Disaggregation of processing pipeline
AssigneeNVIDIA Corporation
InventorsAdit Ranadive, Omri Kahalon, Aviad Shaul Yehezkel, Liran Liss, Gal Shalom, Yorai Itzhak Zack, Tushar Khinvasara
Life of this patent
Estimating the term…
Loading patent…
AssigneeNVIDIA Corporation
InventorsAdit Ranadive, Omri Kahalon, Aviad Shaul Yehezkel, Liran Liss, Gal Shalom, Yorai Itzhak Zack, Tushar Khinvasara
Estimating the term…
A method for processing includes receiving a definition of a processing pipeline including multiple sequential processing stages. The processing pipeline is partitioned into a plurality of partitions. The first partition of the processing pipeline is executed on a first computational accelerator, whereby the first computational accelerator writes output data from a final stage of the first partition to an output buffer in a first memory. The output data are copied over a packet communication network to an input buffer in a second memory. The second partition of the processing pipeline is executed on a second computational accelerator using the copied output data in the second memory as input data to a first stage of the second partition.
A method for processing, comprising:
A processing system, comprising:
A computer program product, comprising a tangible, non- transitory computer-readable medium in which program instructions are stored, which instructions, when read by a computer, cause the computer to receive a definition of a video processing pipeline comprising multiple sequential processing stages for processing an input video stream having a predefined video frame rate, to partition the video processing pipeline into a plurality of partitions to enable completion of all processing stages within a video frame duration that is determined by the predefined video frame rate, each partition comprising one or more successive stages of the video processing pipeline, including at least a first partition and a second partition following the first partition in the video processing pipeline, to instruct a first computational accelerator to execute the first partition of the video processing pipeline, whereby the first computational accelerator writes output data from a final stage of the first partition for a first video frame to an output buffer in a first memory, to instruct a NIC to copy the output data over a packet communication network to an input buffer in a second memory for execution of the second partition of the video processing pipeline on a second computational accelerator using the copied output data in the second memory as input data to a first stage of the second partition, and to cause the first computational accelerator to process the first partition of the video processing pipeline for a next video frame while the second computational accelerator processes the copied output data from the first video frame, thereby enabling real-time processing of consecutive video frames at the predefined video frame rate.
Dependencies are inferred from the claim text.