Re: question on control dependence

dkchen@sp91.csrd.uiuc.edu (Ding-Kai Chen)Wed, 16 Dec 1992 20:04:36 GMT

From comp.compilers

Related articles
[2 earlier articles]
Re: question on control dependence cliffc@rice.edu (1992-12-15)
Re: question on control dependence preston@dawn.cs.rice.edu (1992-12-15)
Re: question on control dependence preston@dawn.cs.rice.edu (1992-12-15)
Re: question on control dependence paco@cs.rice.edu (Paul Havlak) (1992-12-15)
Re: question on control dependence bwilson@shasta.stanford.edu (1992-12-15)
Re: question on control dependence paco@cs.rice.edu (1992-12-16)
Re: question on control dependence dkchen@sp91.csrd.uiuc.edu (1992-12-16)
| List of all articles for this month |

 Newsgroups: comp.compilers From: dkchen@sp91.csrd.uiuc.edu (Ding-Kai Chen) Organization: UIUC Center for Supercomputing Research and Development Date: Wed, 16 Dec 1992 20:04:36 GMT References: 92-12-056 92-12-070 Keywords: design

preston@dawn.cs.rice.edu (Preston Briggs) writes:
>The general case of more than 2 succesoors is handled like this...

> CD(X) = union(P(S)) - intersection(P(S)), for S in successors(X)

>(This is a much simpler version of another answer I posted earlier)

This can be simplified (also with less computation) further:

CD(X) = union(P(S)) - P(X)

I think it is not difficut to prove that

{intersection(P(S)), for S in successors(X)}=P(X)

Similarly,

CD(X,L) = P(X_L) - P(X)

where CD(X,L) is the set of nodes control depend on X with label L and X_L
is the successor of X with label L.
Ding-Kai Chen

--
Ding-Kai Chen University of Illinois
(217)244-0046 Center for Supercomputing R&D,
dkchen@csrd.uiuc.edu 465 CSRL 1308 W. Main St. Urbana, IL 61801

--

Post a followup to this message