# US 2019/0171613 A1 · Method and Apparatus for A Vector Memory Subsystem for Use with A Programmable Mixed-Radix DFT/IDFT Processor

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0171613 A1
- Document ID: US20190171613A1
- Status: Published application
- Kind: A1
- Application number: 16/272,470
- Filed: 2019-02-11
- Published: 2019-06-06
- Earliest claimed: 2015-12-31
- Assignees: Cavium, Inc.
- Applicants: Cavium, LLC
- Inventors: Yuanbin Guo; Hong Jik Kim
- Web page: https://patentgrep.com/patent/US20190171613A1

## CPC

- G06F 15/8061: {Details on data memory access}
- G06F 9/30036: {Instructions to perform operations on packed data, e.g. vector, tile or matrix operations}
- G06F 17/141: {Discrete Fourier transforms}
- G06F 17/142: {Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A vector memory subsystem for use with a programmable mix-radix vector processor (“PVP”) capable of calculating discrete Fourier transform (“DFT/IDFT”) values. In an exemplary embodiment, an apparatus includes a vector memory bank and a vector memory system (VMS) that generates input memory addresses that are used to store input data into the vector memory bank. The VMS also generates output memory addresses that are used to unload vector data from the memory banks. The input memory addresses are used to shuffle the input data in the memory bank based on a radix factorization associated with an N-point DFT, and the output memory addresses are used to unload the vector data from the memory bank to compute radix factors of the radix factorization.

## Claims

### Claim 1

A programmable device configured to provide discrete Fourier transform (“DFT”) and inverse DFT (“IDFT”) operations for data processing, comprising:
a vector data path pipeline configured to generate scaled vector data via scaling input data retrieved from a vector memory bank;
a finite state machine controller coupled to the vector data path pipeline and configured to generate radix engine control signals in accordance an input of index; and
a programmable vector mixed-radix engine coupled to the finite state machine and capable of being configured in accordance with the radix engine control signals for facilitating computation of a radix result in response the scaled vector data.

### Claim 2 (depends on 1)

The device of claim 1, wherein the vector data path pipeline includes a vector load unit, vector dynamic scaling unit, vector input staging buffer, and vector data twiddle multiplier.

### Claim 3 (depends on 2)

The device of claim 2, wherein the vector data path pipeline carries twelve data values per clock cycle.

### Claim 4 (depends on 2)

The device of claim 2, wherein the vector scaling unit scales vector data from a memory to generate scaled vector data.

### Claim 5 (depends on 2)

The device of claim 2, wherein the vector staging buffer stores the scaled vector data in a temporary memory in a first order, and outputs the scaled vector data from the temporary memory in a second order.

### Claim 6 (depends on 2)

The device of claim 2, wherein the vector twiddle multiplier multiples the scaled vector data by twiddle factors to form multiplied scaled vector data.

### Claim 7 (depends on 1)

The device of claim 1, wherein the programmable vector mixed-radix engine is configurable to perform a selected radix computation selected from a plurality of radix computations for generating the radix result.

### Claim 8 (depends on 7)

The device of claim 7, wherein the plurality of radix computations includes radix3, radix4, radix5, and radix6 computations.

### Claim 9 (depends on 1)

The device of claim 1, further comprising an output staging buffer configured to buffer one or more radix results.

### Claim 10 (depends on 1)

The device of claim 1, further comprising a vector feedback path coupled to the programmable vector mixed-radix engine and configured to pass the radix result to a vector store unit.

### Claim 11 (depends on 10)

The device of claim 10, wherein the vector store unit stores the radix result in a vector memory bank.

### Claim 12

A data processing device configured to provide discrete Fourier transform (“DFT”) and inverse DFT (“IDFT”) operations for data processing, comprising:
a vector input shuffling controller configured to receive input data from an external component;
a configuration look up table (“LUT”) coupled to the vector input shuffling controller and configured to store index values selectable by an index; and
a finite state machine controller coupled to the configuration LUT and configured to inform the vector input shuffling controller how to store the input data in a vector memory bank based on the index, the finite state machine controller configured to determine size (N) of DFT/IDFT for facilitating programming a programmable vector mixed-radix engine in accordance with the index.

### Claim 13 (depends on 12)

The device of claim 12, further includes a vector data path pipeline configured to generate scaled vector data by scaling input data received from a vector memory bank.

### Claim 14 (depends on 12)

The device of claim 12, wherein the programmable vector mixed-radix engine is capable of being configured in accordance with the radix engine control signals for facilitating producing a radix result in response the scaled vector data.

### Claim 15 (depends on 12)

The device of claim 12, further comprising a vector data path pipeline coupled to the finite state machine controller and configured to receive the vector data from a memory;

### Claim 16 (depends on 15)

The device of claim 15, wherein the vector data path pipeline carries twelve data values per clock cycle.

### Claim 17 (depends on 15)

The device of claim 15, wherein the vector data path pipeline includes a vector scaling unit that scales the vector data to generate scaled vector data.

### Claim 18

A method of providing a discrete Fourier transform (“DFT”) and inverse DFT (“IDFT”) operation for data processing, comprising:
receiving an index from an external component;
selecting one of index values representing size of DFT/IDFT stored in a configuration look up table (“LUT”) based on the index;
instructing a vector input shuffling controller how to store input data in a vector memory bank based on selected index value; and
facilitating to program a programmable vector mixed-radix engine in accordance with the index value.

### Claim 19 (depends on 18)

The method of claim 18, further comprising generating input memory addresses for storing input data into a vector memory bank and generating output memory addresses for retrieving vector data from the vector memory bank to compute radix factors of the radix factorization.

### Claim 20 (depends on 18)

The method of claim 18, further comprising staging the vector data from the memory bank and staging outputs of vector data from a temporary memory.
