Reconfigurable networked processing elements partial differential equations system
A method for using a system to compute a solution to a partial differential equation (PDE) broadly comprises the steps of determining the true accuracy required (TAR) to solve the PDE, determining an architecture according to the TAR that performs a plurality of calculations to solve the PDE, determining a time allowed (TA) and a time required (TR) based on the architecture to solve the PDE, rejecting the PDE if the TR is less than or equal to the TA, configuring a plurality of programmable devices with the architecture, initiating the calculations, and ceasing the calculations when an accuracy criteria is met or when the TA expires. The system broadly comprises a plurality of programmable devices, a plurality of storage elements, a device bus, a plurality of printed circuit (PC) boards, and a board to board bus.
1. A method of computing a solution to a partial differential equation (PDE) utilizing a plurality of programmable devices, the method comprising the steps of:
a) determining the true accuracy required (TAR) to solve the PDE;
b) determining an architecture according to the TAR that performs a plurality of calculations to solve the PDE;
c) determining a time allowed (TA) and a time required (TR) based on the architecture, wherein TA depends on usage limitations of a system and TR depends on a number of calculations to be performed to solve the PDE;
d) rejecting the computation of the PDE if the TR is less than or equal to the TA; and
e) accepting the computation of the PDE if the TR is greater than the TA, configuring the programmable devices with the architecture, initiating the calculations, and ceasing the calculations when an accuracy criteria is met or when the TA expires.
2. The method of claim 1 , further including the step of providing an initial PDE solution to the programmable devices.
3. The method of claim 1 , further including the step of reconfiguring the programmable devices to possess a different architecture during the calculations.
4. The method of claim 1 , wherein the TAR controls the size of a dataword used in the calculations.
5. A method of computing a solution to a partial differential equation (PDE) utilizing a plurality of programmable devices, the method comprising the steps of:
a) determining the true accuracy required (TAR) to solve the PDE;
b) determining an architecture according to the TAR that performs a plurality of calculations to solve the PDE;
c) determining a time allowed (TA) and a time required (TR) based on the architecture, wherein TA depends on usage limitations of a system and TR depends on a number of calculations to be performed to solve the PDE;
d) rejecting the computation of the PDE if the TR is less than or equal to the TA; and
e) accepting the computation of the PDE if the TR is greater than the TA, configuring the programmable devices with the architecture, providing an initial PDE solution to the programmable devices, initiating the calculations, reconfiguring the programmable devices to possess a different architecture during the calculations, and ceasing the calculations when an accuracy criteria is met or when the TA expires.
6. The method of claim 5 , wherein the TAR controls the size of a dataword used in the calculations.
7. The method of claim 1 , wherein the TAR is determined by the resolution of data that is input to the programmable devices.
8. The method of claim 1 , wherein the TAR is determined by a number of bits output from an analog to digital converter that supplies data to the programmable devices.
9. The method of claim 5 , wherein the TAR is determined by the resolution of data that is input to the programmable devices.
10. The method of claim 5 , wherein the TAR is determined by a number of bits output from an analog to digital converter that supplies data to the programmable devices.