Backends and performance

Choose a backend that supports the model, then size the calculation and tune its execution. Backend choice changes where and how the calculation runs; the physical and numerical settings still determine what it computes. Installation begins in the introduction. These chapters cover the native backend, memory, CPU threads, and GPU execution. The best speed on a given machine comes from calibrating solprop to it, once, with solprop calibrate.

To measure a change’s performance, or to see roughly what each backend and method costs, read the developer pages: Benchmarks gives the measurement rules and indicative results, and Running the benchmarks is the manual for uv run poe bench.

• Calibrating solprop to your machine
• Choosing a backend
• The native backend: fast CPUs and GPUs
• Building the native backend
• Execution resources
• Progress and logging
• Performance tuning
• Single precision