onsdagens / esp32c3-rt-benchmarks

Simple benchmark showcasing performance of direct vectoring against regular interrupt handling via the HAL

esp32-c3 runtime benchmarks

The goal of this crate is highlighting the gains of hooking the interrupt vector table directly over using the generic HAL handler.

Results

The amount of cycles taken between the triggering of an interrupt and handler entry was measured to 473 cycles for the direct option, and 3492 cycles for the default HAL option, an 86% decrease in latency.

Run it yourself

NOTE: The benchmarks are intended for the ESP32-C3 microcontroller, and tested using the ESP32-C3-DevKit-RUST-1. The idea is compatible with the other Espressif RISC-V MCUs, but the benchmarks themselves may require some adjustments to compile and flash.

cargo-embed is required for flashing. We use the imc target.

rustup target add riscv32imc-unknown-none-elf
cargo install cargo-embed

cargo embed --example benchmark-hal --release

To run the HAL driven example.

cargo embed --example benchmark-direct --features direct --release

To run the direct example.