[<<][rtl][>>][..]
Sat May 4 08:03:42 EDT 2019

Go back to RAI

RAI uses the observation that loops just introduce context = loop
variables, and arrays are just index-parameterized variables.  Single
assignment is preserved, except for accumulators.

So arrays themselves are easy: just add some annotation.  Assume that
loop indices are just scalar base types.

Accumulators are the sore point.  Can they be written as
single-assignment as well?  E.g they are just arrays, but it should be
written in such a way that there are no back-references.

What about generalizing this: an accumulator is really a "triangle".
Allow for back-references, but optimize them out in case of an
accumulator.




[Reply][About]
[<<][rtl][>>][..]