I quick question regarding pardiso/pardiso_64: When iparm(5)=1 (0-based indexing) so that the solution is put in the rhs vector b, do I need to allocate any memory for the argument 'x' (the solution vector) ? I would think not, but the documentation is not specific (https://software.intel.com/en-us/node/470284).
I'm using mkl 11.3 update 3.
Ok, thanks! However, according to https://software.intel.com/en-us/node/470298#IPARM6, "The array x is always used." What does that mean?
also, if I just pass a std:vector<double>().data() as the 'b' parameter, pardiso crashes.
I'm afraid I misled you earlier, sorry for any inconvenience.
The point is x-array is used anyway on solving step of Pardiso (this is what the Note is supposed to be about), so it should be allocated and iparm is just indicates the storage for the solution.
About the crash: can you please provide a short reproducer for your case?
Thank you for your questions!