Accelerating Scientific Workloads with Numba - Siu Kwan Lam

[@cometor1]

I just went ahead and used the jit wrapper. Decreased my calculation time by 80%.
But I'm also thinking about switching to JULIA if similar calculations have to be done a few more times.

[@Shreendg]

What did you find in the end? Is Julia faster than Numba?

[@cometor1]

@@Shreendg don't know, not for me. I've been sticking to Python. Julia may he the future, but if python does all things you need just stick to it.

[@nycoshouse]

numba don't support infinite integer :'(

[@Stopinvadingmyhardware]

infinites are rational.

[@nycoshouse]

@@Stopinvadingmyhardware ( you can't use numba to compute on huge integers )

[@justinzhang9935]

As a heavy user of Pandas, dealing with numbers and strings all day, I have found that Numba and Modin are just jokes for me. Maybe establishing all projects in the future on JULIA is the only hope. Native LLVM, no GIL, easy grammar for Python users ... Wow.

[@stxnw]

numba is for mathematical calculations not data cleaning or whatever u are doing.

[@keijioza3479]

@@stxnw lmao, exactly

[@keijioza3479]

Lol... with what you're doing, I'd say either use sql for other things as Pandas is only great for dataframes but is kinda slow on other uses... then use Cython. Python can be fast

[@NeoKailthas]

After spending few days on this, just learn C and save your self some time. This is like reinventing the wheel

[@baodau2230]

you have to live it

[@NeoKailthas]

@@baodau2230 I got it to work. What a waste of a week lol

[@MohamedAbomokh]

you really think so ?

[@bernardfinucane2061]

I think it takes more a week to get proficient at C.

[@NeoKailthas]

@@bernardfinucane2061 haha yes assuming you already know c or c++. Also if you gonna do it, use cuda jit kernels. Don't waste time on vectorize or even worse guvectorize