how jit compilers are implemented and fast: pypy, luajit, graal, and more

Interpreted languages still go down to bytecode
Function reuse happens, if you compile multiply(1,2), and then wrote multiply(2,3) then the already compiled version will be used
- llvm bitcode can be seen
jits don’t compile ahead of time, but still compiles source to machine code
- not strictly about compiling code just-in-time, but compiling optimial code at an optimal time
how do we generate optimal code? not all assembly instructions are created equal
- humans typically write better assemly than a computer
jits usually are more expensive but make up in optimized compiled code