Nvidia Corporation

Other

Pastemak Solutions LLC

Rajeshwaran Selvanesan

Steven James Heinrich

Stewart Glenn Carlton

Anjana Rajendran

Charles McCarver

Michael Alan Fetterman

Shirish C Gadre

Alexander Lev Minkin

Omkar Paranjape

Eric Lyell Hill

Douglas J Hahn

Stephen C Purcell

Jack H Choquette

John H Edmondson

Scott A Kimura

Kevin Mitchell

Brett Coon

John R Nickolls

Ming Yiu Siu

Yan Yan Tang

Robert Anthony Alfieri

Olivier Giroux

Tezaswi Raja

Sachin Satish Idgunji

Brian Matthew Fahs

Robert Steven Glanville

Xiaogang Qiu

Manan Patel

Robert J Stoll

Stewart G Carlton

Divya Ramakrishnan

John Douglas Tynefield, Jr

Benjamin Faulkner

James P Reilley

Ewa Kubalska

Aniket Naik

Kevin Wilder

Fei (Walter) Li

Prasad Gharpure

Alex Gu

Daniel Rodriguez

Raj Jayakar

Brett W Coon

Systems and techniques for improving the performance of circuits while adapting to dynamic voltage drops caused by the execution of noisy instructions (e.g. high power consuming instructions) are provided. The performance is improved by slowing down the frequency of operation selectively for types of noisy instructions. An example technique controls a clock by detecting an instruction of a predetermined noisy type that is predicted to have a predefined noise characteristic (e.g. a high level of noise generated on the voltage rails of a circuit due to greater amount of current drawn by the instruction), and, responsive to the detecting, deceasing a frequency of the clock. The detecting occurs before execution of the instruction. The changing of the frequency in accordance with instruction type enables the circuits to be operated at high frequencies even if some of the workloads include instructions for which the frequency of operation is slowed down.

Techniques for configuring a processor to execute instructions efficiently

An integrated circuit such as, for example a graphics processing unit (GPU), includes a dynamic power controller for adjusting operating voltage and/or frequency. The controller may receive current power used by the integrated circuit and a predicted power determined based on instructions pending in a plurality of processors. The controller determines adjustments that need to be made to the operating voltage and/or frequency to minimize the difference between the current power and the predicted power. An in-system reinforced learning mechanism is included to self-tune parameters of the controller.

Techniques for efficiently operating a processing system based on energy characteristics of instructions and machine learning

One embodiment of the present disclosure sets forth an optimized way to execute pre-scheduled replay operations for divergent operations in a parallel processing subsystem. Specifically, a streaming multiprocessor (SM) includes a multi-stage pipeline configured to insert pre-scheduled replay operations into a multi-stage pipeline. A pre-scheduled replay unit detects whether the operation associated with the current instruction is accessing a common resource. If the threads are accessing data which are distributed across multiple cache lines, then the pre-scheduled replay unit inserts pre-scheduled replay operations behind the current instruction. The multi-stage pipeline executes the instruction and the associated pre-scheduled replay operations sequentially. If additional threads remain unserviced after execution of the instruction and the pre-scheduled replay operations, then additional replay operations are inserted via the replay loop, until all threads are serviced. One advantage of the disclosed technique is that divergent operations requiring one or more replay operations execute with reduced latency.

Pre-scheduled replays of divergent operations

One embodiment of the present disclosure sets forth an effective way to maintain fairness and order in the scheduling of common resource access requests related to replay operations. Specifically, a streaming multiprocessor (SM) includes a total order queue (TOQ) configured to schedule the access requests over one or more execution cycles. Access requests are allowed to make forward progress when needed common resources have been allocated to the request. Where multiple access requests require the same common resource, priority is given to the older access request. Access requests may be placed in a sleep state pending availability of certain common resources. Deadlock may be avoided by allowing an older access request to steal resources from a younger resource request. One advantage of the disclosed technique is that older common resource access requests are not repeatedly blocked from making forward progress by newer access requests.

Mechanism for waking common resource requests within a resource management subsystem

One embodiment of the present invention sets forth a technique for processing load instructions for parallel threads of a thread group when a sub-set of the parallel threads request the same memory address. The load/store unit determines if the memory addresses for each sub-set of parallel threads match based on one or more uniform patterns. When a match is achieved for at least one of the uniform patterns, the load/store unit transmits a read request to retrieve data for the sub-set of parallel threads. The number of read requests transmitted is reduced compared with performing a separate read request for each thread in the sub-set. A variety of uniform patterns may be defined based on common access patterns present in program instructions. A variety of uniform patterns may also be defined based on interconnect constraints between the load/store unit and the memory when a full crossbar interconnect is not available.

Uniform load processing for parallel thread sub-sets

A method for managing a parallel cache hierarchy in a processing unit. The method including receiving an instruction that includes a cache operations modifier that identifies a level of the parallel cache hierarchy in which to cache data associated with the instruction; and implementing a cache replacement policy based on the cache operations modifier.

Cache operations and policies for a multi-threaded client

Resource management subsystem that maintains fairness and order

One embodiment of the present invention sets forth an approach for executing replay operations for divergent operations in a parallel processing subsystem. Specifically, the streaming multiprocessor (SM) includes a multistage pipeline configured to batch two or more replay operations for processing via replay loop. A logic element within the multistage pipeline detects whether the current pipeline stage is accessing a shared resource, such as loading data from a shared memory. If the threads are accessing data which are distributed across multiple cache lines, then the multistage pipeline batches two or more replay operations, where the replay operations are inserted into the pipeline back-to-back.

Batched replays of divergent operations

Mechanism for tracking age of common resource requests within a resource management subsystem

One embodiment of the present invention sets forth a technique for performing a shaped access of a register file that includes a set of N registers, wherein N is greater than or equal to two. The technique involves, for at least one thread included in a group of threads, receiving a request to access a first amount of data from each register in the set of N registers, and configuring a crossbar to allow the at least one thread to access the first amount of data from each register in the set of N registers.

Shaped register file reads

The invention sets forth an L1 cache architecture that includes a crossbar unit configured to transmit data associated with both read data requests and write data requests. Data associated with read data requests is retrieved from a cache memory and transmitted to the client subsystems. Similarly, data associated with write data requests is transmitted from the client subsystems to the cache memory. To allow for the transmission of both read and write data on the crossbar unit, an arbiter is configured to schedule the crossbar unit transmissions as well and arbitrate between data requests received from the client subsystems.

Sharing data crossbar for reads and writes in a data cache

One embodiment sets forth a technique for dynamically mapping addresses to banks of a multi-bank memory based on a bank mode. Application programs may be configured to perform read and write a memory accessing different numbers of bits per bank, e.g., 32-bits per bank, 64-bits per bank, or 128-bits per bank. On each clock cycle an access request may be received from one of the application programs and per processing thread addresses of the access request are dynamically mapped based on the bank mode to produce a set of bank addresses. The bank addresses are then used to access the multi-bank memory. Allowing different bank mappings enables each application program to avoid bank conflicts when the memory is accesses compared with using a single bank mapping for all accesses.

Dynamic bank mode addressing for memory access

One embodiment of the present invention sets forth a technique for providing a L1 cache that is a central storage resource. The L1 cache services multiple clients with diverse latency and bandwidth requirements. The L1 cache may be reconfigured to create multiple storage spaces enabling the L1 cache may replace dedicated buffers, caches, and FIFOs in previous architectures. A 'direct mapped' storage region that is configured within the L1 cache may replace dedicated buffers, FIFOs, and interface paths, allowing clients of the L1 cache to exchange attribute and primitive data. The direct mapped storage region may used as a global register file. A 'local and global cache' storage region configured within the L1 cache may be used to support load/store memory requests to multiple spaces. These spaces include global, local, and call-return stack (CRS) memory.

Configurable cache for multiple clients

One embodiment of the present invention sets forth a technique for arbitrating requests received by an L1 cache from multiple clients. The L1 cache outputs bubble requests to a first one of the multiple clients that cause the first one of the multiple clients to insert bubbles into the request stream, where a bubble is the absence of a request. The bubbles allow the L1 cache to grant access to another one of the multiple clients without stalling the first one of the multiple clients. The L1 cache services multiple clients with diverse latency and bandwidth requirements and may be reconfigured to provide memory spaces for clients executing multiple parallel threads, where the memory spaces each have a different scope.

Cache arbitration between multiple clients

One embodiment of the present invention sets forth a technique for arbitrating requests received from one of the multiple clients of an L1 cache and for providing hints to the client to assist in arbitration. The L1 cache services multiple clients with diverse latency and bandwidth requirements and may be reconfigured to provide memory spaces for clients executing multiple parallel threads, where the memory spaces each have a different scope.

Cache interface protocol including arbitration and hints

Growing community of inventors

Milpitas, CA, United States of America

Rajeshwaran Selvanesan