Markov text generation

Train a bigram Markov model on a 1 MiB corpus and emit 10,000 words.

Compile time · median (cold)

median of 7 runs

Native Gocompiled

107 ms78.2×

Pipitbytecode VM

1.37 msbaseline

CPython 3.14bytecode VM

443 µs0.32×

PyPy 7.3tracing JIT

335 µs0.24×

Ttengobytecode VM

232 µs0.17×

Sscriggobytecode VM

411 µs0.30×

Mmvmbytecode VM

386 µs0.28×

YyaegiAST walker

375 µs0.27×

Runner	N	Compile	Runtime	P95	Stddev	RSS	vs pipit	Status
Native Gocompiled	7	107 ms	5.35 ms	5.98 ms	444 µs	1.62 GiB	78.2×	OK
Pipitbytecode VM	7	1.37 ms	23.0 ms	24.8 ms	655 µs	591 MiB	1.00×	OK
CPython 3.14bytecode VM	7	443 µs	27.5 ms	28.0 ms	264 µs	n/a	0.32×	OK
PyPy 7.3tracing JIT	7	335 µs	11.3 ms	12.2 ms	641 µs	n/a	0.24×	OK
tengobytecode VM	7	232 µs	51.2 ms	70.8 ms	6.97 ms	1.69 GiB	0.17×	OK
scriggobytecode VM	7	411 µs	50.5 ms	51.4 ms	401 µs	1.69 GiB	0.30×	OK
mvmbytecode VM	7	386 µs	57.5 ms	58.1 ms	256 µs	1.69 GiB	0.28×	OK
yaegiAST walker	7	375 µs	43.0 ms	61.0 ms	8.06 ms	1.69 GiB	0.27×	OK

Workload & symmetry rules

Workload

Build a bigram model from a 1 MiB lorem corpus. Sample 10,000 tokens from a deterministic LCG-seeded RNG. Print a fingerprint of the output.

Source code