This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[middle-end, patch 0/3] Indirect inlining cleanups


Hi,

before I left for vacation more  than four weeks ago I submitted three
cleanup patches  for indirect inlining.   They were not picked  up and
Honza's changes to  the inliner and ipa-cp meant I  had to update them
to apply to today's trunk.

I understand that we're now in  stage3 but all of these were submitted
in  stage 1  and really  make a  lot of  things a  lot  more sensible,
therefore I hope they can still get in.

The patches are:

1.  Merge of parameter describing arrays and naming cleanup
    (previous version is at
    http://gcc.gnu.org/ml/gcc-patches/2008-08/msg00060.html) 

    This patch combines various  arrays describing parameters into one
    array  of structures.   It also  does some  renaming  requested by
    Honza (the _ith_ part of  accessors functions is removed) and adds
    the unused  attribute where  it should be  used instead  of stupid
    hacks.

2.  Update ipa-prop structures after all kinds of inlining
    (previous version is at
    http://gcc.gnu.org/ml/gcc-patches/2008-08/msg00296.html)

    This patch  simplifies the  inlining code a  little bit  by making
    cgraph_mark_inline_edge invoke indirect inlining updates rather
    than mandating the caller does that immediately afterwards.  

    This  is  not  only  a  cleanup  but  actually  ensures  that  the
    structures are updated in all  cases, whereas I am not really that
    sure they are as it is now.

3.  Keep always_inlining until the cgraph stabilizes
    (previously at
    http://gcc.gnu.org/ml/gcc-patches/2008-08/msg00509.html)

    This function  keeps redoing always_inlining as long  as new edges
    are discovered  to somewhat  honor the semantics  of always_inline
    (even though obviously  taking their address is a  weird thing but
    if can inline them, let's do so).

I'm  currently bootstrapping  and testing  the patches  but  they have
passed both previously.

Thanks for any comments/reviews/approvals :-)

Martin


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]