IP Library › Granted Patent US 10,846,362
Granted Patent B2
US 10,846,362 · App. 16/080,159 · Granted Nov 24, 2020

Information processing apparatus, information processing method, data structure and program

Inventor: Takuya Araki (Tokyo, JP)
Assignee: NEC CORPORATION
G06F17/16G06F9/3004G06F9/30036
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 10,846,362
App. No.
16/080,159
Granted
Nov 24, 2020
Kind
B2
Abstract

An information processing apparatus comprises a first conversion part storing a partial matrix for columns having a predetermined number or more of non-zero elements among columns in a matrix in a first format, and a second conversion part storing a partial matrix for the other columns in the matrix in a second format.

Claims (42)

1. An information processing apparatus, comprising:

a first conversion part storing a partial matrix for columns having a predetermined number or more of non-zero elements among columns in a matrix in a first format; and

a second conversion part storing a partial matrix for the other columns in the matrix in a second format,

wherein the first format is a format in which elements in the matrix are stored in a column major order and the second format is a format in which elements in the matrix are stored in a row major order.

2. The information processing apparatus of claim 1 , wherein the first format is JDS (Jagged Diagonal Storage) format.

3. The information processing apparatus of claim 2 , wherein the second format is a format in which elements in the matrix are stored in a row major order.

4. The information processing apparatus of claim 2 , wherein

the first conversion part left-shifts non-zero elements in the matrix, and stores a partial matrix for columns having a predetermined number or more of non-zero elements among columns in the left-shifted matrix in the first format; and

the second conversion part stores a partial matrix for the other columns in the left-shifted matrix in the second format.

5. The information processing apparatus of claim 2 , comprising

a matrix vector multiplication calculation part which calculates multiplication products of the partial matrix stored in the first format and a vector, calculates multiplication products of the partial matrix stored in the second format and the vector, and sums up the calculated multiplication products so as to obtain multiplication products of the matrix and the vector.

6. The information processing apparatus of claim 1 , wherein the second format is CRS (Compressed Row Storage) format.

7. The information processing apparatus of claim 6 , wherein

the first conversion part left-shifts non-zero elements in the matrix, and stores a partial matrix for columns having a predetermined number or more of non-zero elements among columns in the left-shifted matrix in the first format; and

the second conversion part stores a partial matrix for the other columns in the left-shifted matrix in the second format.

8. The information processing apparatus of claim 6 , comprising

a matrix vector multiplication calculation part which calculates multiplication products of the partial matrix stored in the first format and a vector, calculates multiplication products of the partial matrix stored in the second format and the vector, and sums up the calculated multiplication products so as to obtain multiplication products of the matrix and the vector.

9. The information processing apparatus of claim 1 , wherein

the first conversion part left-shifts non-zero elements in the matrix, and stores a partial matrix for columns having a predetermined number or more of non-zero elements among columns in the left-shifted matrix in the first format; and

the second conversion part stores a partial matrix for the other columns in the left-shifted matrix in the second format.

10. The information processing apparatus of claim 9 , comprising

a matrix vector multiplication calculation part which calculates multiplication products of the partial matrix stored in the first format and a vector, calculates multiplication products of the partial matrix stored in the second format and the vector, and sums up the calculated multiplication products so as to obtain multiplication products of the matrix and the vector.

11. The information processing apparatus of claim 1 , comprising

a matrix vector multiplication calculation part which calculates multiplication products of the partial matrix stored in the first format and a vector, calculates multiplication products of the partial matrix stored in the second format and the vector, and sums up the calculated multiplication products so as to obtain multiplication products of the matrix and the vector.

12. The information processing apparatus of claim 1 , wherein

the first conversion part left-shifts non-zero elements in the matrix, and stores a partial matrix for columns having a predetermined number or more of non-zero elements among columns in the left-shifted matrix in the first format; and

the second conversion part stores a partial matrix for the other columns in the left-shifted matrix in the second format.

13. The information processing apparatus of claim 1 , wherein

the first conversion part left-shifts non-zero elements in the matrix, and stores a partial matrix for columns having a predetermined number or more of non-zero elements among columns in the left-shifted matrix in the first format; and

the second conversion part stores a partial matrix for the other columns in the left-shifted matrix in the second format.

14. The information processing apparatus of claim 1 , comprising

a matrix vector multiplication calculation part which calculates multiplication products of the partial matrix stored in the first format and a vector, calculates multiplication products of the partial matrix stored in the second format and the vector, and sums up the calculated multiplication products so as to obtain multiplication products of the matrix and the vector.

15. The information processing apparatus of claim 1 , comprising

a matrix vector multiplication calculation part which calculates multiplication products of the partial matrix stored in the first format and a vector, calculates multiplication products of the partial matrix stored in the second format and the vector, and sums up the calculated multiplication products so as to obtain multiplication products of the matrix and the vector.

16. An information processing method, comprising:

storing a partial matrix for columns having a predetermined number or more of non-zero elements among columns in a matrix in a first format; and

storing a partial matrix for the other columns in the matrix in a second format,

wherein the first format is a format in which elements in the matrix are stored in a column major order and the second format is a format in which elements in the matrix are stored in a row major order.

17. A data structure having features:

to store a partial matrix for columns having a predetermined number or more of non-zero elements among columns in a matrix in a first format; and

to store a partial matrix for the other columns in the matrix in a second format,

wherein the first format is a format in which elements in the matrix are stored in a column major order and the second format is a format in which elements in the matrix are stored in a row major order.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Aug 27, 2018
From: ARAKI, TAKUYA
To: NEC CORPORATION
Reel/Frame 046713/0192 →
Priority Claims (1)
JP 2016-046028 · Mar 9, 2016 · national
Continuity (1)
Related Publication 20190050371A1 · Feb 14, 2019