23. High Performance in Dynamic Languages

[@valeriuok]

Pat yourself on the back if you went through the whole course. Great stuff!

[@DrewryPope]

this was amazing thank you i went out and bought a book on julia right after this, amazing

[@leixun]

*My takeaways:* 1. Julia 3:51 2. Why Julia can be fast 10:16

[@skyrim7183]

🤣

[@academicalisthenics]

Very insightful, great lecture!

[@ВадимН-о9х]

it's actually not that insightful if you have been programming in dynamic languages, cause if you did, you would intuitively know pretty much everything from this lecture, like i did, cause at my day job i use PHP and JS

[@luisepinos]

Insightful and useful content. Congrats!

[@nikolayzapryanoff1032]

Amazing lecture!

[@JustHereForDaContest]

Holy crap jupyter is cool!

[@zsoltbr]

I would have never thought computer programming is that a physical activity: he's clearly out of breath at the end of the lecture. :)

[@chuanqisun]

The professor squeezed out yet another 1.4x improvement by overclocking his brain.

[@PhuongNguyen-gq8yq]

Well 49:31 multiple dispatch is just cheating

[@MrSmeeds]

At 1:15:30 - "there's a small core [of Julia] that's implemented in C for bootstrapping". Would it be possible to implement the whole language in julia?

[@8292-d6n]

nope, there always has to be 1 little thing that knows how to initiate this bootstrapping mechanic. However you could write a small compiler in C that compiles Julia Core to Assembly, let's call that C1 and then write a full Julia compiler in Julia which we can call C2. Now you can compile C2 with C1 and get the binary for a full Julia compiler. Now you can throw away C1 and only use C2 which is entirely written in Julia and compiles Julia code but you have to keep the binary so you moved the C portion of the compiler to a binary file.

[@green6578]

As the guy with the number-as-a-name said, you ALWAYS need something to implement the bootstrap compiler initially. You don't have to do it in in C; if you want to be a maximum completionist you can do it in machine code, if you want to be a little lazier you can do it in assembly, if you want to be practical it is always done in C. This is because C is fast and the most portable language to ever exist (C can be compiled for basically any architecture). As the lecturer said, they already convert to LLVM to do machine-level code generation, so even if all of their code was in Julia they'd still be dependent on tons of C and C++.