Nvidia Corporation

Alan Menezes

Jerome Francis Duluk, Jr

Pranav Vaidya

James Leroy Deming

Wishwesh Anil Gandhi

Gregory Scott Palmer

Shailendra Singh

Manas Mandal

Praveen Joginipally

Feiqi Su

Lacky V Shah

Samuel Hammond Duncan

Timothy John Purcell

Jonathon Stuart Ramsey Evans

Eric Rock

Luke Durant

Gentaro Hirota

Brian S Pharris

Jeffrey Scott Tuckey

Abhijat Ranade

Shounak Kamalapurkar

Ramon Matas Navarro

Sonata Gale Wen

Roger L Allen

Timothy J McDonald

Thomas Melvin Ogletree

Siddharth Sharma

Jonathon Stuart Ramsay Evans

Gregory Paul Smith

Tom Ogletree

Jin Ouyang

A parallel processing unit (PPU), operating in a traditional processing environment or in a virtualized processing environment, can be divided into partitions. Each partition is configured to operate similarly to how the entire PPU operates. A given partition includes a subset of the computational and memory resources associated with the entire PPU. Software that executes on a CPU partitions the PPU for an admin user. A guest user is assigned to a partition and can perform processing tasks within that partition in isolation from any other guest users assigned to any other partitions. Because the PPU can be divided into isolated partitions, multiple CPU processes can efficiently utilize PPU resources.

Techniques for configuring a processor to function as multiple, separate processors in a virtualized environment

Various embodiments include a system for generating performance monitoring data in a computing system. The system includes a unit level counter with a set of counters, where each counter increments during each clock cycle in which a corresponding electronic signal is at a first state, such as a high or low logic level state. Periodically, the unit level counter transmits the counter values to a corresponding counter collection unit. The counter collection unit includes a set of counters that aggregates the values of the counters in multiple unit level counters. Based on certain trigger conditions, the counter collection unit transmits records to a reduction channel. The reduction channel includes a set of counters that aggregates the values of the counters in multiple counter collection units. Each virtual machine executing on the system can access a different corresponding reduction channel, providing secure performance metric data for each virtual machine.

Generating power efficient secure multiclient performance monitoring data

A parallel processing unit (PPU) can be divided into partitions. Each partition is configured to operate similarly to how the entire PPU operates. A given partition includes a subset of the computational and memory resources associated with the entire PPU. Software that executes on a CPU partitions the PPU for an admin user. A guest user is assigned to a partition and can perform processing tasks within that partition in isolation from any other guest users assigned to any other partitions. Because the PPU can be divided into isolated partitions, multiple CPU processes can efficiently utilize PPU resources.

Techniques for configuring a processor to function as multiple, separate processors

Techniques for reconfiguring partitions in a parallel processing system

Embodiments of the present invention set forth techniques for allocating execution resources to groups of threads within a graphics processing unit. A compute work distributor included in the graphics processing unit receives an indication from a process that a first group of threads is to be launched. The compute work distributor determines that a first subcontext associated with the process has at least one processor credit. In some embodiments, CTAs may be launched even when there are no processor credits, if one of the TPCs that was already acquired has sufficient space. The compute work distributor identifies a first processor included in a plurality of processors that has a processing load that is less than or equal to the processor loads associated with all other processors included in the plurality of processors. The compute work distributor launches the first group of threads to execute on the first processor.

Dynamic partitioning of execution resources

Performance monitors are placed on computational units in different clock domains of an integrated circuit. A central dispatcher generates trigger signals to the performance monitors to cause the performance monitors to respond to the trigger signals with packets reporting local performance counts for the associated computational units. The data in the packets are correlated into a single clock domain. By applying a trigger and reporting system, the disclosed approach can synchronize the performance metrics of the various computational units in the different clock domains without having to route a complex global clock reference signal to all of the performance monitors.

Efficient performance monitoring of integrated circuit(s) having distributed clocks

Growing community of inventors

San Jose, CA, United States of America

Alan Menezes