IP Library Granted Patent US 12,524,244
Granted Patent B2
US 12,524,244 · App. 18/287,192 · Granted Jan 13, 2026

Accelerator control system using control data to perform arithmetic processing on a plurality of accelerators

Inventors: Tomoya Yokono (Musashino, JP); Yoshiro Yamabe (Musashino, JP); Teruaki Ishizaki (Musashino, JP)
Assignee: NTT, Inc.
G06F9/3881G06F9/3885G06F13/1673
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,524,244
App. No.
18/287,192
Granted
Jan 13, 2026
Kind
B2
Abstract

An accelerator control system includes an accelerator control device and a plurality of accelerators, wherein the accelerator control device includes first processing circuitry configured to store control data including a location of data which is an arithmetic processing target, information on an accelerator for performing arithmetic processing, and information specifying content of the arithmetic processing of the accelerator, and determine completion of the arithmetic processing by the accelerator when the control data which has been subjected to the arithmetic processing by the accelerator is stored in a storage, and each of the accelerators includes second processing circuitry configured to acquire the control data from the storage, and perform arithmetic processing on the data which is an arithmetic processing target according to the location of the data which is an arithmetic processing target and information specifying the content of the arithmetic processing of the accelerator included in the acquired control data.

Claims (38)

1 . An accelerator control system comprising:

an accelerator control device; and

a plurality of accelerators,

wherein the accelerator control device includes:

first processing circuitry configured to:

store control data including a location of data which is an arithmetic processing target, information on an accelerator for performing arithmetic processing, and information specifying content of the arithmetic processing of the accelerator; and

determine completion of the arithmetic processing by the accelerator when the control data which has been subjected to the arithmetic processing by the accelerator is stored in a storage, and

each of the accelerators includes:

second processing circuitry configured to:

acquire the control data from the storage;

perform the arithmetic processing on the data which is an is the arithmetic processing target according to the location of the data which is an is the arithmetic processing target and the information specifying the content of the arithmetic processing of the accelerator included in the acquired control data;

store the control data in the storage when the arithmetic processing is completed and the control data does not include information on a transfer destination accelerator;

transfer the control data to the transfer destination accelerator when the control data includes information on the transfer destination accelerator; and

perform the arithmetic processing on the data which is the arithmetic processing target according to the location of the data which is the arithmetic processing target and the information specifying the content of the arithmetic processing included in the control data when the control data has been received from another accelerator.

2 . The accelerator control system according to claim 1 , wherein the second processing circuitry is further configured to check whether or not the control data is receivable from another accelerator.

3 . The accelerator control system according to claim 2 , wherein the second processing circuitry is further configured to retransmit the control data when the transfer destination accelerator does not notify of whether or not the control data is receivable within a predetermined time.

4 . The accelerator control system according to claim 1 , wherein the second processing circuitry is further configured to:

store a result of the arithmetic processing performed on the data which is the arithmetic processing target in the storage when the control data has been transferred to the transfer destination accelerator, and

read the result of the arithmetic processing stored in the storage and set the result of the arithmetic processing as the arithmetic processing target.

5 . The accelerator control system according to claim 1 , wherein the second processing circuitry is further configured to monitor whether or not the control data is present in the storage.

6 . The accelerator control system according to claim 1 , wherein the first processing circuitry is further configured to monitor whether or not the control data which has been subjected to the arithmetic processing is present in the storage.

7 . An accelerator control method executed by an accelerator control system including an accelerator control device and a plurality of accelerators, wherein the accelerator control system includes a storage configured to store control data including a location of data which is an arithmetic processing target, information on an accelerator for performing arithmetic processing, and information specifying content of the arithmetic processing of the accelerator, the accelerator control method comprising:

acquiring the control data from the storage;

performing the arithmetic processing on the data which is the arithmetic processing target according to the location of the data which is the arithmetic processing target and the information specifying the content of the arithmetic processing of the accelerator included in the acquired control data;

determining that the arithmetic processing is completed and the control data does not include information on a transfer destination accelerator;

storing the control data in the storage when the accelerator control system has determined that the arithmetic processing is completed and the control data does not include information on a transfer destination accelerator;

determining that the control data which has been subjected to the arithmetic processing is stored in the storage;

determining completion of the arithmetic processing when the accelerator control system has determined that the control data which has been subjected to the arithmetic processing is stored in the storage;

transferring the control data to the transfer destination accelerator when the control data includes information on the transfer destination accelerator;

determining that the control data has been received from another accelerator; and

performing the arithmetic processing on the data which is the arithmetic processing target according to the location of the data which is the arithmetic processing target and the information specifying the content of the arithmetic processing included in the control data when the accelerator control system has determined that the control data has been received from another accelerator.

8 . A non-transitory computer-readable recording medium storing therein an accelerator control program that causes a computer to execute a process comprising:

acquiring control data from a storage configured to store the control data including a location of data which is an arithmetic processing target, information on an accelerator for performing arithmetic processing, and information specifying content of the arithmetic processing of the accelerator;

performing the arithmetic processing on the data which is the arithmetic processing target according to the location of the data which is the arithmetic processing target and the information specifying the content of the arithmetic processing of the accelerator included in the acquired control data;

storing the control data in the storage when the arithmetic processing is completed and the control data does not include information on a transfer destination accelerator;

determining completion of the arithmetic processing when the control data which has been subjected to the arithmetic processing is stored in the storage;

transferring the control data to the transfer destination accelerator when the control data includes information on the transfer destination accelerator; and

performing the arithmetic processing on the data which is the arithmetic processing target according to the location of the data which is the arithmetic processing target and the information specifying the content of the arithmetic processing included in the control data when the control data has been received from another accelerator.

Assignments (2)
CHANGE OF NAME Recorded Aug 20, 2025
From: NIPPON TELEGRAPH AND TELEPHONE CORPORATION
To: NTT, INC.
Reel/Frame 072556/0180 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 17, 2023
From: YOKONO, TOMOYA; YAMABE, YOSHIRO; ISHIZAKI, TERUAKI
To: NIPPON TELEGRAPH AND TELEPHONE CORPORATION
Reel/Frame 065246/0303 →
Continuity (1)
Related Publication 20240202033A1 · Jun 20, 2024
References Cited (7)
US 11321254B2 · Kim · 2022 [cited by examiner]
US 20080244126A1 · Hundley · 2008 [cited by applicant]
US 20180095750A1 · Drysdale et al. · 2018 [cited by applicant]
US 20210191728A1 · Ahn · 2021 [cited by examiner]
Yang et al., “When Poll is Better than Interrupt”, Fast, vol. 12, Available Online at: https://www.usenix.org/system/files/conference/fast12/yang.pdf, 2012, pp. 1-7. [cited by applicant]
“CUDA Toolkit Documentation v11.2.2”, Nvidia, Available Online at: https://docs.nvidia.com/cuda/, Retrieved from the net on: Mar. 29, 2021, 1 page. [cited by applicant]
Gulati et al., “GPU Architecture and the CUDA Programming Model”, Hardware acceleration of EDA algorithms, Available Online at: https://link.springer.com/content/pdf/10.1007%2F978-1-4419-0944-2_3.pdf, 2010, pp. 23-30. [cited by applicant]