Speed comparison between C++ and Python for a program that counts to 1,000,000,000.

2.4s? Was your C++ binary writing all those numbers to the console?

Nope, not at the console. In both programs (in C++ and Python), only a for loop was established, and the final output, i.e., 1 billion, was outputted to the console.

I’m too old to put up with that nonsense anymore. I’d rather perfect the programs appearance and logic in Python then convert and optimize portions or the entire program later. Speed during design and development is a metric that matters far more than execution speed during design and development phase. You can figure out if a program and workflow will work with real users MUCH FASTER with a language like Python. Anything can be optimized and/or rebased later.

You clearly made a mistake and did a poor job in your python code.

Stupid engagement bait for weak programmers.

If I'm not wrong it mostly shows C++ compiler optmizn deleting empty loops, vs .py intrptr overhead. Running python through PyPy or NumPy narrows that gap instantly. Great example of compiled vs. intrptr design though

@Adi_Verse_ performance critical parts are implemented in C in numpy.

This means nothing without also posting the source code used.

Code used (C++): int main() { size_t n = 0; while (n < 1'000'000'000) n++; std::cout << n << std::endl; } Code used (Python): n = 0 while (n < 1000000000): n += 1 print(n)

only 0.2 sec with python here

Now do machine language.

The optimising phase of the C compiler sees that the program produces no output, so all the counting can be optimised out, or at least replaced by just setting the variable to the end value.

Complines down to; C++ --> Machine Code Python --> C++ --> Machine Code

Show the code because that looks just like compiler optimization. C loops should be faster regardless but like not THAT fast

Apples and oranges. ASM will have C++ eating its dust.

I code in ASM. difference of speed between C and ASM is not even noticeable. even tho asm is obviously faster. to humans, gap between <1s and <2s is irrelevant.

@TopCurlsoid People who don’t see real value behind the different purposes of languages they use are easily measuring the things they understand; speed, time or delay. But if that is necessary to have just a loop like that, why don’t you write it as asm function and call it from Python prog?

@wholyv Precisely. And if you need to optimize coefficient's of a horner's scheme, use Python, not C or ASM.

are we serious still doing this? in 2026?

🐍 is still better for ML

@grok 何か胡散臭いんだけど…

It's bs to compare a compiled language and interpreted one

w/o seeing the code this post is useless

you need to compile it with -O0 to avoid constant propagataion and loop unrolling. C++ doesn't actually run the loop if the optimizations elliminate it completely before generating assembly code. rookie mistake, terrible example

I got even larger difference.

it's the comparison between monster and baby chicken!

Different use cases for these languages. Plus your test case is silly because it serves no real purpose

how about rust ?

C++ makes it harder for you to shoot yourself in the foot but once you do it it blows away your entire leg

😅

Interpreted Vs compiled of course the interpreted one will be slow.

Given the post AND the comments, I understand most people don't know the differences in execution time magnitude between compiled and interpreted code, and why that doesn't really matters when most of your computation is done by a compiled lib...

Delete your account

コンパイル言語なんだから当たり前の話だと思いますが・・・

That's not what Python is design for. Python is a glue code, when it needs speed it calls C++ libraries. It's strength is how easy it is to learn and use. Each language has its pros and cons, use the right tool for the job.

The real trick is to call the C++ from python, and use python for everything that is IO bound instead of CPU bound.

There used for different things so this is pointless

@grok compare avec rust

#include <stdio.h> #include <time.h> int main(){ volatile int count=0;clock_t start=clock(); for(int i=1;i<=1000000;i++)count=i;clock_t end=clock();double time_ms=1000.0*(double)(end-start)/CLOCKS_PER_SEC; printf("%.3f ms\n",time_ms,count); return 0; } 3.000 ms

It does not matter anymore. Programming language were a filter for binary so us humans can work with machine. With AI, this becomes completely obey

Python is slower than C++ in raw computation, but its real world speed comes from how quickly you can build and iterate making it faster for development even if it's slower for execution. E.g on Msc project doc retriever used python later embedded c for more faster retriever