Method and apparatus for performing numerical calculations

A computing device comprising a stream processor operable to process an iterative computation, the stream processor being operable to perform calculations in a fixed point numerical format having a scaling factor that is selectively variable for different steps of the calculation in order to prevent overflow and to reduce underflow, the computing device to:

• input a plurality of arrays of data to the stream processor;
• scale each array to a common scaling factor selected to prevent overflow and to reduce underflow;
• cause the stream processor to calculate a step of a numerical computation in a fixed point format using the scaled array data;
• apply, using the stream processor, a predetermined change to the scaling factor with each step of the iterative calculation, the value of the predetermined change being the same for each step of the iterative calculation;
• update, using a processor, the scaling factor based upon the resulting values of the stream processor calculation;
• and
• repeat the input, scale, cause, apply, and update steps for the next step of the numerical computation using the updated scaling factor.

A computing device comprising:

• a central processing unit;
• and
• a further processor, the further processor being operable to perform numerical computations comprising a plurality of stepwise iterations in a fixed point numerical format having a scaling factor that is selectively variable in order to prevent overflow and to reduce underflow, and to apply a predetermined change to the scaling factor of the data during each step of the iterative computation, the value of the applied predetermined change being the same for each step of the iterative calculation, the numerical computations being performed on a plurality of arrays of data, each array being scaled to a common scaling factor during each step of the iterative calculation;
• and
• the central processing unit being operable to update the scaling factor for the next step based at least in part upon the resulting values of a previous step during each step of the iterative computation.

A method of performing an iterative computation on a computing device comprising a central processing unit and a further processor, the method comprising:

• performing, on the further processor, numerical computations comprising a plurality of stepwise iterations in a fixed point numerical format having a scaling factor that is selectively variable in order to prevent overflow and to reduce underflow, the numerical computations being performed on a plurality of arrays of data, each array being scaled to a common scaling factor during a given each step of the iterative computation;
• applying, using the further processor, a predetermined change to the scaling factor of the data during each step of the iterative computation, the value of the applied predetermined change being the same for each step of the iterative calculation;
• and
• updating, on the central processing unit, the scaling factor for the next step based upon the resulting values of a previous step during each step of the iterative computation.

A method of processing an iterative computation on a computing device comprising at least one processor by performing, on a stream processor, an iterative calculation on data in a fixed point numerical format having a scaling factor that is selectively variable for different steps of the calculation in order to prevent overflow and to reduce underflow, the method comprising:

• inputting data to the stream processor;
• scaling the inputted data using a scaling factor selected to prevent overflow and to reduce underflow;
• calculating, by the stream processor, a step of a numerical computation in a fixed point format using the scaled data;
• applying, using the stream processor, a predetermined change to the scaling factor of the data, the value of the applied predetermined change being the same for each step of the iterative calculation;
• updating, using a processor, the scaling factor for the next step based upon the resulting values of the calculation step;
• repeating the steps of inputting, scaling, calculating, applying, and updating for the next step of the numerical computation using the updated scaling factor.

A computing device comprising a stream processor operable to process an iterative computation on a computing device, the stream processor being operable to perform calculations in a fixed point numerical format having a scaling factor that is selectively variable for different steps of the calculation in order to prevent overflow and to reduce underflow, the computing device being operable to carry out the steps of:

• inputting data to the stream processor;
• scaling the inputted data using a scaling factor selected to prevent overflow and to reduce underflow;
• calculating, on the stream processor, a step of a numerical computation in a fixed point format using the scaled data;
• applying, using the stream processor, a predetermined change to the scaling factor, the value of the applied predetermined change being the same for each step of the iterative calculation;
• updating, using a processor, the scaling factor for the next step based upon the resulting values of the calculation step;
• repeating the steps of inputting, scaling, calculating, applying, and updating for the next step of the numerical computation using the updated scaling factor.