT == typenum)) {
        return cblas_matrixproduct(typenum, ap1, ap2, out);
    }
#endif

When you stick your whole 3-dimensional A array into dot, NumPy takes a slower path, going through an nditer object. It still tries to get some use out of BLAS in the slow path, but the way the slow path is designed, it can only use a vector-vector multiply rather than a matrix-matrix multiply, which doesn’t give the BLAS anywhere near as much room to optimize.

More Related Contents:

  1. Using numpy to build an array of all combinations of two arrays
  2. Difference between numpy.array shape (R, 1) and (R,)
  3. Understanding NumPy’s einsum
  4. Intuition and idea behind reshaping 4D array to 2D array in NumPy
  5. How does numpy.newaxis work and when to use it?
  6. What is the difference between flatten and ravel functions in numpy?
  7. Why is numpy’s einsum faster than numpy’s built in functions?
  8. Calculate mean across dimension in a 2D array
  9. How to count the occurrence of certain item in an ndarray?
  10. Concatenate Numpy arrays without copying
  11. How do I count the occurrence of a certain item in an ndarray?
  12. Convert a 1D array to a 2D array in numpy
  13. Subsetting a 2D numpy array
  14. Selecting Random Windows from Multidimensional Numpy Array Rows
  15. How to split a numpy array in fixed size chunks with and without overlap?
  16. How to find the index of a value in 2d array in Python?
  17. What is the difference between ndarray and array in NumPy?
  18. Why using an array as an index changes the shape of a multidimensional ndarray?
  19. 3d Numpy array to 2d
  20. Vectorized spatial distance in python using numpy
  21. Axes from plt.subplots() is a “numpy.ndarray” object and has no attribute “plot”
  22. Python: how to store a numpy multidimensional array in PyTables?
  23. Access n-th dimension in python [duplicate]
  24. Is there a numpy/scipy dot product, calculating only the diagonal entries of the result?
  25. How does the axis parameter from NumPy work?
  26. Indexing numpy array with another numpy array
  27. How can I turn a flat list into a 2D array in python?
  28. Get the indices of N highest values in an ndarray
  29. Mapping a NumPy array in place
  30. How to convert a 3d numpy array to 2d

Leave a Comment