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] Disable CSE skip-blocks

I have faith than when complete the resulting
objects will be smaller and faster.  It doesn't make much sense
to enquire about "SPEC with -fno-cse-skip-blocks" because it may
be unreasonable to expect all the intermediate steps to be monotonic
improvements.  If anything is to get done, we need to take a little
risk once in a while.  Indeed early stage 1 is when we should take
those risks.

The nice thing is that we have bugzilla to track regressions. We
can fix missed-optimization regressions from now all the way through
stage 3 of 4.3's release cycle. However, I'd expect 4.3 to benchmark
better than 4.2 once Steven is through.
Thanks Roger for expressing what Steven and I have been stating for months/years.

I think everybody by now has understood that the benefit of tree-SSA was not to replace RTL optimizers, but to make them serve better their purpose. Removing things such as ADDRESSOF, CONSTANT_P_RTX, QUEUED allows now to focus on low-level optimizations that cannot be expressed on the tree level, to understand what needs to be done on the RTL level, and to do it well.

Everything that will be removed from CSE will most likely not be done on the tree level (though while developing fwprop we did find one case or two), but will still happen at the RTL level in a less intertwined and more easy to maintain way.


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