From the Press Release: “Intel Corp today announced version 5.0 of the Intel C++ Compiler for Linux and the Intel Fortran Compiler for Linux. The compilers are specifically designed to help developers fully utilize the architectural innovations in the Intel Itanium and Pentium 4 processors, allowing for easy access to all of the performance features of Intel’s latest processors.” Our Take: The big news is not just to dryily report this release for Linux. The real news here is that Linux can now have an incredibly good compiler. Sharing a house with 4 ex-Be engineers in the past taught me at least one thing: the Intel compilers are many times faster (in generated executable code quality and compilation speed) than GCC 3.X and even VC++. While GCC is free and Intel’s Proton costs $399, still, for professional use or for apps where execution speed matters (games or scientific software for example), Proton and VTune seems like the clear choice.
Intel Launches New Compilers for Linux
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Ooops…
I changed some stuff last night in the commenting section and I did not realise that people could not comment on new stories because of a (stupid) bug in my code…
Fixed now though. You can now freely flame me for my opinion about gcc!
Even if it is faster… 400 dollars for a compiler?! Ouch!
I guess, when you are getting 25-30% faster than the rest of the compilers (because we are talking about such an improvement), I think it is ok to buy it… Think you are John Carmack for a moment (but don’t think of it too much
. Compiling with Proton will almost guarantee that your game will have a minimum requirement, let’s say, PII-233 instead of PII-300. That is already a huge step for a larger market. The money you paid for the compiler, will come back to you later.
Also, in a scientific software where having the job done FAST is very important, it may save lots of time (for example, analyzing a blood stream or another chemical reaction) which may be very important for the whole operation.
For users or amateur developers, Gcc may be good enough, but for professional stuff, Proton is unbeatable.
30%? Wow. Hmm, I see your point though. Some times it really is important to get the job done fast. If your developing comercial software $400 really isn’t that much.
$400 is really cheap. That’s quite less than one day worth of work by a software engineer. And it’s not unusual for an engineer to spend a day on a single routine to try to optimize it and not even make it 30% faster.
Check out rule number 9 of http://joel.editthispage.com/stories/storyReader$180“>The . While you’re at it, check all the other rules too…
–jbq
How much VisualC++ costs? I mean, the compiler-only, not the whole Visual Studio interface/package. Does Microsoft sells their compiler individually?
Actually, VC++ is usually based on a the Intel compiler available at the time of its release. The reason Intel released their compiler on its own is because MS hasn’t released a new VC++ version in over a year, since they’re too busy working on Visual C++.NET . That will probably include the new compiler.
Compaq C and C++ can provide significant performance improvements in applications previously built for Linux Alpha. In some floating-point intensive applications, for example, improvements of a factor of two or more have been seen, while more modest improvements on the order of 15% to 30% are more often seen for integer codes.
Since march 2000 and free (but) only for Alpha
Maybe a little off-topic, but is there any realistic hope that gcc will catch up evenutally? Is gcc 3.0 significantly better optimized than previous versions? I had heard a year or so ago that Intel was providing some engineering help to the gcc/egcs project, and it would be great if someday we had a free compiler with real state-of-the-art performance.
gcc 3.0 is a big improvement over previous CPUs. I have had opportunitites to disassemble snippets of code generated by different variants of gcc 2.9x and by a version of gcc 3.0, and the code itself looked much better. I’ve never run any benchmark. For a really good comparison, I’d like to see it use to compile a specint2000/specfp2000 suite for pentium3 and pentium4.
A few years ago, I had had the opportunity to see a detailed study of the code quality and speed of Be’s gcc (as used in R4) versus the versions of MSVC and proton that were available at the time. Proton won on almost every test (MSVC was able to be slightly better in a few occastions), not really by pulling any magic trick, but by almost never violating any optimization rule.
–jbq