It was thus said that the Great Chris M once stated:
Some of the optimizations
performed by the top-
end compilers are amazing to the point of one
smacking one's forehead
and saying, "Damn, that's clever!"
This is going to sound like a stupid question, but
given it's typical applications - scientific and
engineering - why the need for such speed?
Mathematical simulations.
Back in college, I wrote a program (in C, didn't really have a Fortran
compiler for the platform) that did a horrendous number of calculations
(looking for attractors---please don't ask for more of an explanation as I
didn't even really understand what my employer was looking for, but I could
write the program to his specifications) that took a year to run. Even a
10% improvement in speed would have nocked a month off the run time [1].
Another
stupid question. Would a modernish pc port lend itself
well to writing some kind of modernish game?
Probably not. Games typically simplify the physics, and use a lot of
tricks to get "acceptable" looking results, not "accurate" results.
There's
a difference there.
-spc (Who's old enough to have taken Fortran in college ... )
[1] Last year, I did some tests, and was sad (and amazed) to see that
the same problem I did over a decade previously would have taken
maybe a week on a modern quad-core machine. Then again, had I such
equipment available to me back then, the work off the backend would
have possibly taken a year [2].
[2] The program generated images that needed to be run together into an
animation and transferred to VCS tape. The editing of the results
alone took several days of constant work.