This is the mail archive of the 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]: Code movement is a pain in the ass.

On Sunday, August 10, 2003, at 2:39 PM, Diego Novillo wrote:

On Sun, 2003-08-10 at 14:01, Daniel Berlin wrote:
On Sunday, August 10, 2003, at 1:11 PM, Diego Novillo wrote:

On Sun, 2003-08-10 at 01:42, Daniel Berlin wrote:

Anybody mind if i make bsi_remove take an argument to convince it i'm
only removing the statement temporarily (ie moving it), and then add
an argument to remove_stmt to tell it *not* to reset definitions and

That way, i can just remove it from one place, and move it to another?

Or, how about we create a bsi_move (stmt, from, to)?  'from' and 'to'
may be iterators or some structure that includes enough info to remap
the statement from the old location into the new one.

Ideally, it would still use bsi_remove and bsi_insert_after/before
internally, so we'd still need to be able to tell bsi_remove and
remove_stmt to not invalidate definitions.
Otherwise we end up copying code from remove_stmt, which seems silly
when we can just surround 20 lines of it with a if (update_annotations)
and have it take an extra argument.

I meant implementing bsi_move in terms of bsi_remove and bsi_insert. Is
that feasible?
Yes, but it will require adding an extra argument to bsi_remove and remove_stmt.


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