Hi everyone,
I would like to propose extending the XML node dumping feature for the
compiler in order to help with development of new compiler features,
especially where node optimisation is concerned (e.g. pure functions,
vectorisation). Currently, I have added support to show some elements
of newly-defined classes and records in the unit, mostly in order to
show the byte offsets and lengths of each field (to aid vectorisation):
https://bugs.freepascal.org/view.php?id=36882
One thing I like to add is to dump the node tree as it appears after the
first pass, or at least right before it is converted into
platform-specific assembly language, because that will allow developers
to see how everything appears after simplication and node optimisation,
especially if nodes are removed. My question though... considering node
dumps generate many megabytes of data and a single procedure can produce
a massive tree, should the optimised tree appear in the same dump file
as the rest, under a different XML branch, or in a separate node dump
file to make visual side-by-side comparison with the regular node dump
easier?
Gareth aka. Kit
--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
_______________________________________________
fpc-devel maillist - fpc-devel@lists.freepascal.org
https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel
