Repository navigation
Benchmarks and CI
0x4171341 edited this page Jul 24, 2026
·
2 revisions
The primary workflow is .github/workflows/ci.yml.
It validates:
- king20 smoke benchmarks
- scalability phase analysis
- ownership and parity tests
- Windows COFF / PE direct-link path
- macOS host-link path
- module and stdlib checks
The Benchmark Package Workflow is .github/workflows/bpw.yml.
It runs a multi-platform comparison across:
- L++
- Rust
- Go
Benchmarks include:
- CPU-heavy:
fib(40)plus prime counting - RAM-heavy: large list allocation/fill
- File I/O: write/read workload
- Multi-file import project
- Generated large-project compile benchmark
Each benchmark records:
- average time
- min/max
- standard deviation
- binary size
- system information
- compiler versions
- output correctness
Very small millisecond timings are noisy. BPW runs multiple iterations and reports aggregate data, but benchmark numbers should still be read as directional rather than absolute.
The strongest observed L++ advantages are:
- very small direct-linked binaries
- fast compile times
- competitive native runtime on simple workloads
The BPW workflow has separate platform jobs and a merge job:
bpw-linux
bpw-windows
bpw-merge
The merge job downloads platform artifacts, combines JSON reports, prints a cross-platform table, and uploads a merged report artifact.
Typical artifact contents:
- source benchmark files
- per-platform
bpw_report.json - merged
bpw_merged.json - selected binaries where workflow configuration preserves them