Smoke: integer addition

Sum 1..N in a tight loop, the smallest workload in the suite. It exists to surface interpreter dispatch overhead at the smallest possible payload.

Runtime Compile time

Runtime · median per inner-loop window

median of 10 runs

Native Gocompiled

24.9 µs0.06×

Piko interpbytecode VM

444 µsbaseline

CPython 3.13bytecode VM

1.53 ms3.44×

PyPy 7.3tracing JIT

481 µs1.08×

Ttengobytecode VM

5.06 ms11.4×

Sscriggobytecode VM

4.48 ms10.1×

Mmvmbytecode VM

2.81 ms6.33×

YyaegiAST walker

2.26 ms5.09×

Full statistics

Runner	N	Compile	Runtime	P95	Stddev	RSS	vs piko	Status
Native Gocompiled	10	178 ms	24.9 µs	28.3 µs	1.01 µs	68 MiB	0.06×	OK
Piko interpbytecode VM	10	452 µs	444 µs	449 µs	2.77 µs	78 MiB	1.00×	OK
CPython 3.13bytecode VM	10	185 µs	1.53 ms	1.84 ms	116 µs	n/a	3.44×	OK
PyPy 7.3tracing JIT	10	166 µs	481 µs	510 µs	13.9 µs	n/a	1.08×	OK
tengobytecode VM	10	140 µs	5.06 ms	5.66 ms	341 µs	84 MiB	11.4×	OK
scriggobytecode VM	10	131 µs	4.48 ms	4.56 ms	537 µs	70 MiB	10.1×	OK
mvmbytecode VM	10	138 µs	2.81 ms	3.43 ms	511 µs	57 MiB	6.33×	OK
yaegiAST walker	10	193 µs	2.26 ms	2.28 ms	6.01 µs	58 MiB	5.09×	OK

Workload & symmetry rules

Workload

Sum 1..1,000,000 in a single tight loop, print the result mod 2^32.

Why this benchmark exists

It's the simplest possible interpreter workload. If dispatch dominates anything, it dominates this. End-to-end mode is therefore mostly startup time; inner-loop mode is mostly add+jump throughput.

Symmetry rules

One loop, two locals, no function calls in the hot path.
No sum() / numpy / strconv shortcuts.

Source code

piko / Go
piko_source.go
native Go
native_main.go
CPython / PyPy
cpython.py
tengo
script.tengo