> With todays optimizations done by C/C++ compilers you would have to
> try real hard to make a piece of code a lot faster in Assembly...
Hmm, not sure about that. Sure, anyone can write bad assembler, but the
functionality of standard code is inevitably going to be generalist because
the compiler has no idea why you are doing what you are. In assembler you can
write exactly what you want it to do. C++ is always going to be slower that C
or assembly because of the increased indirection. No earth-shatteringly
slower, but slower notheless. COM is always going to be a slower method of
invoking functionality because of the run time linking. Yes modern compilers
are producing quicker code than there predecessors but it tends to be much
larger code.and if therefore slower to load, which will give the perception
of slowness.
Here is an interesting story, that parallels what I am saying:
The machine which was used to crack the German WW2 Enigma code and I think
was called Colossus, was made up largely of moving parts. Relatively recently
the same logic was ported to a compiled program on an early Pentium machine.
Astonishingly, Colossus was faster! This was because it was tailored to crack
the Enigma code and nothing else where as the Pentium is an all purpose
generalist machine. The same generalist versus specific argument applies to
assembler versus C/C++.
Absolutely, I never want to write any assembler ever again.
Love the cat....
Cheers,
Paul
