C486: Parallel Versions of Stone’s Strongly Implicit (SIP) Algorithm 

This is a solid paper that should be published. Two minor points
1) Please explain super linear speed up seen in some red black sizes
   Why does efficiency vary so drastically with size?

2) Discuss quality of red-black method. Is it -- like in Gauss Seidel
as good or better than than original sequential case? Here I mean
"good" as measured by quality of solution for a given number of iterations