Slashdot Mirror
Free Optimizing C++ Compiler from Microsoft
FortranDragon writes "Microsoft has made the command line toolkit for Visual C++ available for a free download. You can use the toolkit to build applications and redistribute them if you want (though you should read the EULA for the details, as always). This is a nice boon for those that have to deal with cross-platform compatibility, especially since Microsoft has tried to make Visual C++ more conformant to the ISO C++ standard. Go forth and compile your favorite OSS or FS programs today. ;-)"
MS seems to be doing a lot of this lately. It's nowhere near as open as F/OSS solutions, but it's freeing up access to what's possible with Windows far more than previously.
One of the reasons for the success of OSX is the general geek crowd's appreciation of it's *IX background, but without free dev tools that's nothing but another flavour of unix. Add the ability to dive into developing instantly and there's tens of thousands more developers working for the company.
Microsoft has spent over a decade essentially supporting only ONE processor architecture, x86. The GNU project has to worry about applying optimization to a plethora of architectures, including the quirks associated with each particular implementation.
Not Microsoft - it gets to focus on how to produce the most amount of work out of a processor of at least 80486 grade instructions. How easy is that? They get to throw a hundred developers to extract every bit of performance possible out of one processor. Every now and then make a modification to support a new supplemental instruction set (MMX, 3dnow!, SSE, etc.)
If you read their optimization whitepapers, you will notice that much of their optimization is done at the math level - nothing Win32 specific. Also, their memory optimization, loop unrolling, inlining, etc. is considered top notch by many software developers.
Ayup
I remember a certain 12 year old student who couldn't afford MASM and didn't like TASM so he wrote his own macro preprocessor in C. Then the code was fed back to Debug.com and BEHOLD it worked.
Young programmers today don't realize how spoilt they are. Back in the day we all had our copy of the Intel386 assembler document, and a heavily modified version of Abrash's Zen Timer. Have you ever seem a man shuffle assembler instructions by hand to alleviate register contention, or deliberately NOT'ing AX and reversing the following branch to lull the Pentium into correctly predicting the jump, wasting one cycle but saving 4-7 (depending on whether the code following the branch accessed RAM or not, since the prefetch would have been invalidated).
For that matter, do you even remember when we used fancy tricks to either cope with the 64kb code segment limit, or trounce all over it. I hate to say it, but I miss those days, because back then it took more than an optimizing compiler and a few www tutorials for someone to be called a programmer. You actually had to know at least a little bit what was going on under the hood. It wasn't about compilers and libraries and distributed object frameworks, it was about making a limited machine do limitless things.
-Billco, Fnarg.com