Nvidia Corporation

Wish Gandhi

Shirish C Gadre

Jack H Choquette

Lacky V Shah

John H Edmondson

Ronny Krashinsky

Manan Patel

Ravi Kiran Manyam

Prakash Bangalore Prabhakar

Apoorv Parle

Steve Heinrich

Jerome Francis Duluk, Jr

Olivier Giroux

Naveen Cherukuri

Luke Durant

Gentaro Hirota

Brian S Pharris

Ze Long

Jeff Tuckey

Vaishali Kulkarni

Gobikrishna Dhanuskodi

Rajballav Dash

Adam Hendrickson

Greg Palmer

Sandeep Navada

Raymond Wong

Eric Werness

Naman Govil

Shayani Deb

Alexander Lev Minkin

A new level(s) of hierarchy—Cooperate Group Arrays (CGAs)—and an associated new hardware-based work distribution/execution model is described. A CGA is a grid of thread blocks (also referred to as cooperative thread arrays (CTAs)). CGAs provide co-scheduling, e.g., control over where CTAs are placed/executed in a processor (such as a GPU), relative to the memory required by an application and relative to each other. Hardware support for such CGAs guarantees concurrency and enables applications to see more data locality, reduced latency, and better synchronization between all the threads in tightly cooperating collections of CTAs programmably distributed across different (e.g., hierarchical) hardware domains or partitions.

Cooperative group arrays

Distributed shared memory (DSMEM) comprises blocks of memory that are distributed or scattered across a processor (such as a GPU). Threads executing on a processing core local to one memory block are able to access a memory block local to a different processing core. In one embodiment, shared access to these DSMEM allocations distributed across a collection of processing cores is implemented by communications between the processing cores. Such distributed shared memory provides very low latency memory access for processing cores located in proximity to the memory blocks, and also provides a way for more distant processing cores to also access the memory blocks in a manner and using interconnects that do not interfere with the processing cores' access to main or global memory such as hacked by an L2 cache. Such distributed shared memory supports cooperative parallelism and strong scaling across multiple processing cores by permitting data sharing and communications previously possible only within the same processing core.

Distributed shared memory

This specification describes a programmatic multicast technique enabling one thread (for example, in a cooperative group array (CGA) on a GPU) to request data on behalf of one or more other threads (for example, executing on respective processor cores of the GPU). The multicast is supported by tracking circuitry that interfaces between multicast requests received from processor cores and the available memory. The multicast is designed to reduce cache (for example, layer 2 cache) bandwidth utilization enabling strong scaling and smaller tile sizes.

Programmatically controlled data multicasting across multiple compute engines

Apparatuses, systems, and techniques for supporting fairness of multiple context sharing cryptographic hardware. An accelerator circuit includes a copy engine (CE) with AES-GCM hardware configured to perform both encryption and authentication of data transfers for multiple applications or multiple data streams in a single application or belonging to a single user. The CE splits a data transfer of a specified size into a set of partial transfers. The CE sequentially executes the set of partial transfers using a context for a period of time (e.g., a timeslice) for an application. The CE stores in a secure memory for the application one or more data for encryption or decryption (e.g., a hash key, a block counter, etc.) computed from a last partial transfer. The one or more data for encryption or decryption are retrieved and used when data transfers for the application is resumed by the CE.

Fairly utilizing multiple contexts sharing cryptographic hardware

The subject application relates to computing an authentication tag for partial transfers scheduled across multiple direct memory access (DMA) engines. Apparatuses, systems, and techniques are described for computing an authentication tag for a data transfer when the data transfer is scheduled as partial transfers across a specified number of direct memory access (DMA) engines. An orchestration circuit stores partial authentication tags, computed by the DMA engines, and corresponding adjustment exponents during one or more rounds in which the partial transfers are scheduled and processed by the specified number of DMA engines. During a last round, a combined authentication tag can be computed based on the partial authentication tags and the corresponding adjustment exponents stored by the orchestration circuit during the rounds.

Computing an authentication tag for partial transfers scheduled across multiple direct memory access (DMA) engines

Growing community of inventors

Sunnyvale, CA, United States of America

Wish Gandhi