Approach for significant improvement of FFT performance in microcontrollers

A method, comprising:

• receiving N input items, by a control unit, from a memory bank;
• performing Fast Fourier Transform (FFT) computations, by the control unit, over the N input items, based on Merged radix-2 butterfly calculations;
• storing computed N items to the memory bank;
• and
• wherein the control unit is configured to use a Merged radix-2 butterfly calculation to process FFT calculations over four input items across two consecutive stages;
• further comprising, by the control unit, executing a second instruction wherein the second instruction comprises:
• performing a radix-2 butterfly calculation by reading two input items from the control unit and using a twiddle factor stored in the memory bank to generate two output butterfly values.

A processor, comprising:

• a memory bank;
• and
• a control logic coupled to the memory bank is configured to perform Fast Fourier Transform (FFT) computations over N input items using Merged radix-2 butterfly calculations;
• wherein the control unit accesses the memory bank to retrieve input items and stores the input items in registers in the control unit for the Merged radix-2 butterfly calculations;
• and
• wherein the control unit retrieves the input items, from the registers in the control unit, for the Merged radix-2 butterfly calculations.