Nvidia Corporation

Gary Ward

M Wasiur Rashid

Philip Browning Johnson

Wei-Je Robert Huang

Samuel Hammond Duncan

Dennis K Ma

Lincoln G Garlick

Jonathon Evans

Wojciech Jan Truty

Embodiments of the present invention set forth techniques for resolving page faults associated with a copy engine. A copy engine within a parallel processor receives a copy operation that includes a set of copy commands. The copy engine executes a first copy command included in the set of copy commands that results in a page fault. The copy engine stores the set of copy commands to the memory. At least one advantage of the disclosed techniques is that the copy engine can perform copy operations that involve source and destination memory pages that are not pinned, leading to reduced memory demand and greater flexibility.

Low overhead copy engine fault and switch mechanism

A copy subsystem within a processor includes a set of logical copy engines and a set of physical copy engines. Each logical copy engine corresponds to a different command stream implemented by a device driver, and each logical copy engine is configured to receive copy commands via the corresponding command stream. When a logical copy engine receives a copy command, the logical copy engine distributes the command, or one or more subcommands derived from the command, to one or more of the physical copy engines. The physical copy engines can perform multiple copy operations in parallel with one another, thereby allowing the bandwidth of the communication link(s) to be saturated.

Managing copy operations in complex processor topologies

A multi-threaded processing unit includes a hardware pre-processor coupled to one or more processing engines (e.g., copy engines, GPCs, etc.) that implement pre-emption techniques by dividing tasks into smaller subtasks and scheduling subtasks on the processing engines based on the priority of the tasks. By limiting the size of the subtasks, higher priority tasks may be executed quickly without switching the context state of the processing engine. Tasks may be subdivided based on a threshold size or by taking into account other consideration such as physical boundaries of the memory system.

Technique for improving performance in multi-threaded processing units

Non-contiguous or tiled payload data are efficiently transferred between peers over a fabric. Specifically, a client transfers a byte enable message to a peer device via a mailbox mechanism, where the byte enable message specifies which bytes of the payload data being transferred via the data packet are to be written to the frame buffer on the peer device and which bytes are not to be written. The client transfers the non-contiguous or tiled payload payload data to the peer device. Upon receiving the payload data, the peer device writes bytes from the payload data into the target frame buffer for only those bytes enabled via the byte enable message. One advantage of the present invention is that non-contiguous or tiled data are transferred over a fabric with improved efficiency.

Growing community of inventors

Santa Clara, CA, United States of America

Gary Ward