Matrix scale() function will not run

[Benjamin Beach]

Hello Blaze team,

I would like to use the built-in scale() function to perform elementwise matrix-matrix or vector-vector multiplication, but when I try to run:

  blaze::StaticMatrix<dfloat, 2, 2> M1();
  M1=1;
  blaze::StaticMatrix<dfloat, 2, 2> M2(M1);
  M2.scale(M1);

I get the following error:

[path2blaze]/blaze/include/blaze/math/dense/StaticMatrix.h:1960:21 error: invalid operands of types ‘double’ and ‘blaze::StaticMatrix<double, 2ul, 2ul>(’ to binary ‘operator’ v_[iNN+j] = scalar; ~~~^~~~~ [path2blaze]/blaze/include/blaze/math/dense/StaticMatrix.h:1960:21 error: in evaluation of ‘operator=(double, class blaze::StaticMatrix<double, 2ul, 2ul>())’

The same thing occurs when I attempt to use dynamic or sparse matrices, or even dynamic vectors. I can get around this with custom functions, but it would be extremely useful to simply be able to call this builtin function.

Best, Ben Beach

[Benjamin Beach]

Update: When I said run, I really meant compile; the errors are compiler errors.

[Klaus Iglberger]

Hi Benjamin!

We have to apologize for the apparently misleading or incomplete documentation of the scale() function. scale() does not perform a componentwise vector-vector or matrix-matrix multiplication, but applies the given scalar element to all elements of the vector or matrix:

DynamicMatrix<int> A;

A *= 5;        // Scaling matrix A by 5; only works for fundamental data types and complex numbers
A.scale( 5 );  // Same effect as the previous function call; works for all possible data types

In contrast to the multiplication assignment operator, which only works for fundamental data types and std::complex, the scale() function works for all possible data types. This might for instance be necessary for block vectors and matrices.

The componentwise multiplication for vectors is available via operator*():

DynamicVector<int> a, b, c;

c = a * b;  // Element-wise multiplication

The componentwise matrix multiplication (i.e. the Hadamard product or Schur product) is one of the big new features of Blaze 3.2 and will be available shortly (see issue #60).

Again, we are sorry for the confusing and will try to update the documentation of the scale() function accordingly.

Best regards,

Klaus!