On 07/09/2015 08:03 AM, mliska wrote:
Hello.
It's going to be almost a year Richard advised me to utilize IPA ICF
infrastructure in tree-ssa-tail-merge, currently using value numbering which
is quite hard to maintain.
Following small patch set is kick-off and I am opened for advices. Meanwhile,
I'm going to send statistics about merged basic blocks before/after the patch
set.
Patch set can bootstrap on x86_64-linux-pc and survives regression tests.
Thanks for tackling this -- I never mentioned the tail merging code
explicitly, I'm glad Richi did.
Re-using parts of the infrastructure you built for ICF for anything that
wants to be proving equivalency of gimple code (such as tail merging)
has been in the back of my mind as soon as you posted the initial ICF code.
jeff
