Iterative Fibonacci

Compute fib(100k) mod 2^64 with hand-rolled integer-to-decimal. Tests integer arithmetic and tight-loop dispatch.

Runtime · median per inner-loop window

median of 7 runs

Native Gocompiled

114 µs0.04×

Pipitbytecode VM

3.00 msbaseline

CPython 3.14bytecode VM

18.8 ms6.25×

PyPy 7.3tracing JIT

10.3 ms3.43×

Ttengobytecode VM

no data

n/a

Sscriggobytecode VM

15.4 ms5.12×

Mmvmbytecode VM

19.0 ms6.31×

YyaegiAST walker

34.7 ms11.6×

Runner	N	Compile	Runtime	P95	Stddev	RSS	vs pipit	Status
Native Gocompiled	7	109 ms	114 µs	115 µs	1.19 µs	83 MiB	0.04×	OK
Pipitbytecode VM	7	505 µs	3.00 ms	4.14 ms	439 µs	58 MiB	1.00×	OK
CPython 3.14bytecode VM	7	246 µs	18.8 ms	19.2 ms	238 µs	n/a	6.25×	OK
PyPy 7.3tracing JIT	7	183 µs	10.3 ms	10.7 ms	382 µs	n/a	3.43×	OK
tengobytecode VM	0	n/a	n/a	n/a	n/a	n/a	n/a	no data
scriggobytecode VM	7	130 µs	15.4 ms	18.3 ms	1.02 ms	64 MiB	5.12×	OK
mvmbytecode VM	7	207 µs	19.0 ms	30.6 ms	4.11 ms	64 MiB	6.31×	OK
yaegiAST walker	7	228 µs	34.7 ms	43.8 ms	6.28 ms	64 MiB	11.6×	OK

Workload & symmetry rules

Workload

100,000 iterations of next = (a + b) & 0xFFFFFFFFFFFFFFFF, then hand-rolled int-to-decimal of the result.

Hand-rolled int-to-decimal in every runner (no strconv.FormatUint, no str(), no f"{n}").
Pure integer arithmetic; no math calls.
Explicit 64-bit mask in Python so it matches Go's uint64 wrap.

Every step of the work is interpreter-bound; halving dispatch overhead halves the wall time. There is no library call on the hot path.

Source code