Our brains are amazing machines. Whenever we form a new memory, or learn a new trick, our brain forms new connections (
synapses) between its cells. The synapses act as the storage (loosely speaking) of both the data (the memory) and the code (the technique behind the trick). Next time when we do the trick, this hard-wired information helps us perform much much better than the last time!
The way computers are currently made has a big drawback: once the microprocessors are made, the instructions inside are fixed. Software can be written to implement any code, but the software approach is almost always slower than a hardware-implemented code.
FPGA: The blocks can be switched on/off during reprogramming (Courtesy: Caltech)However, an experimental supercomputer made from hardware that can reconfigure itself to tackle different software problems is
being built by researchers in Scotland :):). The system is being built by the
EPCC (Edinburgh Parallel Computing Centre), and uses
Field Programmable Gate Array (FPGA) chips instead of conventional microprocessors. A FPGA is a device that can be reprogrammed after it is manufactured, rather than having its programming fixed during the manufacturing. Therefore, it has some definite advantages over conventional microprocessors; the programming can be changed depending upon how the FPGA will be used.
The new computer incorporates 64 FPGA processing units and will run at 1 teraflop - one trillion mathematical operations per second. This is fairly modest by modern supercomputing standards, as the fastest machines can operate at hundreds of teraflops. This seems modest, But the Edinburgh system will be up to 100 times more energy efficient than a conventional supercomputer of equivalent computing power. The 64-node FPGA machine will also need only as much space as four conventional PCs, while a normal 1 teraflop supercomputer would fill a room:).
If this is successful, researchers will try to transfer several existing supercomputer programs onto the new hardware. If they work, then the supercomputer structure can be further scaled up for higher speeds.