Simple benchmarks to compare go-lua performance against the reference C implementation of Lua 5.3.
| Script | What it measures |
|---|---|
fib.lua |
Recursive Fibonacci — function call overhead |
loop.lua |
Tight arithmetic loop (10M iterations) — VM dispatch |
table.lua |
Array insert/read (1M), hash insert/iterate (500k) |
string.lua |
table.concat (100k), string.gmatch pattern matching |
sort.lua |
table.sort on 500k random integers |
cd benchmarks
go run . .Or build and run:
cd benchmarks
go build -o run-benchmarks .
./run-benchmarks .for f in benchmarks/*.lua; do
echo "--- $(basename $f) ---"
lua5.3 "$f"
echo
doneMeasured on Apple M4, macOS, Go 1.24, Lua 5.3.6.
| Benchmark | C-Lua 5.3 | go-lua | Factor |
|---|---|---|---|
| fib(35) | 0.42 s | 1.02 s | ~2.4x |
| loop 10M | 0.05 s | 0.39 s | ~8x |
| table | 0.22 s | 0.63 s | ~3x |
| string | 0.02 s | 0.05 s | ~2.5x |
| sort 500k | 0.11 s | 0.57 s | ~5x |
go-lua is roughly 2-8x slower than C-Lua 5.3, which is expected for a pure Go implementation. The overhead comes mainly from Go interface dispatch, bounds checking, and garbage collection differences.
