MPI communication of GPU buffers

US9178747B2 · kind B2 · utility

Assignee

• NVIDIA Corporation · US

Inventors

• Rolf VandaVaart · Harvard, US
• Timothy Murray · Mountain View, US
• Peter Buckingham · Parramatta, AU

Key dates

• Filing date: Nov 29, 2012
• Grant date: Nov 3, 2015
• Expiry date: Nov 29, 2033

Classification

• Technology area (CPC —): General

Abstract

A technique for enhancing the efficiency and speed of data transmission within and across multiple, separate computer systems includes the use of an MPI library/engine. The MPI library/engine is configured to facilitate the transfer of data directly from one location to another location within the same computer system and/or on separate computer systems via a network connection. Data stored in one GPU buffer may be transferred directly to another GPU buffer without having to move the data into and out of system memory or other intermediate send and receive buffers.