Why is Python SO slow!? 💀 #developer #tech #coder #softwareengineer #software #technology #tech

[jaredwilliams]

As someone who has to debug code, I would consider dynamic types a con as well.

[Martynyuu]

Oh, I thought because a Python has no legs

[Anonymous-urj]

Python developers explaining why their language being 1000x slower than every other one is a good thing:

[Gameplayer]

Remember, python is used for AI

It's the special limitation to prevent creating skynet

[rdwells]

One of Python's strengths is its use as a front end to libraries like NumPy, SciPy, and TensorFlow, which are written in high-level compiled languages like C and C++. I'm guessing that the particle simulation program Lewis described was written entirely in Python. If it had been written using NumPy, it may very well have been faster than the Java version.

[abdulniyas]

I remember one twitter post by Francois chollet(author of keras) in which he says, “The complaints "Python is slow" or "Python is unsafe" seem misguided.

The point of Python isn't to be fast or safe, it's to be flexible and hackable, and to interface well with everything else. It has become successful by serving as a frontend from which to call other libraries”

[actuallyasriel]

Python *is* compiled to machine code, just on the fly. You can also compile Python to intermediate code just like other languages and this improves the runtime a lot.

[nebula]

Can't wait for when the Python JIT compiler is finally complete

[kenbutsuri]

You can always combine c/c++ and python. And its relatively easy.

[planktonfun]

you can use python to convert python code to assembly and then run that assembly

[pretzel]

It's important to clarify what the test is when comparing speeds. Python looks much worse on shorter tasks due to the overhead Python requires on startup. This can be mitigated somewhat when compiling python to .pyc files. At medium tasks, python looks much better. At massive tasks we see the gap widen again, and this can be somewhat mitigated by compiling with optimized bytecode into .pyo files.

Dev time is worlds faster for python and it's portability is fantastic. Debugging is all just personal familiarity, because neither are that hard if you work with it. Portability can be matched by few compiled languages, but most include varying levels of additional steps/setup/tests.

Fighting over languages is silly, because at the end of the day the reason you use one over the other is because:
I'm paid to use [My Language]
I like the control/safety of [My Language]
I like the syntax/structure of [My Language]
I only know [My Language]
I'm conditioned to make fun of [Your Language]

At the end of the day there are workarounds to alter the standard behavior, mitigate the shortcomings, and interact with most APIs with whatever language you choose. If it's Turing complete, then you just need to get gud, lol.

[leodepew]

Honestly, I used to be a certified Python hater, but after working in it for a while, it is so nice to use, especially its JSON modules, makes parsing through data so easy.

[metgaming]

I just wish there was a mode to have it so you can specify a variables type yourself so Python doesn’t have to do it for you. Would significantly speed up run time.

[pretro]

You also forgot about the compatibility, Python can easily be paired with fast languages like c or rust. I like to write most of the code in python, and write the important functions in rust, that way my code is fast and easity to read.

[xx_Ashura_xx]

things to note is languages like java and c# are also "interptred" or rather use a runtime. but they also have JIT compilers which convert it into the target machine code on the fly

[colinmaharaj]

I'm sticking with C++ Builder and I don't do ML or AI yet.

[senzmaki]

technically the actual code is not executed line by line, that would be horribly inefficient, it first gets compiled into byte code which is what's executed "line by line"


















🤓 ☝️

[KangoV]

Python is also single threading. Most libraries are C/C++ underneath.

[litjellyfish]

And Java can also be considered slow depending on what you compare it to. Also if parts of the Python code is really slow you can create that function in C / C++ and just call it.

[rustygear]

A good point about using python in AI: if it used C++ or assembly, it would be already running the world.