Re: Question on scalar replacement

[Diego Novillo <dnovillo at redhat dot com>]

On Sun, 2004-06-13 at 10:56, Richard Kenner wrote:

> Scalar replacements:
> T.47.P_BOUNDS -> SR.196
> FRAME.64.M12b -> SR.198
> T.47.P_ARRAY -> SR.195
> 
> The first and last seem OK, but there is no assignment to SR.198 in the
> .t32 file but instead, the assignment
> 
> <L10>:;
>   FRAME.64.M12b = system__secondary_stack__ss_mark ();
> 
> from the .t31 file is still present.
> 
Test case?  It's as if the main loop in scalarize structures didn't find
any virtual operands on that store.  Does -fdump-tree-sra-vops show
virtual operands in that statement?  There should be a V_MAY_DEF to
FRAME.64.

> There is no documenation in the front of tree-sra.c
>
Every function is documented.  But I see that tree-sra.c:tree_sra
contains some outdated stuff.  I'll update it.

> (again, I thought that
> such documentation was supposed to be a required criteria for the tree-ssa
> merge)
>
And I thought that my repeated calls for documentation review had been
answered.  There is a documentation PR for tree-ssa (13756) to which you
never seem to have contributed.  We did our best effort to have all the
internals documented before the merge.  Feedback we received via mail
and/or PR13756 were replied to.  Some things fell through the cracks and
will be fixed as we move along.

I thank you for pointing out missing documentation, but if you saw that
a bunch of documentation was missing, why didn't you stop the merge? 
Why didn't you review the documentation before the merge?  It's now
post-merge time, so anything that is missing will just have to be
fixed.  No need to be whiny about it.


> Also, when is the missing documentation supposed to be added?
>
When we discover that is missing, of course.  If you let me know what
you find difficult to understand in tree-sra.c, I'll happily add more
explanations to the code.


Diego.