Most of us have computers on our desk that would have been considered supercomputers not long ago. We always wonder how many of them get any actual workout other than decoding video. If you want to simulate circuits you may very well start chewing up significant CPU time, so you might consider Xyce, an open source high-performance analog circuit simulator from Sandia National Labs. As you’d expect from a giant government lab it is able to support large scale parallel computing, but will also work on common desktop systems. On Linux, it will do what they call “small-scale parallelism.” In addition, it can deal with simulations of things as diverse as neural networks and power grids.
The code is open source, but oddly you do have to register to download it. Xyce has been around for a bit, but version 7.0 just arrived in April. Many of the changes are to improve compatibility with other Spice programs, notably HSpice.
The program is in C++ and not a derivative of Spice even though it tries to be compatible. It does, however, have a variety of features that support more efficient modeling and the use of some powerful solving engines and techniques.
Like traditional Spice, you’ll have to create netlists to use Xyce. Of course, for many circuits, you could probably create your netlist in another Spice tool and export the netlist. Why would you not just use the other tool? Xyce offers the potential to run very large simulations using multiple CPUs or clustered computers.
While it probably isn’t for everyone, Xyce might be just the excuse you need to build that Linux supercomputer you’ve always wanted. We imagine you’d need something more than ESP32s, though. We wonder if it will support CUDA or OpenCL.