Generic numeric pipeline

Generic Sum / Max / Min over 500 slices each of uint32 and int64. Exercises type-parameterised dispatch.

Runtime Compile time

Compile time · median (cold)

median of 7 runs

Native Gocompiled

108 ms65.4×

Pipitbytecode VM

1.65 msbaseline

CPython 3.14bytecode VM

370 µs0.22×

PyPy 7.3tracing JIT

307 µs0.19×

Ttengobytecode VM

240 µs0.15×

Sscriggobytecode VM

unsupported in scriggo

n/a

Mmvmbytecode VM

402 µs0.24×

YyaegiAST walker

453 µs0.28×

Full statistics

Runner	N	Compile	Runtime	P95	Stddev	RSS	vs pipit	Status
Native Gocompiled	7	108 ms	696 µs	700 µs	1.96 µs	1.62 GiB	65.4×	OK
Pipitbytecode VM	7	1.65 ms	17.7 ms	18.0 ms	127 µs	592 MiB	1.00×	OK
CPython 3.14bytecode VM	7	370 µs	32.5 ms	32.6 ms	171 µs	n/a	0.22×	OK
PyPy 7.3tracing JIT	7	307 µs	10.2 ms	11.1 ms	584 µs	n/a	0.19×	OK
tengobytecode VM	7	240 µs	70.9 ms	86.2 ms	6.75 ms	1.69 GiB	0.15×	OK
scriggobytecode VM	0	n/a	n/a	n/a	n/a	n/a	n/a	unsupported
mvmbytecode VM	7	402 µs	40.0 ms	50.2 ms	3.66 ms	1.69 GiB	0.24×	OK
yaegiAST walker	7	453 µs	42.0 ms	57.8 ms	6.94 ms	1.69 GiB	0.28×	OK

Workload & symmetry rules

Workload

Generate 500 slices of 100 uint32 values via LCG; apply GenericSum, GenericMax, GenericMin to each, XOR-folding results. Repeat with 500 slices of 100 int64 values, the same generic functions in a second instantiation. Total 6,000 generic dispatches per outer call, each iterating 100 elements.

Symmetry rules

Go uses real [T Numeric] generics; Python uses the equivalent duck-typed function (it has no separate generic-dispatch primitive, so the comparison is what Python's runtime cost looks like on the same workload).

Why this benchmark exists

Distinct from interface dispatch (bench 13) and closure dispatch (bench 17). Measures whether pipit's generics path has competitive specialisation.

Source code

pipit / Go
pipit_source.go
native Go
native_main.go
CPython / PyPy
cpython.py
tengo
script.tengo