Re: [tree-ssa] lower complex operations

[law at redhat dot com]

In message <20040114083937.GA16912@twiddle.net>, Richard Henderson writes:
 >Thinking about this while testing was underway, I'm wondering if it
 >might be best, actually, to do this lowering from directly from dom1.
 >This would reuse the value number that we've already done, as well
 >as avoid yet another pass over the function.  Thoughts, Jeff?
 >
 >Anyway, bootstrapped and tested on i686-linux.
I don't particularly like the idea of mixing lowering with the
dominator optimizer either.

However, you can probably get most of the benefit by doing your
lowering via a dominator walk with a trivial amount of optimization
built-in at the same time.

In fact, the dominator walker was built with that kind of usage in
mind.  If you give me a better idea of what you want to do I could
probably give you some skeleton code to get you going.

jeff