I am using cusparseDenseToSparse subroutines to convert Dense matrix to Sparse Matrix. I found an issue when the data type was set to CUDA_C_64F. The number of non zero value is not correct during conversion.
e.g. 77 no zero values in real part and 81 no zero value in image part. The number of non zero value returned from subroutine cusparseSpMatGetSize is 77. Not 81.
Is anyone aware of this issue?