Abstract: Garbage collection processes are automatically tuned on a system comprising a plurality of virtual machines running on a physical processing device with multiple physical processors. Tuning is implemented for each garbage collection process, by updating a load variable to reflect that a new garbage collection process has started, the load variable having a value that is based at least in part on a number of garbage collection processes that are currently running on the system. Tuning further comprises creating a number of garbage collection threads based at least in part on the current value of the load variable and/or number of processors, where the garbage collection threads are used to perform at least a portion of at least one phase of the new garbage collection process. Tuning further comprises re-updating the load variable to reflect completion of the new garbage collection process.

Abstract: An enhanced function-descriptor-based dispatch in a multi-linkage environment receives user code containing a function compiled in a supplementary linkage convention of a caller to form an invoked function and determines whether the supplementary linkage convention of the caller for the invoked function matches a supplementary linkage implementation provided by a library. Responsive to a determination that the supplementary linkage convention of the caller for the invoked function matches a supplementary linkage implementation provided by the library, an embodiment selects the supplementary linkage implementation provided by the library and dispatches the invoked function in the selected supplementary linkage implementation provided by the library.

Abstract: A method, system and computer program product for sharing class data among virtual machine applications running on one or more guests in a virtualized environment. A control program in a virtual operating system is used to manage the user portions of the virtual operating system, each commonly referred to as a guest. A guest operating system runs on each guest and applications can run on each guest operating system. A memory management facility manages shared memory which includes a class cache configured to store class data. The shared memory may be mounted onto each guest using a cluster file system or accessed via an API interface thereby allowing the class cache to be shared across the guests. By sharing the class cache among the guests, multiple copies of the same class data are no longer necessary thereby optimally using the physical memory on the host.

Abstract: Shared memory and sockets are used to protect shared resources in an environment where multiple operating systems execute concurrently on the same hardware. Rather than using spinlocks for serializing access to the shared resources, when a thread is unable to acquire a shared resource because that resource is already held by another thread, the thread that was unable to acquire the resource creates a socket with which it will wait to be notified that the shared resource has been released. The sockets may be network sockets or in-memory sockets that are accessible across the multiple operating systems; if sockets are not available in a particular implementation, communication technology that provides analogous services between operating systems may be used instead. In an optional aspect, fault tolerance is provided to address socket failures, in which case one or more threads may fall back (at least temporarily) to using spinlocks.

Abstract: A computer-implemented method for referencing a mutable out-of-line monitor for a shared object in read-only memory, comprising creating a lockword with a pointer field set to zero and a pre-set inflated bit when the object is stored in read-only memory. An out-of-line monitor table is referenced when encountering the lockword with the pointer field set to zero and the pre-set inflated bit. The desired monitor referenced in the out-of-line monitor table is identified based on the address of the lockword.