Return to the
comp.compilers page.
Search the
comp.compilers archives again.
| Related articles |
|---|
| reducible loops preston@cs.rice.edu (1992-02-24) |
| Re: reducible loops rockwell@socrates.umd.edu (Raul Deluth Miller-Rockwell) (1992-02-25) |
| Re: reducible loops preston@dawn.cs.rice.edu (1992-02-26) |
| Re: reducible loops lee@dumas.rutgers.edu (1992-02-26) |
| Re: reducible loops glew@pdx007.intel.com (1992-02-26) |
| Re: reducible loops preston@dawn.cs.rice.edu (1992-02-27) |
| Duff's device glew@pdx007.intel.com (1992-02-27) |
| Re: reducible loops joel@decwrl.dec.com (1992-02-28) |
| Re: reducible loops idacrd!desj@uunet.UU.NET (1992-03-09) |
| Re: reducible loops preston@dawn.cs.rice.edu (1992-03-10) |
| Newsgroups: | comp.compilers |
| From: | glew@pdx007.intel.com (Andy Glew) |
| Keywords: | optimize |
| Organization: | Intel Corp, Hillsboro, Oregon |
| References: | <9202220021.AA28153@ucbvax.Berkeley.EDU> 92-02-122 |
| Date: | Wed, 26 Feb 1992 23:33:33 GMT |
Let me add Duff's device to the discussion about irreducible loops.
Duff's device is a technique for unrolling loops with minimal overhead.
Let's see if I can get an example right:
j = i / 8;
switch( i % 8 ) {
do {
case 7:
destptr++ = srcptr++;
case 6:
destptr++ = srcptr++;
case 5:
destptr++ = srcptr++;
case 4:
destptr++ = srcptr++;
case 3:
destptr++ = srcptr++;
case 2:
destptr++ = srcptr++;
case 1:
destptr++ = srcptr++;
case 0:
destptr++ = srcptr++;
} while( --j >= 0 );
}
If you're smart you can replace all of the pointer increments by a
single increment.
(1) Most compilers cannot generate code that uses Duff's device - because
they just don't consider irreducible loops.
(2) Most compilers cannot perform optimizations around Duff's device, because
of the irreducible loop.
--
Andy Glew, glew@ichips.intel.com
Intel Corp., M/S JF1-19, 5200 NE Elam Young Pkwy,
Hillsboro, Oregon 97124-6497
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.