Hewlett-packard Development Company, L.p.

Hewlett-packard Development, L.p.

Christopher Philip Ruemmler

Jonathan Kent Ross

Matthew L Fischer

Colin Edward Honess

Patrick O Kilfoyle

Fani A Pavlova

A method of co-locating threads and corresponding system are described. The method comprises a first thread executing on a first processor awakening a second thread for execution on a second processor and assigning the second thread to execute on the first processor based on a determination that the first thread awakened the second thread at a prior awakening of the second thread.

Wakeup-attribute-based allocation of threads to processors

A method of controlling thread access to a synchronization object is described. The method comprises waking a sleeping first thread after release, by a second thread, of a synchronization object in a normal state and causing, by the first thread, the synchronization object to transition to a want-lock state, the synchronization object being acquirable by the first thread prior to being acquirable by the second thread.

Method of controlling thread access to a synchronization object

An embodiment of the invention provides an apparatus and a method of dynamically controlling the number of busy waiters in for a synchronization object. The apparatus and method perform the steps of increasing a number of allowed busy waiters if there is a waiter in a sleep state and there are no current busy waiters when a requester releases the synchronization object, and decreasing the number of allowed busy waiters if a busy waiter moves from a busy waiting state to the sleep state.

Dynamically controlling the number of busy waiters in a synchronization object

One embodiment disclosed relates to a method of reducing access latency to a task priority register (TPR) of a local programmable interrupt controller unit within a microprocessor. A command is received to write an interrupt mask value to the TPR, and the interrupt mask value is written to the TPR. In addition, the interrupt mask value is also written into a shadow copy of the TPR. The shadow copy is written each time that the TPR is written. Another embodiment disclosed relates to a method of reducing a latency to read a TPR of an IPF type microprocessor. When a command is received to read an interrupt mask value from the TPR, the interrupt mask value is read from the shadow copy at a memory location, instead of from the task priority register itself.

Reducing latency, when accessing task priority levels

A computer-implemented method for handling pending interrupt vectors of a pending interrupt list is disclosed. The method includes batch-reading the set of pending interrupt vectors into a working list of working interrupt vectors. The method also includes performing interrupt handling of the working interrupt vectors using an interrupt handling arrangement until the working list is empty. The interrupt handling process permits a first incoming interrupt vector that is received by the pending interrupt list after the batch reading to temporarily interrupt the performing interrupt handling of the working interrupt vectors and to be handled on a priority basis by the interrupt handling arrangement if a priority level of the first incoming interrupt vector is higher than a priority level of a first working interrupt vector being currently handled by the interrupt handling arrangement.

Batch processing of interrupts

One embodiment disclosed relates to a computer system. The computer system includes a microprocessor, an operating system, and a memory system. The microprocessor includes a register stack and a register stack engine (RSE), and the operating system includes a kernel. The memory system is configured to have a single memory page that includes both a kernel stack and an RSE stack. The memory system may be further configured such that the kernel stack and the RSE stack grow in opposite directions and such that a uarea data structure is located between those two stacks.

Growing community of inventors

Morgan Hill, CA, United States of America

Christopher Philip Ruemmler