Re: [patch] handle new cgraph nodes in ipa-cp pass

[Richard Henderson <rth at redhat dot com>]

On 04/06/2010 05:49 AM, Jan Hubicka wrote:
> However in this case I guess main problem is where we want TM pass to lie in the
> WHOPR queue.  I.e. can it benefit from whole program and are we going to implement
> it in WHOPR friendly way? In that case I guess it should be IPA pass, but I guess
> current implementation is just touching the function bodies and then it should
> be SMALL_IPA pass that will be executed at compilation stage.

The TM pass can benefit from having access to the entire call graph that lies
within a __transaction block.  In principal this could be anything in the program.

In practice, it probably won't be, since the same program wouldn't compile with -O0.
Functions that are outside the current translation unit would have to be properly
marked as transaction_safe or whathaveyou.  So we probably don't gain much beyond
being a SMALL_IPA pass.

Just sort of thinking aloud here...


r~