Browse files

Added a note about YNF reduction, a potential infinite unfolding solu…

  • Loading branch information...
1 parent 86f8715 commit 78ee9b0b89eed379760e88046c8c4e49fb2f5e36 @luqui committed Jun 29, 2009
Showing with 15 additions and 0 deletions.
  1. +15 −0 notes/YNF.txt
@@ -0,0 +1,15 @@
+Trying to solve the "infinite unfolding" problem, as discussed
+in Thyer's "Lazy Specialization", section 6.2.
+Keep the terms normalizing, by convention or using a type system.
+Introduce a constant Y into the language, standing for the fixed
+point combinator but with different unfolding rules.
+Otherwise use HNF or complete laziness. The term "normal form"
+refers to the normal forms of those strategies.
+If F -> G then YF -> YG.
+If F is a normal form of the form λx.λy.E, then YFA -> F(YF)A
+The question is whether we lose optimization by self-fusing. Are
+we repeating any significant computations?

0 comments on commit 78ee9b0

Please sign in to comment.