Programming Language and compiler
Benchmarks

Codon VS Lua benchmarks

Current benchmark data was generated on Tue Mar 25 2025, full log can be found HERE

CONTRIBUTIONS are WELCOME!

[x86_64][4 cores] AMD EPYC 7763 64-Core Processor (Model 1)

* -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.)

helloworld

Input: QwQ

lang code time stddev peak-mem mem time(user) time(sys) compiler compiler/runtime
lua 1.lua 1.5ms 0.2ms 2.1MB 0ms 0ms lua 5.4.7
lua 1.lua 1.5ms 0.1ms 2.4MB 0ms 0ms luajit 2.1.0-beta3
codon 1-m.py 11ms 0.7ms 8.6MB 20ms 0ms codon 0.18.2

nbody

Input: 5000000

lang code time stddev peak-mem mem time(user) time(sys) compiler compiler/runtime
lua 4.lua 717ms 7.0ms 2.3MB 703ms 0ms luajit 2.1.0-beta3
codon 1.py 1322ms 1.2ms 8.8MB 1630ms 0ms codon 0.18.2
lua 4.lua timeout 0.0ms 2.3MB 4990ms 0ms lua 5.4.7

Input: 500000

lang code time stddev peak-mem mem time(user) time(sys) compiler compiler/runtime
lua 4.lua 77ms 2.5ms 2.3MB 70ms 0ms luajit 2.1.0-beta3
codon 1-m.py 176ms 3.4ms 9.0MB 487ms 0ms codon 0.18.2
lua 4.lua 1108ms 50ms 2.3MB 1100ms 0ms lua 5.4.7

nsieve

Input: 12

lang code time stddev peak-mem mem time(user) time(sys) compiler compiler/runtime
codon 1-m.py 461ms 8.3ms 47.7MB 763ms 0ms codon 0.18.2
codon 2.py 1572ms 19ms 100.8MB 1877ms 27ms codon 0.18.2

Input: 10

lang code time stddev peak-mem mem time(user) time(sys) compiler compiler/runtime
codon 1-m.py 125ms 0.2ms 18.6MB 430ms 0ms codon 0.18.2
codon 2-m.py 397ms 1.3ms 27.1MB 707ms 3ms codon 0.18.2