GCC has generally had perf numbers that lag the commercial compilers. Although GCC has narrowed this gap recently as C/C++ has generally fallen out of favor by the major vendors (and commercial vendors aren't working much on raw C/C++ performance anymore).
I think thats because GCC will compile anything anywhere. MS is only Windows platform, and ICC prefers Intel x86 CPUs. Since the other compilers didn't stand a chance to compete with GCC on portability, the only other thing they could compete with it on was performance. Since they focused on it, they should clearly be more performant in their niche.
Since the other compilers didn't stand a chance to compete with GCC on portability, the only other thing they could compete with it on was performance. Since they focused on it, they should clearly be more performant in their niche.
I think your skewed world view is apparent here. MS and Intel certainly didn't come and say, "One of our top priorities is portability. How can we target our compilers to platforms that we have nothing to do with? Oh wait, GCC already does that!"
Of course not. Honestly, GCC has been an afterthought for most of the history of C/C++. It's only now that it is effectively dead that GCC has started to become competitive. MS was focused on Borland for years (and it actually didn't generally revolve around perf, but generally build throughput). Likewise, Intel was focused on SPEC against xlc and other architecture specific compilers.
And as I said in another post, they (GCC, Intel, MS) have different priorities. If your goal is to have a somewhat portable C/C++ compiler then GCC is the way to go. If you want the best perf/throughput on a specific target then MS and Intel are your best choice.
On pretty much all benchmarks -- just Google scholar search. We used GCC but only because it's portable. Performance wise on IA-32(e) GCC is {EDIT:Language} not as good.
To answer your question, probably not. icc has a long history of being less good on AMD processors, and almost certainly not on accident. With that said, they don't have a monopoly on x86 compilers. Microsoft's Visual C++ has been about equally good on both platforms and is far more dominant.
But with that said, I'm sure you know, that's not really relevant to the discussion at hand :-)
So, the Intel's compiler trounces GCC when the code it emits run on certain x86 processors. And Microsoft's compiler does a much better job when the target runs on Windows on x86 processors.
If we narrow the target enough, all comparisons become interesting. I am quite sure GCC will blow both Intel's and Microsoft's compilers out of the water when it comes to compile code for IBM mainframes. I am sure it's much better than Intel's when it comes to target my phone's ARM processor.
And my blender makes better smoothies than GCC. So your point is that GCC is better at things that other tools aren't attempting at all? OK, good point.
Your point was that GCC was not as good at specific things that are the sole focus of the products you mention. Intel's compiler only targets Intel's x86s and Microsoft's compiler only targets the (two or three) OSs Microsoft makes. It's not really an apples to apples comparison.
For me the Intel compiler makes sense in certain specific applications (many of my servers are not even x86), but the Microsoft compiler is absolutely worthless.
My point was that GCC is not good at specific things that customers care about (and it is good at other things, but I'd argue these things are more niche, such as retargeting for custom embedded devices).
For example, code size, throughput, debugging, and optimization are all places that the Visual C++ compiler is superior than GCC. How many Windows games are written with GCC? How many top-tier Windows apps are written with GCC? If I'm writing a Windows C/C++ application, I'm going to either use Visual C++ or icc. I'll choose icc for code with lots of compute intensive loops or vectorization opportunity. VC is better when icache is an overriding concern.
Now that may not be what you do, but to say, "Who in their right mind writes programs for Windows?", I think says more about you than it does about the respective compilers.
From my point of view, Windows desktop software is a niche. It's a good, big and profitable one, but one I do not take part in. I agree Visual C++ is a great compiler for that specific target.
Still, there is a lot more to programming than desktop applications that run on x86 processors under Windows. And embedded ARM processors outnumber desktop x86s by a large margin. I carry at least three of them with me at all times.
And embedded ARM processors outnumber desktop x86s by a large margin. I carry at least three of them with me at all times.
And Visual C++ targets ARM also, and has for years. One of the reasons I've heard bandied about as to why WP7 first party apps are so fast and responsive is because of the quality of the Visual C++ compiler. :-)
Yes, the number of non MS-ARM devices is probably several of orders of magnitude more popular. But you'd be surprised at how many embedded devices run Windows CE.
Regarding WP7, 3rd party apps are managed. But the core OS is native and 1st party apps can be native. For example, IE and the mail app are both native C++ apps.
Honestly, I think allowing 3rd party native apps would be a mistake in general. Although I'd create a partners program where you require devs to pay $10k and then go through a certification process, which allows them to ship native apps. Basically its the XBox model. Indie devs use managed code. Apps that are willing to go through the more expensive process get native access.
Why? Native apps introduce a bunch of problems that are a pain to deal with. For example, there a bunch of security issues and stability issues that native apps introduce. Additionally you have to target native apps to an architecture. It may lock WP7 into ARM, due to the ecosystem, and block it from moving, for example, to a new Intel architecture.
