This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: 3.2 vs 3.1 speed


   From: "Vladimir N. Makarov" <vmakarov@redhat.com>
   Date: Wed, 05 Jun 2002 23:41:42 -0400
   
     Genattrtab makes some optimizations in attribute evaluation.   I don't
   remember how it behaves for cpu.  But I can guess it is evaluated every
   time when the attribute value is needed.
   
I understand now.

So one improvement we could still make is to somehow mark attributes
as having fixed values after compiler command line arguments have
been evaluated.  Then, we can walk all reservations and determine
which reservations and state machines are still valid.

>From your description, once an insn is evaluated for DFA and we know
the reservation which applies, we cache that result.  That is the
premise which I use to suggest the above improvement.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]