Codon is a high-performance Python compiler that compiles Python code to native machine code without any runtime overhead.
The compiler was written with C++ and uses LLVM as the final assembly code optimizer. Unlike Python, Codon supports native multithreading, which can lead to even greater speedups.
Installation
You can compile Codon on your own machine, however there are precompiled binaries for Linux and macOS.
In the case of Linux just run this command below:
You must have cURL installed.
/bin/bash -c "$(curl -fsSL https://exaloop.io/install.sh)"At the end of the installation it will ask you to confirm the addition of the binary to your $PATH variable, press y for yes. If it doesn’t work, even opening it in a new terminal, as he advises, run the following commands:
echo 'export PATH=${HOME}/.codon/bin:${PATH}' >> ~/.bashrc
exec $SHELLTo test run:
codon --versionUsage
Suppose you have this code Python which is a Fibonacci function, you can interpret the code with Codon itself:
def fib(n):
a, b = 0, 1
while a < n:
print(a, end=' ')
a, b = b, a+b
print()
fib(1000)To run without compiling:
codon run fib.pyHowever, the performance will not be very good. The correct thing is to compile this code to a binary and then run the binary, example:
codon build -release -exe fib.py
./fibYou can still compile with [LLVM] optimization(https://terminalroot.com/tags#llvm)
codon build -release -llvm fib.py
./fibAnalyzing the performance
When we used a for loop with 1 million cycles in Python, as we did in this video, Codon was about 600% faster than native interpreter (version 3.10.9).
Test file/code can be obtained here.
Result using /usr/bin/python
time python main.py
1000000Ok
real 0m6,264s
user 0m3,530s
sys 0m2,415sThat is, actual execution: more than 6 seconds. 😞
Result using codon
codon build -release -exe main.py
time ./main
1000000Ok
real 0m0.795s
user 0m0.254s
sys 0m0,063sActual execution: in less than 1 second!!! 😲
