From: | David Chase <chase@naturalbridge.com> |
Newsgroups: | comp.compilers |
Date: | 4 Aug 1998 23:22:51 -0400 |
Organization: | NaturalBridge LLC |
References: | 98-08-011 98-08-023 98-08-031 |
Keywords: | C, C++, performance |
Mayan Moudgill wrote:
> My comment: I know (and probably agree with) the trend; it means that
> in practice, C compilers are being restricted to the level of the
> other compilers. It would be interesting to see what would happen if
> somebody said - lets write _just_ a C compiler, and throw the same
> kinds of resources at that.
No, you have this backwards. The Sun compiler was a C/Fortran
compiler first, and having Fortran in the mix meant the compiler
did more, not less, since by the time I left it Sun it was loaded
with all sorts of dependence analysis and auto-parallelization
goodies. Sometimes, at high levels of optimization on the right
code, it would attempt to vectorize (software pipeline) loops
written in C, just like Fortran. At least one engineer tried the
experiment of telling the compiler to attempt to autoparallelize
some C code (it didn't run that well, because the aliasing information
from C is generally too conservative to be useful). Remember,
Fortran compilers are generally better than C compilers, not
worse. Late in the game, C++ was also added to that compiler, but
relatively little had to change to include it.
David Chase
NaturalBridge LLC
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.