# US 2022/0035760 A1 · Processing accelerator architectures

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

- Number: US 2022/0035760 A1
- Document ID: US20220035760A1
- Status: Published application
- Kind: A1
- Application number: 16/945,619
- Filed: 2020-07-31
- Published: 2022-02-03
- Earliest claimed: 2020-07-31
- Assignees: Alibaba Group Holding Limited
- Applicants: Alibaba Group Holding Limited
- Inventors: Jilan LIN; Dimin NIU; Shuangchen LI; Hongzhong ZHENG; Yuan XIE
- Web page: https://patentgrep.com/patent/US20220035760A1

## CPC

- G06F 13/4027: {using bus bridges (G06F13/4022 takes precedence)}
- G06F 13/1684: {using multiple buses}
- G06F 13/4282: {on a serial bus, e.g. I2C bus, SPI bus (on daisy chain buses G06F13/4247)}
- G06F 13/1657: {Access to multiple memories}
- G06F 2213/0026: PCI express
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A processing element/unit can include a plurality of networks, a plurality of cores, crossbar interconnects, a plurality of memory controllers and local memory on an integrated circuit (IC) chip. The plurality of cores can be coupled together by the plurality of networks on chip. The crossbar interconnects can couple the networks of cores to the plurality of memory controllers. The plurality of memory controllers can be configured to access data stored in off-chip memory. The local memory can be configured to cache portions of the accessed data. The local memory can be directly accessible by the network of processing cores, or can be distributed across the plurality of memory controllers. The memory controllers can be narrow channel (NC) memory controllers having widths of 4, 8, 12, 16 or a multiple of 4 bits.

## Claims

### Claim 1

A processing element/unit comprising:
one or more networks of a plurality of processor cores;
a local memory configured to cache data for use by the plurality of processor cores, wherein the local memory is shared by the plurality of processor cores such that any of the plurality of processor cores can access any of the local memory;
crossbar interconnections configured to couple the one or more networks of the plurality of processor cores to a plurality of memory controllers; and
the plurality of memory controllers configured to access data through a plurality of narrow memory channel interfaces implemented by a plurality of portions of a standard memory channel.

### Claim 2 (depends on 1)

The processing element/unit of claim 1, wherein the local memory is coupled to the one or more networks of the plurality of processor cores.

### Claim 3 (depends on 1)

The processing element/unit of claim 1, wherein the local memory is distributed across the plurality of memory controllers.

### Claim 4 (depends on 1)

The processing element/unit of claim 1, wherein the local memory comprises static random-access memory.

### Claim 5

(canceled)

### Claim 6 (depends on 1)

The processing element/unit of claim 1, wherein the plurality of memory controllers comprise a plurality of narrow memory (NC) channel dynamic random-access memory controllers.

### Claim 7 (depends on 1)

The processing element/unit of claim 1, wherein the one or more networks of the plurality of processor cores comprise one or more mesh networks, one or more tour networks, one or more octagon networks, one or more tree networks, one or more ring networks and one or more star networks.

### Claim 8 (depends on 1)

The processing element/unit of claim 1, further comprising:
one or more communication interfaces configured to couple the one or more networks of the plurality of processor cores to a host processor.

### Claim 9 (depends on 8)

The processing element/unit of claim 8, wherein the crossbar interconnects couple the one or more communication interfaces to the one or more networks of the plurality of processor cores.

### Claim 10 (depends on 8)

The processing element/unit of claim 8, wherein the one or more communication interfaces comprise one or more peripheral component interface express (PCIe) interfaces or one or more compute express link (CXL) interfaces.

### Claim 11 (depends on 1)

The processing element/unit of claim 1, wherein the plurality of processor cores comprise a plurality of reduced instruction set (RISC-V) cores or a plurality of advanced RISC machine (ARM) cores.

### Claim 12 (depends on 1)

The processing element/unit of claim 1, wherein the local memory comprises static random-access memory (SRAM).

### Claim 13 (depends on 1)

The processing element/unit of claim 1, wherein the plurality of memory controllers are configured to access dynamic random-access memory (DRAM).

### Claim 14 (depends on 1)

The processing element/unit of claim 1, wherein the one or more networks of the plurality of processor cores, the local memory, the crossbar interconnections and the plurality of memory controllers comprise an integrated circuit (IC) chip.

### Claim 15

A processing element/unit comprising:
a plurality of networks;
a plurality of processor cores coupled together by the plurality of networks;
a plurality of narrow channel (NC) memory controllers configured to access data stored in off-chip memory through a plurality of portions of a standard memory channel;
local memory coupled to the plurality of narrow channel (NC) memory controllers accessible to all of the plurality of cores, wherein the local memory is configured to cache a portion of the data; and
crossbar interconnects configured to couple the plurality of networks to the plurality of narrow channel (NC) memory controllers.

### Claim 16 (depends on 15)

The processing element/unit of claim 15, wherein the plurality of networks comprises:
respective bi-directional communication channels;
a plurality of communication channel interfaces configured to couple the plurality of processor cores to corresponding bi-directional communication channels and the corresponding bi-directional communication channels to corresponding network controllers; and
network buffers configured to couple the corresponding network controllers to the crossbar interconnects.

### Claim 17 (depends on 16)

The processing element/unit of claim 16, wherein the plurality of networks comprise a plurality of ring networks.

### Claim 18 (depends on 17)

The processing element/unit of claim 17, wherein respective sets of processor cores include groups of processor cores coupled to respective ones of the plurality of ring network.

### Claim 19 (depends on 17)

The processing element/unit of claim 17, further comprising:
one or more communication interfaces configured to couple the plurality of networks of the plurality of processor cores to a host processor.

### Claim 20 (depends on 19)

The processing element/unit of claim 19, wherein the one or more communication interfaces comprise one or more peripheral component interface express (PCIe) interfaces or one or more compute express link (CXL) interfaces.

### Claim 21 (depends on 15)

The processing element/unit of claim 15, wherein the plurality of processor cores comprise a plurality of reduced instruction set (RISC-V) cores or a plurality of advanced RISC machine (ARM) cores.

### Claim 22 (depends on 15)

The processing element/unit of claim 15, wherein the plurality of narrow channel (NC) memory controllers include multiple respective dedicated data buses and a shared command bus configured to couple to the off-chip memory.

### Claim 23 (depends on 15)

The processing element/unit of claim 15, wherein the plurality of narrow channel (NC) memory controllers and the off-chip memory are configured in sets of the plurality of narrow channel (NC) memory controller and local memory.
