| |
 |
Current
Parallel Computers Are Inefficient.
Since the dawn of
computing, computer designers have tried to accelerate computation by
doing more than one thing at the same time. Efforts to harness parallelism
have taken many forms. The Cray-1 could perform 64 operations at the same
time inaugurating "vector" computers.
|
|
| |
|
The multiple-issue
pipelines of today's fastest microprocessors attempt to harness "instruction-level"
parallelism. Attempts to use several processors on the same problem have
generally been called "parallel processors," and those machines with hundreds
or thousands of processors are called "massively-parallel processors"
or MPP. Machines with a dozen or so processors are called "lightly-parallel
processors," but never by their manufacturers.
Over the years,
those that have tried to build lightly-parallel processors have had modest
success, while those that have build massively-parallel processors have
failed miserably. The model of computation in U.S.
Patent No. 5,867,649 issued on February 2, 1999 was invented specifically
to overcome the limitations that hobbled all previous attempts to efficiently
employ large numbers of processing elements. Multitude computers perform
of computation, incorporating other devices also claimed in U.S.
Patent No. 5,867,649.
|
|
