Software Ag, Inc.

Mark Horsburgh

Christopher Reed

Matthew Alastair Johnson

Richard M Bentley

Robert Scott Mitchell

A memory management system for managing objects which represent memory in a multi-threaded operating system extracts the ID of the home free-list from the object header to determine whether the object is remote and adds the object to a remote object list if the object is determined to be remote. The memory management system determines whether the number of objects on the remote object list exceeds a threshold. If the threshold is exceeded, the system batch-removes the objects on the remote object list and then adds those objects to the appropriate one or more remote home free-lists.

System and method for managing the allocating and freeing of objects in a multi-threaded system

A method and/or computer for a tuned spin count in a multithreaded system determines a re-calculation time interval at which to re-calculate a current spin lock value. Then, a spin-lock-re-calculation is repeatedly executed at the re-calculation time interval to perform: observing a current environment of the multithreaded system, determining, using a second-order tuning and values of the current environment, a dynamically calculated heuristic to provide the newly-recalculated spin lock value, and memorizing the newly re-calculated spin lock value, in a memory, as the current spin lock value. Meanwhile, thread(s) in the multithreaded system which want to execute a spinlock will obtain the current spin lock value which is memorized in the memory, and execute the spin lock using the current spin lock value to set a length of the spin lock.

Second-order tuning of spin loops

A system and method for replicating a distributed computer system state among a plurality of deterministic event processing engines are disclosed. A sender module receives a plurality of input events or input event streams and generates a serialized message sequence describing an order in which input events are processed. The serialized message sequence is then transmitted to a plurality of deterministic event processing engines for processing. As the event processing engines are deterministic, each event processing engine produces the same ordered sequence responsive to the received serialized message sequence. The ordered sequences from each deterministic event processing engine are transmitted to a merger module which generates an output from the plurality of ordered sequences. The merger module removes one or more redundant ordered sequences and associates the resulting output with a client device.

Growing community of inventors

Cambridge, United Kingdom

Mark Horsburgh