# US 2012/0254587 A1 · Accelerator engine commands submission over an interconnect link

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

- Number: US 2012/0254587 A1
- Document ID: US20120254587A1
- Status: Published application
- Kind: A1
- Application number: 13/077,804
- Filed: 2011-03-31
- Published: 2012-10-04
- Earliest claimed: 2011-03-31
- Assignees: International Business Machines Corporation
- Inventors: Giora Biran; Ilya Granovsky
- Web page: https://patentgrep.com/patent/US20120254587A1

## CPC

- G06F 9/3877: {using a secondary processor, e.g. coprocessor (peripheral processor G06F13/12)}
- G06F 9/547: {Remote procedure calls [RPC]; Web services}
- G06F 13/38: Information transfer, e.g. on bus (G06F13/14 takes precedence)
- G06F 13/4022: {using switching circuits, e.g. switching matrix, connection or expansion network (G06F13/4009 takes precedence)}
- G06F 13/4282: {on a serial bus, e.g. I2C bus, SPI bus (on daisy chain buses G06F13/4247)}
- G06F 15/167: using a common memory, e.g. mailbox
- G06F 2213/0026: PCI express
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

An apparatus and method of submitting hardware accelerator engine commands over an interconnect link such as a PCI Express (PCIe) link. In one embodiment, the mechanism is implemented inside a PCIe Host Bridge which is integrated into a host IC or chipset. The mechanism provides an interface compatible with other integrated accelerators thereby eliminating the overhead of maintaining different programming models for local and remote accelerators. Co-processor requests issued by threads requesting a service (client threads) targeting remote accelerator are queued and sent to a PCIe adapter and remote accelerator engine over a PCIe link. The remote accelerator engine performs the requested processing task, delivers results back to host memory and the PCIe Host Bridge performs co-processor request completion sequence (status update, write to flag, interrupt) include in the co-processor command.

## Claims

### Claim 1

A method of submitting co-processor commands across an interconnect link, said method comprising:
generating a co-processor command structure in a region of main memory;
issuing an accelerator command operative to transfer said co-processor command structure to an execution queue within a host bridge;
retrieving source data associated with said accelerator command from memory; and
transmitting said accelerator command, associated data and co-processor command structure to a remote hardware accelerator engine via said interconnect link.

### Claim 2 (depends on 1)

The method according to claim 1, wherein said interconnect link comprises a PCIe link.

### Claim 3 (depends on 1)

The method according to claim 1, wherein generating a co-processor command structure comprises allocating memory for source data, destination data, status data and data related to said accelerator command.

### Claim 4 (depends on 1)

The method according to claim 1, wherein issuing an accelerator command comprises writing said co-processor command structure into a memory mapped mailbox buffer using PCIe posted memory write TLPs.

### Claim 5 (depends on 1)

The method according to claim 1, wherein said accelerator command, associated data and co-processor command structure are transmitted to a target remote hardware accelerator engine in response to an accelerator command submission trigger.

### Claim 6 (depends on 1)

The method according to claim 1, further comprising receiving co-processor request task completion results from said remote hardware accelerator engine.

### Claim 7 (depends on 1)

The method according to claim 1, further comprising performing a co-processor request completion sequence in response to execution completion of a command request by said remote hardware accelerator engine.

### Claim 8 (depends on 1)

The apparatus according to claim 1, further comprising controlling communications between said host bridge and said remote hardware accelerator engine.

### Claim 9

A method of submitting co-processor commands across a PCIe link to a remote hardware accelerator engine, said method comprising:
generating co-processor commands at a PCIe host bridge for execution on a remote hardware accelerator engine coupled to a PCIe endpoint adaptor; and
submitting said co-processor commands across a PCIe link to said PCIe endpoint adapter for execution on said remote hardware accelerator engine.

### Claim 10

A method of submitting co-processor commands across an interconnect link to a remote hardware accelerator engine, said method comprising:
receiving an accelerator command over said interconnect link from a host bridge at an endpoint adapter and storing said accelerator command in an execution queue;
executing said accelerator command read from said execution queue on said remote hardware accelerator engine; and
sending processing results of said accelerator command execution back to said host bridge.

### Claim 11 (depends on 10)

The method according to claim 10, wherein said interconnect link comprises a PCIe link.

### Claim 12 (depends on 10)

The method according to claim 10, wherein said endpoint adapter comprises a PCIe endpoint adapter.

### Claim 13 (depends on 10)

The method according to claim 10, wherein executing said accelerator command comprises retrieving data required to execute said command requests from host memory.

### Claim 14 (depends on 10)

The method according to claim 10, further comprising performing requested processing tasks in accordance with said command request and sending processing results back to host memory.

### Claim 15 (depends on 10)

The method according to claim 10, further comprising notifying a PCIe host bridge via a PCIe interconnect link upon command request execution completion.

### Claim 16 (depends on 10)

The method according to claim 10, further comprising updating accelerator command delivery status in a PCIe host bridge port thereby allowing submission of new accelerator commands.

### Claim 17

A method of submitting co-processor commands across a PCIe link to a remote hardware accelerator engine, said method comprising:
generating co-processor commands at a PCIe host bridge for execution on a remote hardware accelerator engine coupled to a PCIe endpoint adaptor; and
submitting said co-processor commands across a PCIe link to said PCIe endpoint adapter for execution on said remote hardware accelerator engine;
executing said co-processor commands on said remote hardware accelerator engine; and
sending any processing execution results of said co-processor commands back to said PCIe host bridge

### Claim 18 (depends on 17)

The method according to claim 17, further comprising issuing co-processor requests by one or more client threads requesting service targeting said remote hardware accelerator engine.

### Claim 19 (depends on 17)

The method according to claim 17, wherein said remote hardware accelerator engine effectively appears as a locally integrated accelerator to CPU core processors connected to said PCIe host bridge.

### Claim 20 (depends on 17)

The method according to claim 17, wherein said method effectively enables software executing on a PCIe host bridge to interface with a remote hardware accelerator engine at a PCIe endpoint adaptor without requiring any direct intervention of local management software.

### Claim 21 (depends on 17)

The method according to claim 17, wherein said remote hardware accelerator engine effectively appears as a locally integrated accelerator to CPU core processors connected to said PCIe host bridge thereby enabling a common programming model for both local and remote hardware accelerators.

### Claim 22 (depends on 17)

The method according to claim 17, further comprising performing a co-processor request completion sequence in response to execution completion of a command request by said remote hardware accelerator engine.

### Claim 23 (depends on 17)

The method according to claim 17, wherein generating co-processor commands comprises:
generating a co-processor command structure in a region of main memory;
issuing an accelerator command operative to transfer said co-processor command structure to an execution queue within a host bridge; and
retrieving source data associated with said accelerator command from memory.

### Claim 24 (depends on 17)

The method according to claim 17, wherein executing said co-processor commands comprises:
receiving an accelerator command over said PCIe link from said PCIe host bridge at said PCIe endpoint adapter and storing said co-processor commands in an execution queue; and
executing said accelerator command read from said execution queue on said remote hardware accelerator engine.
