[fortran-dev] constructor work, part I; updated

[Jerry DeLisle]

Steve Kargl wrote:
> On Thu, Dec 24, 2009 at 03:23:09AM +0100, Daniel Franke wrote:
>> On Wednesday 23 December 2009 17:37:01 I wrote:
>>> To see if it actually works as intended, arith.c, decl.c and expr.c were
>>> already transformed to use the new API. As some inserts on tail-pointers
>>> were replaced by list traversals from begining to end, compile-time
>>> performance degraded a bit (testcase gfortran.dg/initialization_21.f90
>>> times out).
>> I spent some more time on this, revamped the iterator interface again and 
>> converted all files but data.c - attached patch contains the accumulated 
>> changes.
>>
>> In the next step, I'll also convert data.c and remove the con_by_offset member 
>> from gfc_expr.
>>
>> Afterwards, drop the list implementation and move to a splay-tree.
>>
>> Regression tested on i686-pc-linux-gnu. Still one performance regression.
>> Ok for fortran-dev?
>>
> 
> Daniel,
> 
> I haven't had time to read the patch, yet.  I believe
> that this type of patch belongs fortran-dev, but can
> you check with Janus to ensure that conflicts with
> his (and pault's) OOP work do not occur.
> 
> PS: I'll read the patch over in the next few days.
> 
I am also reviewing as discussed on IRC.  This version of the patch still has 
the performance regression even with my patch to PR20923.  I am going to have to 
wait a few days before I can investigate it further.

Steve, the plan has been to go to fortran-dev.  If there is an issue with OOP 
cross-talk I think we can hold this in our local trees until its ready to kick 
into 4.6.  This is great work by Daniel and I hope by the time he is done, we 
will have a great basis for significant improvements.

Regards,

Jerry