Open-addressing hash map

Hand-rolled FNV-1a hash table, 100k mixed put/get/delete operations. No built-in map allowed.

Runtime Compile time

Runtime · median per inner-loop window

median of 7 runs

Native Gocompiled

22.6 ms0.08×

Pipitbytecode VM

292 msbaseline

CPython 3.14bytecode VM

659 ms2.25×

PyPy 7.3tracing JIT

94.5 ms0.32×

Ttengobytecode VM

1.60 s5.46×

Sscriggobytecode VM

unsupported in scriggo

n/a

Mmvmbytecode VM

764 ms2.61×

YyaegiAST walker

1.47 s5.01×

Full statistics

Runner	N	Compile	Runtime	P95	Stddev	RSS	vs pipit	Status
Native Gocompiled	7	108 ms	22.6 ms	24.1 ms	553 µs	1.62 GiB	0.08×	OK
Pipitbytecode VM	7	2.44 ms	292 ms	308 ms	6.38 ms	591 MiB	1.00×	OK
CPython 3.14bytecode VM	7	769 µs	659 ms	660 ms	3.23 ms	n/a	2.25×	OK
PyPy 7.3tracing JIT	7	544 µs	94.5 ms	110 ms	7.14 ms	n/a	0.32×	OK
tengobytecode VM	7	408 µs	1.60 s	1.62 s	28.8 ms	1.69 GiB	5.46×	OK
scriggobytecode VM	0	n/a	n/a	n/a	n/a	n/a	n/a	unsupported
mvmbytecode VM	7	738 µs	764 ms	774 ms	4.74 ms	1.69 GiB	2.61×	OK
yaegiAST walker	7	628 µs	1.47 s	1.51 s	31.9 ms	1.69 GiB	5.01×	OK

Workload & symmetry rules

Workload

Build a HashTable with parallel keys[], values[], and state[] arrays (0=empty, 1=occupied, 2=tombstone). Linear probing on collision, doubles capacity when load > 70%. Hash function is byte-by-byte FNV-1a. A deterministic LCG drives 100,000 mixed operations.

Symmetry rules

No built-in map / dict anywhere in the hot path.
No built-in or stdlib hash functions; FNV-1a must be hand-rolled.
Power-of-two capacity, linear probing, tombstone deletes: every runner uses the same algorithm.

Why this benchmark exists

Pure interpreter throughput without the cheat of a C-level hash table. Reveals what the interpreter's array / arithmetic / loop performance looks like under sustained load.

Source code

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