This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Optimization for automatic objects
- From: dewar at gnat dot com (Robert Dewar)
- To: gcc at gcc dot gnu dot org, jbuck at synopsys dot com, js at jku dot at
- Date: Fri, 24 Oct 2003 00:06:23 -0400 (EDT)
- Subject: Re: Optimization for automatic objects
> We discussed 3 - 4 of these this morning during the call with Brett. I hate
> to add to this list but Dave Lemmon's recent note, [C515-011] act#5031
> Multiple load modules and single thread of execution, is one more. Some of
> our programs (SSDP and IOPM; 2 samll ones) build and run under the static
> elaboration model. However, the SSDP does not run under the dynamic elab
> model. The large ones (AOP and FMP), don't bind under the static elab model.
> They do build under the dynamic model. But they don't run under the dynamic
> model due to elab errors. Dave Lemmon stated that some ASP elaboration
> comes after some AOP elaboration with the dynamic model. We know that this
> is wrong. For the AOP and FMP, the static model is not an option at this
> time. There would be too much refactoring/reorganizing of code. So we are
> caught in a catch-22 situation with elaboration problems.
I don't really see any catch-22 situation here. You are dealing with a large
body of non-portable legacy Ada code that needs to be modified to move to a
new compiler. This is normal in porting legacy code. It is true that you can
expect to port Ada 83 code to an Ada 95 compiler with no or minimal changes,
but only if the code is portable in the first place.
Yes, indeed making code portable with respect to elaboration issues can be
hard work if the original code was not written carefully to obey the requirements
for portable code in this respect.
> (3) Debugger issues ([C924-008] ACT #5031 GVD questions)
This is currently on hold, waiting for input from you (you might want to use
GNAT Tracker to see the status of your various reports). Ramon sent a message
on Oct 20th, and is waiting for a reply ...