摘要:
A first software entity occupies a portion of a linear address space of a second software entity and prevents the second software entity from accessing the memory of the first software entity. For example, in one embodiment of the invention, the first software entity is a virtual machine monitor (VMM), which supports a virtual machine (VM), the second software entity. The VMM sometimes directly executes guest instructions from the VM and, at other times, the VMM executes binary translated instructions derived from guest instructions. When executing binary translated instructions, the VMM uses memory segmentation to protect its memory. When directly executing guest instructions, the VMM may use either memory segmentation or a memory paging mechanism to protect its memory. When the memory paging mechanism is active during direct execution, the protection from the memory segmentation mechanism may be selectively deactivated to improve the efficiency of the virtual computer system.
摘要:
A computer system has secondary data that is derived from primary data, such as entries in a TLB being derived from entries in a page table. When an actor changes the primary data, a producer indicates the change in a set data structure, such as a data array, in memory that is shared by the producer and a consumer. There may be multiple producers and multiple consumers and each producer/consumer pair has a separate channel. At coherency events, at which incoherencies between the primary data and the secondary data should be removed, consumers read the channels to determine the changes, and update the secondary data accordingly. The system may be a multiprocessor virtual computer system, the actor may be a guest operating system, and the producers and consumers may be subsystems within a virtual machine monitor, wherein each subsystem exports a separate virtual central processing unit.
摘要:
A computer system includes at least one virtual machine that has a plurality of virtual processors all running on an underlying hardware platform. A software interface layer such as a virtual machine monitor establishes traces on primary structures located in a common memory space as needed for the different virtual processors. Whenever any one of the virtual processors generates a trace event, such as accessing a traced structure, then a notification is sent to at least the other virtual processors that have a trace on the accessed primary structure. In some applications, the VMM derives and maintains secondary structures corresponding to the primary structures, such as where the VMM converts, through binary translation, original code intended to run on a virtual processor into code that can be run on an underlying physical processor of the hardware platform. In these applications, the VMM may rederive or invalidate the secondary structures as needed upon receipt of the notification of the trace event. Different semantics are provided for the notification, providing different choices of performance versus guaranteed consistency between primary and secondary structures. In the preferred embodiment of the invention, a dedicated sub-system is included within the VMM for each virtual processor; each sub-system establishes traces, senses trace events, issues the notification, and performs other operations relating specifically to its respective virtual processor.
摘要:
A multiprocessor, multi-program, stop-the-world garbage collection program is described. The system initially over partitions the root sources, and then iteratively employs static and dynamic work balancing. Garbage collection threads compete dynamically for the initial partitions. Work stealing double-ended queues, where contention is reduced, are described to provide dynamic load balancing among the threads. Contention is resolved by using atomic instructions. The heap is broken into a young and an old generation where parallel semi-space copying is used to collect the young generation and parallel mark-compacting the old generation. Speed and efficiency of collection is enhanced by use of card tables and linking objects, and overflow conditions are efficiently handled by linking using class pointers. A garbage collection termination employs a global status word.
摘要:
Run time sampling techniques have been developed whereby representative object lifetime statistics may be obtained and employed to adaptively affect tenuring decisions, memory object promotion and/or storage location selection. In some realizations, object allocation functionality is dynamically varied to achieve desired behavior on an object category-by-category basis. In some realizations, phase behavior affects sampled lifetimes e.g., for objects allocated at different phases of program execution, and the dynamic facilities described herein provide phase-specific adaptation tenuring decisions, memory object promotion and/or storage location selection. In some realizations, reversal of such decisions is provided.
摘要:
With better knowledge of the behavior of objects in a running application, it is possible to improve execution environment decisions that affect management of such objects. For example, if available, object lifetime statistics could be employed in decisions that affect how and where objects are placed, e.g., on allocation or during operation of automatic dynamic memory management facilities such as a garbage collector. Typically, instrumenting all objects to sample lifetimes or other characteristics would impose an impractical level of overhead. We present a technique for dynamic sampling of a subset of allocated objects that incurs low runtime overheads. Coupled with automatic memory management or collection facilities, this technique allows us to improve the efficiency of a collector by segregating objects, sampled and non-sampled alike, based on observed characteristics such as object lifetime. The sampling techniques facilitate tracking of many kinds of object information. For purposes of illustration, an exemplary implementation is described in which such sampling techniques are exploited to improve performance of generational garbage collectors.
摘要:
In accordance with the present invention a method for modifying a sequence of instructions to improve memory management within a storage device during execution of the instructions, comprises the steps, performed by a processor, of (a) analyzing the sequence of instructions for a conflict indicating an undeterminable variable type, (b) determining the type of conflict, and (c) modifying the sequence of instructions to eliminate the conflict based on the determination.
摘要:
A sequence of input language (IL) instructions of a guest system is converted, for example by binary translation, into a corresponding sequence of output language (OL) instructions of a host system, which executes the OL instructions. In order to determine the return address after any IL call to a subroutine at a target entry address P, the corresponding OL return address is stored in an array at a location determined by an index calculated as a function of P. After completion of execution of the OL translation of the IL subroutine, execution is transferred to the address stored in the array at the location where the OL return address was previously stored. A confirm instruction block is included in each OL call site to determine whether the transfer was to the correct or incorrect call site, and a back-up routine is included to handle the cases of incorrect call sites.
摘要:
A checkpointing fault tolerance network architecture enables a backup computer system to be remotely located from a primary computer system. An intermediary computer system is situated between the primary computer system and the backup computer system to manage the transmission of checkpoint information to the backup VM in an efficient manner. The intermediary computer system is networked to the primary VM through a first connection and is networked to the backup VM through a second connection. The intermediary computer system identifies updated data corresponding to memory pages that have been least recently modified by the primary VM and transmits such updated data to the backup VM through the first connection. In such manner, the intermediary computer system holds back updated data corresponding to more recently modified memory pages, since such memory pages may be more likely to be updated again in the future.
摘要:
A computer system that is programmed with virtual memory accesses to physical memory employs multi-bit counters associated with its page table entries. When a page walker visits a page table entry, the multi-bit counter associated with that page table entry is incremented by one. The computer operating system uses the counts in the multi-bit counters of different page table entries to determine where large pages can be deployed effectively. In a virtualized computer system having a nested paging system, multi-bit counters associated with both its primary page table entries and its nested page table entries are used. These multi-bit counters are incremented during nested page walks. Subsequently, the guest operating systems and the virtual machine monitors use the counts in the appropriate multi-bit counters to determine where large pages can be deployed effectively.