Re: Simple benchmarks with gcc-3_0-branch

[Daniel Berlin <dberlin at redhat dot com>]

Brad Lucier <lucier@math.purdue.edu> writes:

> <Daniel Berlin writes many things about good register allocators> 
> 
> But I'm afraid that this won't help very much when all pseudo's are
> (falsely, but semi-usefully) declared dead at the beginning of these
> blocks that are the targets of abnormal edges.

>From what I remember, this was the simplest thing to do, no?

What are the other options?

I've been working on the new register allocator, and cleaning up SSA value
numbering so i can submit it, and at least the value numbering should
be in submittable shape by next week, so if it's something I can actually
understand (I don't feel like I have a grasp of all the issues
involved in abornmal edges, prevent unsafe movement across them, etc),
I'm happy to make a go at it .


I'm looking at a paper on complete removal of partially redudant
expressions (from PLDI '98), that integrates code motion and control
flow restructing (with code motion being the primary transformation,
to reduce code growth, it only uses control flow restructuring to enable
hoisting through regions that prevent the necessary code motion.).

Would *that* help?

(Excuse my stupidity if i'm misunderstanding the problem, i was up way too late
last night, and probably should be asleep right now.)

--Dan

> 
> Brad