Re: Patch to allow Ada to work with tree-ssa

[Mark Mitchell <mark at codesourcery dot com>]

Richard Kenner wrote:

>    You also managed to add another quadratic algorithm somewhere,
>    see gcc.gnu.org/PR16163.  Can you please fix this asap.
>
> Well, it's only quadratic due to the way the testcase is constructed.  And
> the reason is due to something that I already know about and will be fixed by
> a change I plan to make soon anyway.
>
> The issue here is that the tree walker didn't used to look at the fields of
> types.  So it missed variables in DECL_FIELD_OFFSET, for example.  Now that
> it does, the cost of walking through a type is proportional to the number of
> fields in the type.
>  
If that's the only reason you're doing this walk, then you should just 
have a tag on types that indicates whether or not they have any fields 
with run-time offsets.    There's no reason to be penalizing every 
standard C and C++ program for a feature used rarely in GNU C, never in 
GNU C++, and never in ISO C.

Please fix this ASAP. 

I consider this a major regression. 

Is there another global write person who will support me in reverting 
this patch, if 48 hours go by and this problem has not been solved?

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com