This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: schedule_block
- To: law at cygnus dot com
- Subject: Re: schedule_block
- From: Michael Hayes <m dot hayes at elec dot canterbury dot ac dot nz>
- Date: Wed, 16 Dec 1998 13:00:26 +1300 (NZDT)
- Cc: Michael Hayes <m dot hayes at elec dot canterbury dot ac dot nz>, meissner at cygnus dot com, egcs at cygnus dot com
- References: <"13942.56824.799761.256729"@ongaonga.elec.canterbury.ac.nz><25069.913763947@hurl.cygnus.com>
Jeffrey A Law writes:
> You could do something with MD_* macros, though I would recommend
> against it.
Moving to the haifa scheduler is on my to do list...
> I'm also not sure it is safe for the c4x to enable scheduling before
> allocation. You seem to have classes for pairs of regs which also show
> up explicitly in the RTL. See recent thread about %eax usage on the
> x86.
Yes, I've been following this thread. Fortunately, life is not as
grim for the c4x compared with the x86. One rarely used class does
overlap with registers used for function arguments but the souped up
reload takes this in its stride.
Scheduling before register allocation on the c4x may be of dubious
benefit anyway---I need to run some more tests. The situation may be
different if egcs could schedule on the basis of memory space (for
example, internal memory vs. external memory).
Michael.