PATCH for CSE speedup

[Jeffrey A Law]

  In message < 19990329144042.E7798@cygnus.com >you write:
  > On Sun, Mar 28, 1999 at 11:42:12PM -0700, Jeffrey A Law wrote:
  > > I'll try to take a good look at this tomorrow.  I've also got some codes 
  > which
  > > compile horribly slowly due to cse1/cse2 that I'll try with the new code 
  > :-)
  > 
  > It doesn't help Plum Hall testing -O3 on Alpha. 
  > 
  > Therein, cse1 allocates 187MB worth of rtl, trying to see if it might
  > help.  Then cse2 does the same and we've got 313MB to lug about through
  > the rest of the function compilation.
  > 
  > But I agree that this is very worth while nontheless.
I suspect some of this is due to the alias analysis stuff.  I'm trying to
dream up a better way to do this since more and more passes use this stuff
some of the things we do get painful.

Some of it will be done like sched manages it's dependency lists (remember the
haifa memory explosion you fixed a while back?)

The other issue is we keep canonicalizing certain hunks of rtl in the alias
code when we should canon it once and cache the value.  Similarly when we
try to find the base value and other items.

This is (in my mind) some of the generic alias cleanup that Andrew will be
starting on soon.




jeff