New version of SpeedIT eXtreme introduces computations and memory management on complex numbers in single and double precision. Complex numbers are represented as new types si_scomplex for single precision and si_dcomplex for double precision. They are compatible with float2 and double2 types from CUDA vector types. For memory management Speedit Extreme allows you to:

• Allocate memory space for complex number's arrays on CPU and GPU memory space.
• Copy arrays with complex numbers from CPU and GPU and back.
• Free allocated memory space for complex numbers arrays from CPU and GPU memory space.

For linear algebra systems new functionality allows you to:
• Perform a fast sparse matrix vector multiplications for complex numbers.
• Solve linear equation system with BICGStab and CG solvers for complex numbers.