摘要:
A method for committing memory transactions in an application that includes executing a plurality of sections of the application in parallel, logging a plurality of memory transactions that occur while executing the plurality of sections to obtain a plurality of logs and a plurality of temporary results, wherein the plurality of memory transactions that includes a plurality of writes to at least one memory location, comparing the plurality of logs to identify an optimal list of writes from the plurality of writes, and committing memory transactions corresponding to a subset of the plurality of temporary results, wherein the subset of the plurality of temporary results is identified by the optimal list of writes.
摘要:
A computer system may be configured to dynamically select a memory virtualization and corresponding virtual-to-physical address translation technique during execution of an application and to dynamically employ the selected technique in place of a current technique without re-initializing the application. The computer system may be configured to determine that a current address translation technique incurs a high overhead for the application's current workload and may be configured to select a different technique dependent on various performance criteria and/or a user policy. Dynamically employing the selected technique may include reorganizing a memory, reorganizing a translation table, allocating a different block of memory to the application, changing a page or segment size, or moving to or from a page-based, segment-based, or function-based address translation technique. A selected translation technique may be dynamically employed for the application independent of a translation technique employed for a different application.
摘要:
A computer system may be configured to dynamically select a memory virtualization and corresponding virtual-to-physical address translation technique during execution of an application and to dynamically employ the selected technique in place of a current technique without re-initializing the application. The computer system may be configured to determine that a current address translation technique incurs a high overhead for the application's current workload and may be configured to select a different technique dependent on various performance criteria and/or a user policy. Dynamically employing the selected technique may include reorganizing a memory, reorganizing a translation table, allocating a different block of memory to the application, changing a page or segment size, or moving to or from a page-based, segment-based, or function-based address translation technique. A selected translation technique may be dynamically employed for the application independent of a translation technique employed for a different application.
摘要:
A method for detecting a dependence violation in an application that involves executing a plurality of sections of the application in parallel, and logging memory transactions that occur while executing the plurality of sections to obtain a plurality of logs and a plurality of temporary results, where the plurality of logs is compared while executing the plurality of sections to determine whether the dependence violation exists.
摘要:
A system and method for allocating the nearest available physical memory in a distributed, shared memory system. In various embodiments, a processor node may broadcast a memory request to a first subset of nodes connected to it via a communication network. In some embodiments, if none of these nodes is able to satisfy the request, the processor node may broadcast the request to additional subsets of nodes. In some embodiments, each node of the first subset of nodes may be removed from the processor node by one network hop and each node of the additional subsets of nodes may be removed from the processor node by no more than an iteratively increasing number of network hops. In some embodiments, the processor node may send an acknowledgment to one node that can fulfill the request and a negative acknowledgement to other nodes that can fulfill the request.
摘要:
The use of a token-based memory protection technique may provide memory protection in a computer system employing memory virtualization. A token-based memory protection technique may include assigning a unique identifier to an application, process, or thread, and associating the identifier with a block of memory allocated to that application, process, or thread. Subsequent to assigning the identifier, a packet requesting access to that block of memory may include a token to be compared to the identifier. A memory controller may be configured to associate the identifier with the block of memory and to compare the token in the memory request packet to the identifier before granting access. If a second block of memory is subsequently allocated to the application, process, or thread, the identifier may be disassociated with the first block of memory and associated with the second block of memory.
摘要:
A method for detecting a dependence violation in an application that involves executing a plurality of sections of the application in parallel, and logging memory transactions that occur while executing the plurality of sections to obtain a plurality of logs and a plurality of temporary results, where the plurality of logs is compared while executing the plurality of sections to determine whether the dependence violation exists.
摘要:
A method for executing a loop in an application that includes executing iterations in a first segment of the loop by a base thread, logging memory transactions that occur during execution of iterations in the first segment by a co-inspector thread to obtain a co-inspector log, executing iterations in a second segment of the loop by a co-thread to obtain temporary results, logging memory transactions that occur during execution of iterations in the second segment to obtain a co-thread log, and comparing the co-inspector log and the co-thread log to determine whether a thread interdependency exists.
摘要:
A system for mapping memory of a computer system includes a sorter, a partitioner, and a mapper. The sorter sorts a physical memory space to obtain a sorted virtual memory space, where the physical memory space includes a plurality of memory patterns, wherein each of the plurality of memory patterns is associated with a physical device and includes at least one memory block, wherein the at least one memory block is contiguous and addressable, and the plurality of memory patterns are arranged based on a predefined number of occurrences of each memory pattern, size of contiguous memory address ranges in each of the plurality of memory patterns, size of contiguous memory address ranges in each of the plurality of memory patterns, and arrangement of contiguous address ranges in each of the plurality of memory patterns. The partitioner partitions the plurality of memory patterns in the sorted memory space based on a partition size to obtain a partitioned virtual memory space, and a mapper transposes the partitioned virtual memory space to obtain a system memory space used by the computer system.
摘要:
A system for mapping memory of a computer system includes a sorter, a partitioner, and a mapper. The sorter sorts a physical memory space to obtain a sorted virtual memory space, where the physical memory space includes a plurality of memory patterns, wherein each of the plurality of memory patterns is associated with a physical device and includes at least one memory block, wherein the at least one memory block is contiguous and addressable, and the plurality of memory patterns are arranged based on a predefined number of occurrences of each memory pattern, size of contiguous memory address ranges in each of the plurality of memory patterns, size of contiguous memory address ranges in each of the plurality of memory patterns, and arrangement of contiguous address ranges in each of the plurality of memory patterns. The partitioner partitions the plurality of memory patterns in the sorted memory space based on a partition size to obtain a partitioned virtual memory space, and a mapper transposes the partitioned virtual memory space to obtain a system memory space used by the computer system.