Re: Shared Generic Instance Code

["Pat Rogers" <progers@acm.org>]

david scott gibson <dgibson@snoopy.cis.ohio-state.edu> wrote in article
<5hrkhkINN9ip@snoopy.cis.ohio-state.edu>...
> Hi.  Could someone summarize the advantages and disadvantages of
> having an Ada compiler that when compiling generic units generates
> code that may be shared by multiple instances?  On the negative side,
> I suspect that it increases compiler complexity and could result in
> slower executables.  On the positive side, it could reduce the size of
> executables and perhaps reduce the amount of recompilation in system
> generation.  Are there other issues such as interaction with other
> Ada language features or the presumed utility of code sharing, that
> make one approach more attractive than the other?

The nature of the generic formal parameters can impose a performance
penalty when code sharing is used.  Anything that requires the
implementation to distinguish between different instances at run-time will
cost you.  Not all parameters will do so; it isn't always the case that
code sharing imposes a penalty, but for example, passing formal subprograms
parameters may do it, as well as formal objects of mode "in out",
etc.......  There's always a trade-off.  For real-time applications, this
is yet another of those cases that require you to know what you are doing,
and is often applicable when comparing languages: the ever-popular "How
come when I recode this well-understood piece of code into ADA it is so
slow?" comes time mind.  (Yes, I put the Ada in upper case on purpose
there:)