Re: Register allocation patent (William Schmidt)
Thu, 30 Nov 1995 20:23:58 GMT

          From comp.compilers

Related articles
Register allocation patent (1995-11-27)
Re: Register allocation patent (1995-11-28)
Re: Register allocation patent (1995-11-29)
Re: Register allocation patent (1995-11-29)
Re: Register allocation patent (1995-11-30)
Register allocation patent (1995-12-09)
| List of all articles for this month |

Newsgroups: comp.compilers
From: (William Schmidt)
Keywords: registers, optimize, legal
Organization: IBM Rochester MN
References: 95-11-214
Date: Thu, 30 Nov 1995 20:23:58 GMT (Stephen Clarke) writes:
|> What is the status of IBM's patent on register allocation and spilling
|> via graph coloring? Most people seem to reference Chaitin's work when
|> describing their register allocators, including numerous commercial
|> companies; is the patent unenforceable?

No, the patent is in full force. The state of intellectual property in the
computer world is that almost all of the major companies have signed
cross-license agreements with one another, allowing each to use the other's
patented intellectual property. If a company markets a compiler that
uses Chaitin's technique, and that company does not have an agreement with
IBM to use the technique, this is a violation of U.S. patent law.

Incidentally, thanks to Preston Briggs, Rice University owns a dependent
patent covering Briggs's optimistic coloring heuristics (modifications to the
Chaitin technique). A company using his extensions needs to have an agreement
in place with Rice.

I know all this because I accidentally ran across the Rice patent a year ago,
and had to spend some time with IP attorneys determining whether we had the
appropriate agreement with Rice. (We did; the optimistic coloring work was
done partially with funding support from IBM.)

William Schmidt (507) 253-0852 t/l 553-0852
GZA/030-2 (internal)
IBM Rochester (external)

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.