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]

Re: [tree-ssa] Patch ping


In message <1070478168.9315.2401.camel@p4>, Andrew MacLeod writes:
 >> We could improve SSA_NAME recovery if bsi_remove released them, but I'm
 >> not sure I want bsi_remove to know about this kind of stuff.  Plus bsi_remo
 >ve
 >> is used by code which wants to move a statement to a different location.
 >> Thoughts?
 >> 
 >
 >I dont think I like that, since there are places which assume bsi_remove
 >and friends only change linking,
Precisely.

 > and then do something with the tree that was part of the stmt.
Like bsi_move_{before,after}.

 >I think it ought to continue just to deal
 >with linking issues, or we might have horrible sharing bugs to track
 >down :-)
If it proves necessary what I think I'll probably end up doing is have
a routine for statement deletion which takes care of SSA_NAME recovery
IMM_USES updates, etc etc which in turn uses bsi_remove to actually
deal with the underlying linkage gunk.

I guess I'm just spoiled by PHI node management, which is easier as
they are never moved to new locations.  We create them, muck around
with their arguments and delete them.  Meaning I can recover them all
in the single deletion routine.


Jeff






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