cuda - How to perform Hadamard product with CUBLAS on complex numbers? -

i need compute element wise multiplication of 2 vectors (hadamard product) of complex numbers nvidia cublas. unfortunately, there no had operation in cublas. apparently, can sbmv operation, not implemented complex numbers in cublas. cannot believe there no way achieve cublas. there other way achieve cublas, complex numbers ?

i cannot write own kernel, have use cublas (or standard nvidia library if not possible cublas).

cublas based on reference blas, , reference blas has never contained hadamard product (complex or real). hence cublas doesn't have 1 either. intel have added v?mul mkl doing this, non-standard , not in blas implementations. kind of operation old school fortran programmer write loop for, presume didn't warrant dedicated routine in blas.

there no "standard" cuda library aware of implements hadamard product. there possibility of using cublas gemm or symm , extracting diagonal of resulting matrix, horribly inefficient, both computation , storage stand point.

the thrust template library can trivially using thrust::transform, example:

thrust::multiplies<thrust::complex<float> > op; thrust::transform(thrust::device, x, x + n, y, z, op); 

would iterate on each pair of inputs device pointers x , y , calculate z[i] = x[i] * y[i] (there couple of casts need make compile that, idea). requires compilation of cuda code within project, , apparently don't want that.