Barcelona Supercomputing Center-centro Nacional De Supercomputacion

Francisco Javier Cazorla Almeida

A method for executing several tasks in a multithreaded (MT) processor, each task having, for every hardware shared resource from a predetermined set of hardware shared resources in the MT processor, one associated artificial time delay that is introduced when a task accesses a hardware shared resource, the method including step (a) of establishing, for every hardware shared resource and each task to be artificially delayed, the artificial delay to be applied to each access of each task to every hardware shared resource; step (b) of performing the following steps (b1) to (b4) on the access of each task to be artificially delayed to every hardware shared resource in the predetermined set of hardware shared resources in the MT processor: step (b1) of verifying if a request is ready to be sent to the hardware shared resource; in case of positive result, Step (b2) of verifying if the hardware shared resource is ready to accept the request; in case of positive result, step (b3) of delaying the request by the artificial delay established for the hardware shared resource; step (b4) of allowing the request of the task to access the hardware shared resource.

Growing community of inventors

San Fernando de Maspalomas, Spain

Francisco Javier Cazorla Almeida