> There is no KAI C/C++ compiler for Windows. > Wrong, there have been a Kai C++ Compiler for NT: v3.3 in 1999. KAI C++ for Windows NT is a collaborative effort between KAI and Edinburgh Portable Compilers (EPC). It is a command line compiler and was tested in the Cygwin environment. Unlike KAI C++ on many other platforms, this version for Windows NT generates native code rather than intermediate C code that is later compiled by a backend C compiler,"
Kai also released a pure C source optimizer. It takes a source and outputs a source. > KAI is the best C++ compiler all around. Intel C++ is quite far from > KAI in terms of compliance with the ISO C++ standard. > Intel C++ is intended to compile MSVC projects. And MSVC is not really an ISO compiler. > OpenMP is a vendor-independent technology that both KAI and Intel C++ > supports. > > KAI and Intel C++ possibly shares some small pieces, although I > wouldn't be surprised if they don't share nothing at all. > > Definitely they are different beasts. > Right. > And only to stay on-topic I will say that, as we all know, the best C > compiler can't compete with an experienced assembler programmer > producing efficient code. A different matter is if the long time the > programmer needs to produce the assembler code justifies the gain on > speed. I'm pretty sure we all agree on this point ;-) > Wrong again. Experienced programmers can code faster in assembly than in C code (this was my case ;-)). But now, if you intend portability, a C compiler is the only choice. I think that C compilers can produce BETTER code than assembly programmers in certain cases. For example, the Intel C++ compiler does optimize all branches, and this is really impressive, since it almost counts every cycle. All these optimizations take a lot of time for a programmer, and they are almost always neglected. Also, some parts can be pretty optimized. I'm thinking about a marvellous tool called SuperOptimizer (or Gnu SuperOpt). Its aim is to produce the smallest code possible given the input list and the output list. Of course, the process was very slow, and the longest code found was 7 instructions. _________________________________________________________________________ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers