Re: Fix thinko in cleanup_tree_cfg

[Diego Novillo <dnovillo at redhat dot com>]

On Fri, Aug 19, 2005 at 11:40:30AM -0400, Daniel Berlin wrote:
> Are you sure?
> 
> I didn't believe our goal was to make the two equal in terms of "raw
> things they catch", only in terms of "making them produce good code".
> 
> There is no point, for example, in trying to make them both catch
> exactly every jump threading opportunity, because there is no statistics
> or benchmarks (that i've seen posted) that show that it's useful.  Only
> vague "It helps redundancy elimination, etc" statements (while at the
> same time, it incredibly messes up the CFG in a lot of cases).
> 
Well, I personally have no love nor dislike for jump threading.
It has rarely affected the passes I worked on, so it just doesn't
interest me at all.  I don't have any kind of benchmarks or
statistics on its usefulness (don't remember seeing any, either).

My understanding is that whatever the tree jump threader doesn't
catch, is caught by the path following code in CSE.

As I mentioned on IRC, my only interest here was fixing the
behaviour of cleanup_tree_cfg.  My changes caused it to lie, so I
fixed that.

I'm running timings now to see what effect, if any this has on
compile time.