/sci/ - Science & Math » Thread #11868951

51KiB, 1024x1024, 1024px-Python-logo-notext.svg.png

View Same Google iqdb SauceNAO

>implying Python is slow

Anonymous Sun 05 Jul 03:48:04 2020 No.11868951 View Reply Original Report

Quoted By: >>11868967 >>11869127 >>11869450 >>11869474 >>11869689 >>11870057 >>11870062 >>11871227 >>11871237

>/sci/ says Python is slow
Python functions are implemented in C or Fortran you idiots. There a slight overhead but it's small.

Anonymous

Anonymous Sun 05 Jul 2020 03:55:22 No.11868967 Report

Quoted By: >>11868997 >>11869461

>>11868951
python is slow though, everybody knows that. you even say that it's implemented in C "or Fortran" (kek, no it's not implemented in fortran. idiot). if you admit that, then you admit the python interpreter introduces some (huge) slowdown in how it interprets things into a real programming language

Anonymous

View Same Google iqdb SauceNAO _20200630_115125.jpg, 219KiB, 720x790

Anonymous Sun 05 Jul 2020 04:05:56 No.11868997 Report

Quoted By: >>11869005 >>11870062 >>11870538

>>11868967
Maybe you shouldn't believe everything you read on the internet. This is wall clock time btw.

This board is really full of idiots.

Anonymous

Anonymous Sun 05 Jul 2020 04:08:54 No.11869005 Report

Quoted By: >>11869017

>>11868997
ooooh wow anon imported numpy

remember that numpy is implemented in C. not python or fortran or anything else.

the fact that you said "python functions are implemented in Fortran" or the like means that you are a fucking retard. Fortran is deader than a doornail and anybody who shills fortran is basically a dinosaur or a zombie

Anonymous

Anonymous Sun 05 Jul 2020 04:14:12 No.11869017 Report

Quoted By: >>11869054

>>11869005
I never claimed that python is in any way special because numpy and friends are actually implemented in Fortran, but that's what you use to solve problems. Nobody writes matrix diagonalization him-/herself, you would always resort to libraries with efficient implementations. The same is true for compiled languages. The point that python is extremely slower than C is moot. Sure, it is slower, but not much. These critiques always come from uninformed people who have no idea how to use a high-level language and instead perpetuate myths they heard about on some boomer forum or blog.

Anonymous

Anonymous Sun 05 Jul 2020 04:29:11 No.11869054 Report

Quoted By: >>11869093

>>11869017
>These critiques always come from uninformed people who have no idea how to use a high-level language and instead perpetuate myths they heard about on some boomer forum or blog.
yeah sure python is great for importing code implemented in faster languages. that doesn’t help your argument

Anonymous

Anonymous Sun 05 Jul 2020 04:43:41 No.11869093 Report

Quoted By:

>>11869054
I bet this guy thinks he isn't using Lapack or BLAS in every project.

Anonymous

Anonymous Sun 05 Jul 2020 04:51:09 No.11869105 Report

Quoted By:

Python is training wheels for retards who can't into C

Anonymous

View Same Google iqdb SauceNAO CE6B579F-8D19-4644-AE33-81D1D990 (...).jpg, 122KiB, 735x963

Anonymous Sun 05 Jul 2020 05:04:50 No.11869127 Report

Quoted By: >>11871058

>>11868951
>he doesn’t know about the GIL
also /g/tfo and sageru

Anonymous

Anonymous Sun 05 Jul 2020 07:22:33 No.11869343 Report

Quoted By:

The speed of python doesn’t matter for anyone here. Unless you are doing some crazy shit. Who the fuck cares if your script takes 2 seconds vs .25 seconds. If you are using python for something where speed is important you need to completely rethink what you are doing.

Anonymous

Anonymous Sun 05 Jul 2020 07:41:11 No.11869401 Report

Quoted By: >>11871058

It is slow and it can easily be proven. Speed is not why it's used anyway so this statement is something no one gives a fuck about in the first place.

Anonymous

View Same Google iqdb SauceNAO lightspeedpython.png, 92KiB, 772x858

Anonymous Sun 05 Jul 2020 07:43:32 No.11869406 Report

Quoted By:

Anonymous

Anonymous Sun 05 Jul 2020 07:46:46 No.11869408 Report

Quoted By:

Pure python will almost always be slower than the equivalent C/C++/whatever native language, but that's not the point of using it. The real value come from how quickly you can proto-type/explore new ideas in it. If you need the speed then you start looking at where your runtime is spent, and move that to a native language and speed up from there. And with things like numpy/cython/numba you can usually get away without ever writing a single line of true C or C++ code before things are "fast enough".

Anonymous

Anonymous Sun 05 Jul 2020 08:20:50 No.11869450 Report

Quoted By:

>>11868951
Are you the guy who got his ass blasted in the Julia thread?

Anonymous

Anonymous Sun 05 Jul 2020 08:29:39 No.11869461 Report

Quoted By:

>>11868967
PyPy is mostly faster than CPython

Anonymous

Anonymous Sun 05 Jul 2020 08:32:08 No.11869463 Report

Quoted By: >>11869469

When you can write code with pure numpy functions, you can mitigate the speed issue.
If you find yourself writing double or triple for loops in your Python code when processing large image, you're fucked.
That's my experience as a EO scientist.
I generally do prototyping in Python first with toy data, then rewrite everything in C/C++ when processing real actual data.
People shitting on either of those languages don't understand you need different tools for different tasks, and that there isn't a language better than all the rest.

Anonymous

Anonymous Sun 05 Jul 2020 08:35:06 No.11869469 Report

Quoted By: >>11869492

>>11869463
Pythonistas think their language is suitable for everything

Anonymous

Anonymous Sun 05 Jul 2020 08:36:46 No.11869474 Report

Quoted By:

>>11868951
pylons only go fast if you know what you're doing
it's almost as if working smart requires high iq

Anonymous

Anonymous Sun 05 Jul 2020 08:44:21 No.11869492 Report

Quoted By: >>11869636

>>11869469
No one thinks this. It’s a complete straw man.

Anonymous

Anonymous Sun 05 Jul 2020 10:33:54 No.11869636 Report

Quoted By:

>>11869492
No

Anonymous

Anonymous Sun 05 Jul 2020 10:42:30 No.11869644 Report

Quoted By:

Numpy is fine if your problem falls within its constraints. Problem is that a LOT of problems do not.
Python is really great, but you will need to know some C/C++ to offload your heavier loops for anything nontrivial.

Alternatively take the Julia pill.

Anonymous

Anonymous Sun 05 Jul 2020 11:09:57 No.11869689 Report

Quoted By: >>11869709

>>11868951
Pure python might be slow because it is a script language and not compiled language. BUT you can write code in C and integrate it with cython, If there is a bottleneck in your code, you can always make it faster by writting it as a C extension. There is a reason python is used: it is great for interfacing with other languages.

For instance, most Machine Learning libraries are written in python, but under the hood it is actually C code.

Anonymous

Anonymous Sun 05 Jul 2020 11:21:37 No.11869709 Report

Quoted By: >>11869735 >>11869876

>>11869689
Once again, C is used for the real work. Python is a scripting language for children, and will remain as such

Anonymous

Anonymous Sun 05 Jul 2020 11:37:57 No.11869735 Report

Quoted By:

>>11869709
most Machine Learning libraries are written in python

Anonymous

Anonymous Sun 05 Jul 2020 12:38:52 No.11869847 Report

Quoted By:

Python is used to throw mockups together. It's used for Web development because most web devs start with html and css and then can't learn an actual programming language so they use an easily readable one like python.

Any computationally intensive work will be done in C++, Java or C.

Anonymous

Anonymous Sun 05 Jul 2020 12:51:20 No.11869876 Report

Quoted By:

>>11869709
Ok pal. Whatever you say. I dont have preference of one over another. I use them freely, so that I can take advantage of both.

Anonymous

Anonymous Sun 05 Jul 2020 14:52:13 No.11870057 Report

Quoted By:

>>11868951
If you want speed, you use python as “glue” for modules written in lower level languages.

Anonymous

Anonymous Sun 05 Jul 2020 14:56:33 No.11870062 Report

Quoted By: >>11870094

>>11868951
>>11868997
Jesus fucking Christ this board really is in a literal timeloop we had this exact exchange last week

Anonymous

Anonymous Sun 05 Jul 2020 15:11:02 No.11870094 Report

Quoted By: >>11871058

>>11870062
You're talking to a guy who takes photos of his screen, of course he's retarded.

Anonymous

Anonymous Sun 05 Jul 2020 18:04:12 No.11870538 Report

Quoted By:

>>11868997
Haha, guys look this matrix transpose took less than a millisecond! Haha guys that is super fast right? That is so fucking slow for taking the transpose of a matrix. Do you know how many clock cycles are used in python’s matrix transpose V.S. the ideal? Well given a 1GHz clock means there are roughly 50 million micro-ops, or about 12 million macro operations. For a 100x100 matrix, 100000 entries, that equates to ~1200 per entry in the matrix. That is, it is super inefficient. BTFO pytard

Anonymous

Anonymous Sun 05 Jul 2020 20:55:08 No.11871058 Report

Quoted By: >>11871114

>>11869127
If you didn't understand I know what GIL is your reading comprehension is even below the average in this shithole.
That doesn't matter so much anyway because you can use multiprocessing to speed up embarrassingly parallel problems. When you can rent a machine with 64 cores and then spin ti down after 5 minutes it's really not a big deal.
>>11869401
>It is slow and it can easily be proven.
Where? Numb nut
>>11870094
>You're talking to a guy who takes photos of his screen, of course he's retarded.
Network blocks 4chan. I know it's difficult to understand for you, but maybe if you try reeeeally hard.

Anonymous

Anonymous Sun 05 Jul 2020 21:12:55 No.11871114 Report

Quoted By: >>11871175

>>11871058
>Where? Numb nut
Literally the post below mine. Numb nut
As well as many other examples. Even if you make it do a bunch of basic calculations it will lag behind most other mainstream languages. Have you ever even programmed?

Anonymous

Anonymous Sun 05 Jul 2020 21:33:43 No.11871175 Report

Quoted By:

>>11871114
>Have you ever even programmed?
you're projecting. As I said (I know you can't comprehend written word), I'm well aware of everything that was said. You just don't understand the difference between.. why am I even bothering

Anonymous

Anonymous Sun 05 Jul 2020 21:53:53 No.11871227 Report

Quoted By: >>11871234

>>11868951
syntax wise it's amazing, considering it's 30 years old, and its ecosystem makes it the best choice for scientists

having that said 40x performance overhead compared to pure ansi C (ten times slover than javascript) makes it a laughable pile of shit, and there is no excuse for such state of things. its well funded, with large developer community, why don't you put JIT into it already for fuck sake

there are bindings for fortran, sure, but outside of scripting, when you have to keep long running processes everything tends to go tits up, because of memory handling, i/o or, god knows why. been trying to put a production system with pyspark, ultimately we ended up with scala implementation, because it just couldn't stop crashing at random

Anonymous

Anonymous Sun 05 Jul 2020 21:55:59 No.11871234 Report

Quoted By: >>11871250

>>11871227
>another retarded muppet anon thinks performance is important
into the trash the reply goes.

Anonymous

Anonymous Sun 05 Jul 2020 21:58:50 No.11871237 Report

Quoted By:

>>11868951
for competitive programming,programs made in python are given 5 times the allowed time as c++ programs to account for overhead. Java is given 10 times as much a c++
Yes Python is slower for almost all algorithmic applications..

Anonymous

Anonymous Sun 05 Jul 2020 22:05:25 No.11871250 Report

Quoted By: >>11871308

>>11871234
it has no place in production, outside of scripting and webshit

it can offload computations on native libraries, but fails at garbage collection making it unusable in the long run

Anonymous

Anonymous Sun 05 Jul 2020 22:28:54 No.11871308 Report

Quoted By:

>>11871250
>he thinks you cant use python in production
lmao retard you know nothing about anything. are you a freshmen?

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches