# US 2023/0409204 A1 · Random reads using multi-port memory and on-chip memory blocks

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

- Number: US 2023/0409204 A1
- Document ID: US20230409204A1
- Status: Published application
- Kind: A1
- Application number: 18/230,117
- Filed: 2023-08-03
- Published: 2023-12-21
- Earliest claimed: 2021-02-24
- Assignees: Xilinx, Inc.
- Applicants: Xilinx, Inc.
- Inventors: Abhishek Kumar JAIN; Henri FRAISSE; Dinesh D. GAITONDE
- Web page: https://patentgrep.com/patent/US20230409204A1

## CPC

- G06F 3/0613: {in relation to throughput}
- G06F 16/9014: {hash tables}
- G06F 3/0659: {Command handling arrangements, e.g. command buffers, queues, command scheduling}
- G06F 3/0673: {Single storage device}
- G06F 16/9024: {Graphs; Linked lists (G06F16/9027 takes precedence)}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method includes receiving a value and an identifier from a first memory and hashing the identifier to produce a memory block identifier. The method also includes routing, based on the memory block identifier, a read request to a memory block of a plurality of memory blocks and updating the value received from the first memory based on a property received from the memory block in response to the read request. The memory further includes storing the updated value in the first memory.

## Claims

### Claim 1

A method comprising:
receiving a first value and an associated first identifier from a first memory;
hashing the first identifier to produce a memory block identifier for a memory block of a second memory;
routing, based on the memory block identifier, a read request to the memory block of the second memory;
receiving a property from the memory block of the second memory based on the read request;
updating the first value received from the first memory based on the received property; and
storing the updated first value in the first memory.

### Claim 2 (depends on 1)

The method of claim 1, wherein the first value is a weight of an edge in a graph and the property is a property of a source vertex of the edge.

### Claim 3 (depends on 2)

The method of claim 2, wherein the first identifier identifies the source vertex of the corresponding edge.

### Claim 4 (depends on 1)

The method of claim 1, wherein hashing the first identifier comprises performing a modulus operation on the first identifier based on a number of memory blocks in the second memory.

### Claim 5 (depends on 1)

The method of claim 1, wherein the first memory is a different type of memory than the second memory.

### Claim 6 (depends on 5)

The method of claim 5, wherein the first memory is a high bandwidth memory.

### Claim 7 (depends on 1)

The method of claim 1, further comprising buffering a second value in response to hashing a second identifier for the second value producing a second block identifier equal to the memory block identifier.

### Claim 8 (depends on 1)

The method of claim 1, further comprising:
receiving a second value and a second identifier from the first memory in parallel with receiving the first value and the first identifier;
routing, based on the second identifier, a second read request to a second memory block of the second memory while routing the read request to the memory block; and
updating the second value based on a second property received from the second memory block while updating the second value.

### Claim 9

An apparatus comprising:
a first memory;
a second memory comprising a plurality of memory blocks, wherein the second memory is separate from the first memory; and
a hardware processor communicatively coupled to the first and second memories, the hardware processor configured to:
receive a first value and an associated first identifier from the first memory;
hash the first identifier to produce a memory block identifier for selecting a memory block of the second memory;
route, based on the memory block identifier, a read request to the memory block of the second memory;
receive a property from the memory block of the second memory based on the read request;
update the first value received from the first memory based on the received property; and
store the updated first value in the first memory.

### Claim 10 (depends on 9)

The apparatus of claim 9, wherein the first value is an edge weight for an edge in a graph and the property is a vertex property of a source vertex of the edge.

### Claim 11 (depends on 10)

The apparatus of claim 10, wherein the first identifier identifies the source vertex of the corresponding edge.

### Claim 12 (depends on 9)

The apparatus of claim 9, wherein hashing the first identifier comprises performing a modulus operation on the first identifier based on a number of memory blocks in the plurality of memory blocks.

### Claim 13 (depends on 9)

The apparatus of claim 9, wherein the first memory is a different type of memory than the plurality of memory blocks.

### Claim 14 (depends on 9)

The apparatus of claim 9, wherein the first memory is a high bandwidth memory.

### Claim 15 (depends on 9)

The apparatus of claim 9, wherein the hardware processor is further configured to buffer a second value in response to hashing a second identifier for the second value producing a second block identifier equal to the memory block identifier.

### Claim 16 (depends on 9)

The apparatus of claim 9, wherein the hardware processor is further configured to:
receive a second value and a second identifier from the first memory in parallel with receiving the first value and the first identifier;
route, based on the second identifier, a second read request to a second memory block of the plurality of memory blocks while routing the read request to the second memory block; and
update a second value based on a second property received from the second memory block while updating the second value.

### Claim 17

A system comprising:
a first memory configured to store values and associated identifiers;
a second memory comprising a plurality of memory blocks, wherein the second memory is separate from the first memory; and
a hardware processor communicatively coupled to the first and second memories, the hardware processor configured to:
hash a source identifier received from the first memory to produce a memory block identifier for selecting a memory block of the second memory;
route, based on the memory block identifier, a read request to the memory block of the second memory;
receive a property from the memory block of the second memory based on the read request;
update the first value received from the first memory based on the received property; and
store the updated first value in the first memory.

### Claim 18 (depends on 17)

The system of claim 17, wherein the value is an edge weight for an edge in a graph and the property is for a source vertex of the edge.

### Claim 19 (depends on 18)

The system of claim 18, wherein the source identifier identifies the source vertex of the corresponding edge.

### Claim 20 (depends on 17)

The system of claim 17, wherein hashing the source identifier comprises performing a modulus operation on the source identifier based on a number of memory blocks in the plurality of memory blocks.
