Compiler vs Interpreters

Last updated Mar 24, 2023

• 61 c

See also: Number Representation

C works in compilation. It takes a program, and turns it into architecture-specific machine code (a string of 1’s and 0’s). Architecture-specific because companies like Apple and Intel have their own machine code.

This is unlike Java, which converts to architecture-independent bytecode that may then be compield by a just-in-time compiler. Java has both a compiler and interpreter.

Python does a realtime bytecode compilation. It compiles to bytecode as it runs.

These differ mainly in exactly when your prrogram is converted to machine code.

Compiled languages in general are as fast as it gets. It basically runs on raw silicon.

But then why do we do scientific computation in Python?

• Good libraries for accessing GPU-specific resources.
• Also, many times python allows the ability to drive many other machines very easily.
• Also, Python can call low-level C code: Cython

The “Change -> Compile -> Run” cycle is slow.

Also, we have to compile for each architecture we want to run on. This movement is called “portability”, and is always a pain.

• If you type make -j you can compile in parallel, which is quite fast. Linking, however, is still slow (the bottleneck) because of Amdahl’s Law.