sasaki@harvard.ARPA (Marty Sasaki) (03/16/85)
About 4 years ago I attended a meeting of the Boston VAXLUG of DECUS
where several developers of the VMS C compiler gave a presentation. They
talked about C, and how they put things into VMS, problems with the
run-time and other things. Very useful.
After the formal presentation someone asked one of these guys about the
performance of the VMS compiler versus the BSD compiler. One of the
tests that was done was to recompile the BSD kernel. The VMS compiler
produced code 20% smaller and 20% faster than the BSD compiler.
Someone said that current VMS C compilers compile procedure calls into
slower code because they conform to the VMS calling standard. I don't
know whether the procedure calls are really slower, but the VMS
compilers do not produce code that conforms to the standard. There was a
discussion of this in INFO-VAX. The version 2.0 compiler produced code
that did conform with the standard, but that broke so much code that the
latest versions are back to what 1.x did.
--
Marty Sasaki net: sasaki@harvard.{arpa,uucp}
Havard University Science Center phone: 617-495-1270
One Oxford Street
Cambridge, MA 02138