On Fri, 2007-06-15 at 21:09 -0700, Brian Slesinsky wrote: > On 6/15/07, Don Dailey <[EMAIL PROTECTED]> wrote: > > > These arguments are based on the assumption that the human can't do as > > well as the computer for certain tasks. A custom designed memory > > manager by a person who knows what he is doing will be faster than an > > automatic generalized algorithm and assembler code written by a human > > who actually knows what he is doing will always be faster than C code > > written by a compiler. > > > > Of course the big variable here is how many programmers have the > > expertise to out-code the compiler. These statements are probably > > mostly true for most people and most programs. > > Well, it's not just a matter of skill. There is also the question of > what the code will look like after the assembly language expert is > done. Sure, a person can inline, unroll loops, do common > subexpression elimination, and all the other tricks that compilers can > do. But if you do that at the source code level, what do you do when > you want to change something? It's a dead end.
Yes, of course. I'm not claiming it's always desirable but I'm claiming the basic argument is a bit of a lie. - Don > (On the other hand, what could a good programmer do with an > application-specific code generator using something like LLVM?) > > - Brian > _______________________________________________ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
