Aina Niemetz <aina.niem...@gmail.com> writes: > i'm one of the students who didn't get accepted this year, unfortunately. This > doesn't lessen my motivation to get involved, though. Thus i decided to roll > up > my sleeves and start to work on my proposed project anyway as i think it'd be > just perfect for getting familiar with the code base and the project as a > whole. > > The project I proposed addresses RTX traversals as a possible optimization > candidate for gaining some overall speedup. I plan to make the currently > recursive walks over the RTL non-recursive, which does especially affect > FOR_EACH_RTX and co. My proposal for GSoC is not public, so no link provided > here, but if you're interested i'd be happy to send it along. > > I know that Paolo Bonzini tried something similar quite some time ago and i > would appreciate his thoughts on what i propose here, a lot. Further, some > comments of the reviewers of my proposal, whoever that might be, would be > great, > too. That said, i have to add that i'd appreciate some feedback by anyone of > you > very much :)
Thanks for your continued interest. I think your proposal is a good one. I think there were two reasons that it didn't make the top ten. The first is a question of scope: it seems like a project that should take a couple of weeks, rather than the whole summer. The second is that it's not clear that it really will save compilation time; recursion is not typically the fastest algorithm, but in order to walk RTL you need to keep some form of stack. When you have a trial implementation you will have to do some measurements to confirm that it is better. Ian
