This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [dataflow] [RFC] Remove many almost useless DCE passes
Paolo Bonzini wrote:
I am just trying to compare apples with apples. If you compare
compiler speed I think you should have the same preformance for the
comparison. It might be that removing delete_trivially_dead_insns in
the same places and putting it on the analogous place as
pass_rtl_fast_dce (although it is more powerful) on the mainline
makes the mainline compiler to generate the same quality code as on
the branch after your patch and making the compiler on the mainline
faster too.
My patch was not meant to be committed in the first place, just to get
raw numbers about how much time is actually spent in DCE. Otherwise,
I would have did SPEC myself, and possibly assembly output comparisons.
Paolo, I am sorry. I really don't want you to waste your time and do
this comparison. I really appreciate what you are trying to do to
improve the df-branch.