US 12,095,678 B2
GPU-native packet I/O method and apparatus for GPU application on commodity ethernet
Chan Gue Jung, Suwon-si (KR); Young Hoon Kim, Suwon-si (KR); Su Hwan Kim, Suwon-si (KR); Hong Uk Woo, Suwon-si (KR); and Ik Jun Yeom, Suwon-si (KR)
Assigned to Research & Business Foundation Sungkyunkwan University, Suwon-si (KR)
Filed by RESEARCH & BUSINESS FOUNDATION SUNGKYUNKWAN UNIVERSITY, Suwon-si (KR)
Filed on Oct. 8, 2021, as Appl. No. 17/497,131.
Claims priority of application No. 10-2021-0025097 (KR), filed on Feb. 24, 2021.
Prior Publication US 2022/0272052 A1, Aug. 25, 2022
Int. Cl. H04L 49/102 (2022.01); G06T 1/20 (2006.01); H04L 49/00 (2022.01); H04L 49/351 (2022.01); H04L 67/104 (2022.01)
CPC H04L 49/3063 (2013.01) [H04L 49/102 (2013.01); H04L 49/351 (2013.01); H04L 67/104 (2013.01); G06T 1/20 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A method for inputting/outputting a packet based on a commodity Ethernet device performed by a graphic processing unit (GPU) internal packet input/output device, the method comprising:
being allocated an available packet buffer from a memory pool inside a GPU;
after packets received from a network interface controller (NIC) are directly transferred to the allocated packet buffer, processing the directly transferred packets through a reception (Rx) kernel that updates information for a ring descriptor related to the directly transferred packets;
transmitting a transmission packet to a network through the NIC according to an operation of a transmission (Tx) kernel; and
returning the allocated packet buffer.