I'm developing a tiny game engine in C(11) and I've built profiling into the core, and I profile many of the math-heavy parts separately as well. Clang 3.3 actually almost always does better than gcc 4.8 here. Not by much, but there you have it. You should take a look at the SLP vectorizer, which will come enabled in -O3 as of Clang 3.4 but can already be enable separately with -fslp-vectorize.

So for single threaded code I'm already leaning towards Clang. OpenMP is going to get integrated as well, as of then, all bets are off. Exiting times to be a C/C++ dev... (or any other kind, for that matter, LuaJIT never ceases to amaze me).

He probably just forgot to type the word "former" by accident, as he also mentioned Almaty, which afaik is a city in Kazakhstan (part of the former Sovient Union).

Thank you for a short and _very_ informative post, mod up!

I also use Dev-C++ for my windows development, I find the interface to be superb in not getting in my way, while not being dog-ugly or slow.

Conversely, I was searching for an alternative for linux (yes, I'll learn vi/vim some day, just not now), and what I found was Geany. It starts up at the flick of a finger and manages to look a lot like Dev-C++ (perhaps more polished even). The only thing I dislike about it is its (intentional) weak project support, but that could conceivably be fixed with the nice plugin system. So, Geany is very lightweight and supports quite a lot of languages, it has its default settings such that if you have the necessary compilers installed (which is basically mostly true on most linux systems, or can be done with one line at the shell), it'll work out of the box. The same for python, perl, etc.

Then I noticed that it also has a windows port, that comes with its own GTK+ runtime (all nicely contained in Geany's folder, and nicely uninstalled if you want), I tried it and while the first startup is not as fast (due to the GTK+ libraries needing to be loaded), all the rest is just as snappy. Ofcourse the windows version needs a little more help to get started, but not _that_ much more. All you basically have to do is install MingW and set your $PATH to search MingW/bin.

So basically the only problem with Geany that still remains (for me), is that it doesn't really support projects like Dev-C++ did, so for now you basically have to make your own makefiles. This could also be viewed in a positive light ofcourse, as learning about makefiles will prove to be a good skill if you want to do some open source contribution. But I'm sure it could get tedious as well (haven't done many large projects lately), so someone developing a plugin would be really nice, and probably not too hard as well.

Well, I've been ranting and promoting long enough now, time for Geany to promote itself, give it a spin (it's free, the only cost is your time ;) ).

I fully agree with your post (I prefer C over most other languages myself for some weird reason, but if it n eeds to be made *right now*, I'll use Python/Perl/Bash/...).

However, there is an addendum I'd like to make here:

Some programming languages force you to think in ways that are actually beneficial to the speed of your code, and can outpace the "normal" C program significantly.

For example, a few months ago I was forced to write something concerning an AI algorithm in Prolog. Now, I was cursing and cursing at it, because the constraint solver built into the prolog compiler was so goddamn restrictive, but that's how constraint solving works. Every time I was thinking to myself: "if I'd have been allowed to build this in C, it would be done yesterday, and probably be faster to boot!"

But when I ended up finishing it and running it, it was blindingly fast, and I queried my professor about it. He told me that another student some time ago was thinking the same thing as me, and actually made a C variant. It ran 4x as slow as the prolog equivalent even after spending quite some time optimizing (interchanging algorithms and even looking at the assembler code, he told me).

Then he told me what was causing this discrepancy, as I had always thought that C was the end all be all of performance. It was the restrictive nature of the prolog solver that caused me to put more brain power into the problem, and as such shift work from the computer to the human. Because those same restrictions allowed lots and lots of optimisations (aliasing comes to mind).

I wish I had mod points right now, because I sure as hell was too stupid to realise that (although maybe I can feel less inferior by ascertaining myself that English is not my mother tongue).

I do feel the need to checkup on "6x7" being the last question though