On Fri, Jul 5, 2013 at 3:29 PM, Matt Godbolt <matt@xxxxxxxxxxx> wrote: > Hi all, > > I have occasionally found -finstrument-functions very useful in > tracking performance issues. However, it is a very broad brush to > use, and without lots of judicious exclusions tends to be unwieldy. > In my use case I'd ideally want an inclusion-based approach and having > seen the relevant code (in gimplify.c), it would seem fairly > straightforward to do. [snip] To cut a long story short, I found that I had initialized my pass before gimplification had taken place, and that was the main reason I was having trouble finding the GIMPLE. I've now added by pass BEFORE the "lower" pass and all seems to do what I'd expect. Regards, Matt
