Re: Why C is much slower than Fortran

"Peter Mayne" <Peter.Mayne@compaq.com> writes:
> Greg Lindahl <lindahl@pbm.com> wrote
> > how many hardware architectures have hardware bounds checking?
>
> VAX for one.


And the MC68010 and x86. But, to belabor the obvious, didn't the 801
people write a paper about how you can do software bounds checking for
a rather small cost if you do some optimization of the bounds checks?


As a result, what *modern* hardware architecture has hardware bounds
checking? Zip zilch none.


Software alias analysis of subroutine arguments is even cheaper than
bounds checking. C compilers already have to figure out which
statements contain potential alias problems. In some cases, it's
obvious that it's an argument alias problem.


-- g

Return to the comp.compilers page.
Search the comp.compilers archives again.