Based on https://discourse.julialang.org/t/uniform-scaling-inplace-addition-with-matrix/59928 some improvements can be done in the julia code gen:

• add a function inplace_scale_add!(a,A,B::UniformScaling) which computes A=a*A+B where B=I is UniformScaling with for-loop as in discourse
• add a function inplace_scale_add!(a,A,B::Matrix) which computes A=a*A+B, essentially replacing BLAS.axpby! calls with dot operation. Direct BLAS-calls should be avoided in julia.
• Replace codes like this with a copy-call and subsequently a call to inplace_scale_add!

    memslots[6][:]=coeff1*memslots[2] + coeff2*memslots[3]

The functions should maybe take into account if a coefficient is one since it can be made more efficient. Maybe we need more functions, e.g. just inplace_add and inplace_scale_add_scale! and inplace_add_scale!.