better benchmarks

Author: mkrd

.github/copilot-instructions.md

@@ -0,0 +1,32 @@
+# Swift-BigInt
+
+Swift 4.2 big number library (`BInt`, `BDouble`) with no external dependencies.
+
+## Build & Test
+
+```sh
+swift build                # debug build
+swift test                 # run all tests
+just test-verify           # release-mode correctness checks
+just test-full             # same as swift test
+```
+
+## Benchmarks
+
+```sh
+just bench                 # run benchmarks, print results
+just bench-save            # run benchmarks, save baseline to performance_baseline.json
+just bench-compare         # run benchmarks, compare against saved baseline
+just bench 3               # run each benchmark 3 times, show per-run + average
+just bench-save 3           # save baseline averaged over 3 runs
+just bench-compare 3        # compare averages over 3 runs against baseline
+```
+
+The `Benchmarks` binary accepts subcommands directly:
+- `Benchmarks [runs]` — print results
+- `Benchmarks save <file> [runs]` — save results as JSON
+- `Benchmarks compare <file> [runs]` — compare current run against JSON baseline
+
+Benchmarks are defined declaratively in `Sources/Benchmarks/Benchmarks.swift` as a `[BenchmarkEntry]` array. Each entry has a name and a `setup` closure that returns the timed `body` closure. Setup work (allocations, precomputation) is not included in the measurement.
+
+To add a benchmark, append a `BenchmarkEntry` to `allBenchmarks`. Then run `just bench-save` to update the baseline.