one more note from conference with Kahan and Thomas

[David G. Hough on validgh]

In passing, Kahan commented that one reason for planning to put
quadruple precision in hardware eventually, as envisioned in the
Spectrum==PA-risc and SPARC V8 instruction set architectures,
is to permit grossly unstable algorithms, that parallelize well,
to work adequately most of the time, by pushing the roundoff
level so far down that it doesn't have enough time to cause
trouble for the size of problems likely in the next few years.