This is the mail archive of the
mailing list for the GCC project.
Re: [tcb] Incremental SSA updates
Zdenek Dvorak wrote:
OK, that would be interesting to have, if you can produce them. Thanks.
because we do not do many of them (usually one or two per optimization
pass, at most). I definitely can come up with examples where ignoring
this during incremental ssa updates would cause quadratic behavior.
Oh, we do a pre-pass scan in prepare_blocks_for_update. Your approach
uses DU chains, which saves you from doing this. That's what I'm mostly
interested in testing. I'll let you know as soon as I have something
useful to report.
How do you know the block does not contain uses of affected names,
without scanning it?