Codon VS Lua benchmarks

Current benchmark data was generated on Sun Sep 27 2026, full log can be found HERE

CONTRIBUTIONS are WELCOME!

CPU INFO:[x86_64][4 cores] Intel(R) Core(TM) i7-3687U CPU @ 2.10GHz (Model 58)

* -m in a file name stands for multi-threading or multi-processing

* -i in a file name stands for direct intrinsics usage. (Usage of simd intrinsics via libraries is not counted)

* -ffi in a file name stands for non-stdlib FFI usage

* (You may find time < time(user) + time(sys) for some non-parallelized programs, the overhead is from GC or JIT compiler, which are allowed to take advantage of multi-cores as that's more close to real-world scenarios.)

show numbers without parallelization

show numbers with parallelization

helloworld

Input: QwQ

lang	code	time	stddev	peak-mem mem	time(user)	time(sys)	compiler compiler/runtime
lua	1.lua	1.6ms	0.3ms	2.4MB	0ms	0ms	luajit 2.1.1703358377
lua	1.lua	2.2ms	0.4ms	1.8MB	0ms	0ms	lua 5.4.6
codon	1.py	9.9ms	0.8ms	10.9MB	7ms	3ms	codon 0.20.2

nbody

Input: 5000000

lang	code	time	stddev	peak-mem mem	time(user)	time(sys)	compiler compiler/runtime
lua	4.lua	1312ms	8.1ms	1.6MB	1303ms	0ms	luajit 2.1.1703358377
codon	1.py	2621ms	35ms	11.0MB	2900ms	0ms	codon 0.20.2
lua	4.lua	19839ms	364ms	1.8MB	19823ms	0ms	lua 5.4.6

Input: 500000

lang	code	time	stddev	peak-mem mem	time(user)	time(sys)	compiler compiler/runtime
lua	4.lua	139ms	4.9ms	1.5MB	127ms	0ms	luajit 2.1.1703358377
codon	1-m.py	357ms	59ms	11.0MB	610ms	0ms	codon 0.20.2
lua	4.lua	1941ms	5.9ms	1.8MB	1927ms	0ms	lua 5.4.6

nsieve

Input: 12

lang	code	time	stddev	peak-mem mem	time(user)	time(sys)	compiler compiler/runtime
codon	2.py	1317ms	10ms	110.4MB	1597ms	10ms	codon 0.20.2
codon	1.py	1510ms	4.1ms	49.9MB	1783ms	10ms	codon 0.20.2

Input: 10

lang	code	time	stddev	peak-mem mem	time(user)	time(sys)	compiler compiler/runtime
codon	2-m.py	292ms	4.4ms	29.3MB	570ms	3ms	codon 0.20.2
codon	1-m.py	358ms	6.6ms	20.7MB	643ms	0ms	codon 0.20.2