The Power of the NORC 3760

The idea was that Fortran unlike other labeling was constrainted to 99999 digits. This is a small number on most computers. In part it was boredom between benchmark development runs, and a mix of others misunderstandings about measurement. The idea is to be part of a larger package which I thought of as the equivalent of a "tape measure" workbench for performance measurement. It was hacking.

Sure that's called standardization.

The details are in the How. That's what started my involvement in this thread. Alliant (was once Dataflow Systems) attempted (and later the U. Ill. Cedar project) an interactive-computer engine distinction IPs vs. CEs. It just didn't work commercially. There is enough in common to the problems of interaction and pure compute leaving out human reaction time that you might as well use the same processing unit for both.

Now admittedly, their first IP was a MC68008 (who ever head of an 08 when 10s and 20s came out), but they did...? They needed more IP umph.

No, I am talking about really high end problems which people won't embark on even with a supercomputer and a fair number of those exist. The old metric at places like this was Cray-years not Cray-hours when most people had workstations and shared Dept. machines. Single problems mentioned in those terms (you can tell because they know how many hours are in a year to a digit or 2 significance off the top of their head, and why an hour or 10 doesn't make that much differene).

Some progress. Progress in computing is at best linear to quadratic. A lot of problems are O(n^3) and higher (if not NP-complete). Percolation theory is O(n^5).

Depend what Gates wants to constraint him on and for how much.

A lot of people.

Wearcam.org.

--