Nvidia Corporation

Azul Systems, Inc.

Eric Lyell Hill

Michael Alan Fetterman

Rajeshwaran Selvanesan

Omkar Paranjape

Anjana Rajendran

Shirish C Gadre

Steven James Heinrich

Charles McCarver

Jack H Choquette

John H Edmondson

Stewart Glenn Carlton

Douglas J Hahn

Kevin Mitchell

Shu-Yi Yu

Richard L Schober, Jr

Russell R Newcomb

Hungse Cha

Olivier Giroux

Xiaogang Qiu

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

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

One embodiment of the present invention sets forth a technique for performing high-performance clock training. One clock training sweep operation is performed to determine phase relationships for two write clocks with respect to a command clock. The phase relationships are generated to satisfy timing requirements for two different client devices, such as GDDR5 DRAM components. A second clock training sweep operation is performed to better align local clocks operating on the client devices. A voting tally is maintained during the second clock training sweep to record phase agreement at each step in the clock training sweep. The voting tally then determines whether one of the local clocks should be inverted to better align the two local clocks.

Hardware WCK2CK training engine using meta-EDC sweeping and adjustably accurate voting algorithm for clock phase detection

Cyclic redundancy check (CRC) values are efficiently calculated using an improved linear feedback shift register (LFSR) circuit. CRC value generation is separated into two sub-calculations, which are then combined to form a final CRC value. A programmable XOR engine performs logic functions via a table lookup rather than via a random logic circuit. LCRC and ECRC calculations are performed using a single shared LFSR circuit. Multiple links share the same CRC value generator. One advantage of the present invention is that CRC values are generated using smaller and fewer LFSR circuits relative to conventional circuit designs. As a result, a CRC value generator utilizing the disclosed techniques consumes less surface area of an integrated circuit and consumes less power, resulting in cooler operation.

Cyclic redundancy check generation via distributed time multiplexed linear feedback shift registers

A self-checking test generator program creates a self-checking test program that can test a device under test (DUT). The self-checking test generator selects instructions for a test. Selected instructions are executed on a software DUT model to generate results that can be self-checked by other instructions such as compare and branch instructions. The software DUT model has fuzzy models and unknown models for blocks in the DUT. Fuzzy models generate expected outputs for a block of the DUT. Fuzzy models may propagate unknown data from their inputs to their outputs. Unknown models do not predict expected outputs. Instead, unknown models always output unknown (X). Over time, as more of the DUT logic is modeled, unknown models may be replaced with fuzzy models.

Self-checking test generator for partially-modeled processors by propagating fuzzy states

Growing community of inventors

Palo Alto, CA, United States of America

Eric Lyell Hill