performance

Performance

hyprcross adds SDK and runner configuration around the selected builder. A warm build with no work left is therefore a useful measurement of its fixed overhead.

Fully cached builds

The workload is the hyprcross workspace, using release builds of hyprcross and Cargo. Each row has one checked warm-up followed by 15 checked samples, run sequentially and offline.

Command	Median	Fastest
cargo build --target x86_64-unknown-linux-gnu	60.469 ms	60.000 ms
hyprcross build --builder cargo -t x86_64-unknown-linux-gnu	61.664 ms	60.408 ms
hyprcross build --builder cargo -t aarch64-unknown-linux-gnu	62.088 ms	61.245 ms
hyprcross --version	0.765 ms	0.712 ms

Environment: Intel Core i5-10400F, Linux, rustc 1.98.1. The AArch64 case uses the installed direct SDK. The measured fixed overhead is about 1.2 ms for the host target and 1.6 ms for the cross target. These numbers describe a warm no-op build; compile and link time dominates real builds.

Reproduce the measurement after release builds with:

python3 benches/bench_runner.py \
  --hyprcross target/release/hyprcross \
  --igl ../target/release/igl --runs 15

Pass --builder igl to measure Igel when the workspace lockfile can be resolved by Igel. Raw samples and version details are in benches/benchmark_results.json.