try to use:
LinearAlgebra.BLAS.gemm!('T', 'N', one(T), view(mx, block, :), A, one(T), B)
where block
- Vector{Int}
and have:
ERROR: ArgumentError: strides is invalid for SubArrays with indices of type Vector{Int64}
can I use views with LinearAlgebra.BLAS.gemm!?
Consider using the 5 argument mul!
also from LinearAlgebra
1 Like
The underlying BLAS API doesn’t support views with an arbitrary vector of indices.
(You can use mul!
, but realize that you will pay a performance price compared to BLAS multiplies because it is falling back to the generic multiplication routine.)
1 Like
Thank you! mul!
works fine, but I didn’t make a deep benchmark