Back to General and Gameplay Programming

Why is STL so slow?

Vorlath · 2002-06-29T11:25:00

I was doing some timing tests between a list and my own implementation (for lists only). Why is it that my version is so much faster? It''s not ever worth using STL at all. And I had to take out the comparison on iterator.end() because this slows down STL by a factor of 100. All it does is just compares with NULL and it''s a function call that VC doesn''t make inline. Why do people use STL? It''s certainly not for the speed. list<int>::iterator i; list<int> l(100000); list<int>::iterator end = l.end(); int cnt = 0; for (i=l.begin();i!=end;i++,cnt++) { *i = cnt; } If I replace list with my own linked list implementation, it''s so much faster. It''s even worse with deque and vector.

General and Gameplay Programming Programming

Started by Vorlath June 27, 2002 12:44 AM

34 comments, last by Vorlath 21 years, 9 months ago

UnshavenBastard

376

June 28, 2002 01:45 AM

quote:Original post by Stoffel

quote: "It's not ever worth using STL at all."

All generalizations are false.

hehe, isn't that one, too ?

paradox sentence...headache...

[edited by - UnshavenBastard on June 28, 2002 2:48:55 AM]

DerekSaw

243

June 28, 2002 01:50 AM

quote:Original post by Stoffel
Sounds like pilot error. I'm unaware of std::string leaks in MSVC 6.0's STL. We use strings extensively in several large projects. The one criticism I have heard about std::string is that it tends to make many small allocations, which can be a problem for certain memory models; I believe this can be solved with a custom allocator.

quote:Original post by daerid This is vc.net (vc7), not vc6

Maybe it was the debugger I was using, but I created a tiny char* wrapper that did what I needed, and eliminated the (perceived) leak.

[edited by - daerid on June 27, 2002 12:28:14 AM]

std::string in vc.net somehow maintain some internal buffer for very small bytes of char (say, maybe 16 or less).... and the buffer (just assumption) may not act as it is destructed in destructor (but freed in somewhere else)... so the debugger may detect leakage. so beware.

[edited by - DerekSaw on June 28, 2002 2:58:14 AM]

"after many years of singularity, i'm still searching on the event horizon"

dalleboy

324

June 28, 2002 04:01 AM

quote:Original post by Stoffel
All generalizations are false.

Then why are we using templates? If generalizations are false, then all templates could be optimized to a single false-statement. Including STL would give us noting more than a lot of false-statements....

Anonymous

June 28, 2002 10:42 AM

all science is about generalizations

Fruny

1,658

June 28, 2002 03:08 PM

Look at this. Debunking straight from the source...

"Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it." — Brian W. Kernighan

Arild Fines

968

June 29, 2002 11:25 AM

You just _know_ it''s going to be bad if its crossposted to a *.*.advocacy group.

--AnkhSVN - A Visual Studio .NET Addin for the Subversion version control system.[Project site] [IRC channel] [Blog]

Why is STL so slow?

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Why is STL so slow?

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines