Today's computers typically come with CPUs having more than one core. It will thus be useful to enhance ngspice to make use of such multi-core processors.

Using circuits comprising mostly of transistors and e.g. the BSIM3 model, around 2/3 of the CPU time is spent in evaluating the model equations (e.g. in the BSIM3Load() function). The same happens with other advanced transistor models. Thus this function should be paralleled, if possible. Resulting from that the parallel processing has to be within a dedicated device model. Solving the matrix takes about 10% to 50% of the CPU time, so paralleling the matrix solver is sometimes of secondary interest only! And it is difficult to achive with our Sparse Matrix und KLU solvers.

Another alternative is using CUSPICE, that is ngspice (current version

27. designed for running massively parallel on NVIDIA GPUs. CUDA enhancements to C code are applied. For LINUX, please see the user guide. For MS Windows, an executable is available at the ngspice download pages.