This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch] Lno branch merge part 3 -- ssa form updating improvements

Mark Mitchell wrote:
Well, and to say it bluntly:  we need something *now*, in order to be
able to proceed with the rest of the branch merge, and we won't have
anything else that would be sufficiently tested soon enough.

Has it been definitively decided to merge the LNO branch?

At the Summit, there were some pretty deep reservations expressed about some parts of the LNO branch. I think that there needs to be some discussion about these issues and probably SC involvement before merging in all of the code. I think that we should probably be taking advantage of Ken Zadeck's offer to talk about things as well.

I've been informed offline that my words are being subjected to some sort of Greenspan-like analysis on IRC. Are "deep reservations" worse than "severe reservations"? This is truly frightening to me; I'm sure poor Alan has to spend days planning for a three-minute press conference...

To be more specific, some people at the summit seemed concerned about the scalar evolution code, calling it "too experimental for a production compiler." There seemed to be widespread agreement that the replacements for the RTL loop optimizers were a good thing. Perhaps, after further discussion, consensus has been reached that basically the entire branch is in good shape. If so, that's fine by me!

Are there still people who are worried about the code on the LNO branch? If so, would you please articulate your concerns? Have you looked at the LNO branch in enough detail that you feel that you understand what it's doing? Is it possible

I have also been told that the June 30th Stage 2 date is causing consternation for the LNO people. My expectation here is that, while the rest of the compiler will enter Stage 2 at that time, it's not unreasonable to take some additional LNO functionality in Stage 2, if there's clear consensus on the approach, the code has been thoroughly tested, etc. I am trying to balance the desire (articulated by several at the summit) that we do releases relatively frequently with the equally reasonable desire that we get good optimization technology into the compiler as soon as reasonably feasible.

Mark Mitchell
CodeSourcery, LLC

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]