[trans-mem] ipa pass for tm function cloning
Diego Novillo
dnovillo@google.com
Tue Nov 11 23:13:00 GMT 2008
2008/11/11 Jan Hubicka <hubicka@ucw.cz>:
> There is overall whopr document; but no documentatio on the current
> implementation (yet). I will try to put something together, wonder
> where is the best place to put it to then.
doc/cgraph.texi seems like a good place (it doesn't exist yet). We
are overhauling the IPA harness in the LTO branch, so we can add new
information there.
>> Because really, ignoring the LTO bits, I can't see that any of the
>> existing IPA passes should be anything but simple_ipa_passes. That
>> they aren't merely complicates understanding of what's going on in
>> them.
>
> Yep, without LTO there is no much value of IPA passes, except that it
> prevents inliner from blowing up peak memory use by too much of
> inlining.
Perhaps, but we may simplify our mental model if we amalgamated simple
and regular IPA passes. A regular IPA pass simply has additional
callbacks to read/write its summary. A simple IPA pass would simply
ignore those.
Diego.
More information about the Gcc-patches
mailing list