stack variable liveness analysis
Daniel Berlin
dberlin@dberlin.org
Mon Aug 24 06:29:00 GMT 2009
>> See struct df_problem in df.h
>
> ok. I am actually aware of this -- just did not see any use of it on
> gimple level.
Since roughly all the analysis is done on SSA, iterative dataflow
solvers are rare. I think PRE's compute_antic might be one of the
only ones. It would not be easy to shoehorn that one into DF for a
variety of reasons (i've tried a few times).
>>
>>
>>> The effort of
>>> adapting into existing DF framework seems much larger than having the
>>> private little solver which has very little overhead, but I could be
>>> wrong.
>>
>> Honestly, i think you are wrong ;)
>> The problem is a lot of people felt this way, and we ended up with 30
>> little solvers here and there in GCC to maintain.
>
> I would expect an answer like this from the author of DF:)
DF has been rewritten almost completely since the last time i contributed to it.
In fact, my dataflow solver code was replaced by something else that
was faster ;)
> I guess
> you are right from the long term maintenance point of view. Will
> convert to use existing DF, IF this patch is deemed useful.
Sure.
More information about the Gcc-patches
mailing list