Abstract: Techniques for managing dataset resource instance performance via a data-centric approach are disclosed. A system determines an aggregated level of service demands placed on individual dataset resource instances that may be in a distributed computing system. The system may identify portions of a dataset that are associated with high levels of service demands. Once identified, the system may provide an administrator with the service demand information. The administrator may relocate these high demand dataset portions to other dataset resource instances that are better able to respond to the high levels of demand without impaired performance.
Abstract: A system and method for dynamically protecting against security vulnerabilities in a reconfigurable signal chain. The system includes a signal chain formed from at least a first component connected with a second component. The first component has a set of source outputs and a first authentication block, and the second signal chain component has a set of destination inputs and a second authentication block. The system also includes a signal chain configurator that populates the first authentication block with at least one validated endpoint from the set of destination inputs. A signal chain integrity block, which is communicatively coupled with the first authentication block and the second authentication block, identifies a source-destination pair from one or more endpoint pairs formed from the at least one validated endpoint and the set of source outputs. The signal chain integrity block propagates the source-destination pair to the first authentication block and the second authentication block.
Type:
Grant
Filed:
February 28, 2023
Date of Patent:
June 18, 2024
Assignee:
Texas Instruments Incorporated
Inventors:
Veeramanikandan Raju, Anand Kumar G, Christy Leigh She
Abstract: A system and method of accelerating execution of a NN model, by at least one processor may include: receiving a first matrix A, representing elements of a kernel K of the NN model and a second matrix B, representing elements of an input I to kernel K; producing from matrix A, a group-sparse matrix A?, comprising G tensors of elements. The number of elements in each tensor is defined by, or equal to a number of entries in each index of an input tensor register used for a specific Single Instruction Multiple Data (SIMD) tensor operation, and all elements of A? outside said G tensors are null. The system and method may further include executing kernel K on input I, by performing at least one computation of the SIMD tensor operation, having as operands elements of a tensor of the G tensors and corresponding elements of the B matrix.
Type:
Grant
Filed:
November 4, 2021
Date of Patent:
October 24, 2023
Assignee:
Neuralmagic, Inc.
Inventors:
Alexander Matveev, Dan Alistarh, Justin Kopinsky, Rati Gelashvili, Mark Kurtz, Nir Shavit