公开(公告)号：US20050131865A1

公开(公告)日：2005-06-16

申请号：US10989704

申请日：2004-11-15

申请人： Terry Jones ,  Pythagoras Watson ,  William Tuel ,  Larry Brenner ,  Patrick Caffrey ,  Jeffrey Fier

发明人： Terry Jones ,  Pythagoras Watson ,  William Tuel ,  Larry Brenner ,  Patrick Caffrey ,  Jeffrey Fier

IPC分类号： G06F9/48 ,  G06F17/30

CPC分类号： G06F9/4881 ,  G06F9/4825

摘要： In a parallel computing environment comprising a network of SMP nodes each having at least one processor, a parallel-aware co-scheduling method and system for improving the performance and scalability of a dedicated parallel job having synchronizing collective operations. The method and system uses a global co-scheduler and an operating system kernel dispatcher adapted to coordinate interfering system and daemon activities on a node and across nodes to promote intra-node and inter-node overlap of said interfering system and daemon activities as well as intra-node and inter-node overlap of said synchronizing collective operations. In this manner, the impact of random short-lived interruptions, such as timer-decrement processing and periodic daemon activity, on synchronizing collective operations is minimized on large processor-count SPMD bulk-synchronous programming styles.

摘要翻译： 在包括具有至少一个处理器的SMP节点的网络的并行计算环境中，用于提高具有同步集合操作的专用并行作业的性能和可扩展性的并行感知协同调度方法和系统。 该方法和系统使用全局共同调度器和操作系统内核调度器，其适于协调节点和节点之间的干扰系统和守护进程活动，以促进所述干扰系统和守护进程活动的节点内和节点间重叠，以及 所述同步集合操作的节点内和节点间重叠。 以这种方式，在大型处理器数SPMD批量同步编程风格下，随机短暂中断（如定时器递减处理和周期守护程序活动）对同步集合操作的影响最小化。

公开(公告)号：US07810093B2

公开(公告)日：2010-10-05

申请号：US10989704

申请日：2004-11-15

申请人： Terry R. Jones ,  Pythagoras C. Watson ,  William Tuel ,  Larry Brenner ,  Patrick Caffrey ,  Jeffrey Fier

发明人： Terry R. Jones ,  Pythagoras C. Watson ,  William Tuel ,  Larry Brenner ,  Patrick Caffrey ,  Jeffrey Fier

IPC分类号： G06F9/46

CPC分类号： G06F9/4881 ,  G06F9/4825

摘要： In a parallel computing environment comprising a network of SMP nodes each having at least one processor, a parallel-aware co-scheduling method and system for improving the performance and scalability of a dedicated parallel job having synchronizing collective operations. The method and system uses a global co-scheduler and an operating system kernel dispatcher adapted to coordinate interfering system and daemon activities on a node and across nodes to promote intra-node and inter-node overlap of said interfering system and daemon activities as well as intra-node and inter-node overlap of said synchronizing collective operations. In this manner, the impact of random short-lived interruptions, such as timer-decrement processing and periodic daemon activity, on synchronizing collective operations is minimized on large processor-count SPMD bulk-synchronous programming styles.

摘要翻译： 在包括具有至少一个处理器的SMP节点的网络的并行计算环境中，用于提高具有同步集合操作的专用并行作业的性能和可扩展性的并行感知协同调度方法和系统。 该方法和系统使用全局共同调度器和操作系统内核调度器，其适于协调节点和节点之间的干扰系统和守护程序活动，以促进所述干扰系统和守护程序活动的节点内和节点间重叠，以及 所述同步集合操作的节点内和节点间重叠。 以这种方式，在大型处理器数SPMD批量同步编程风格下，随机短暂中断（如定时器递减处理和周期守护程序活动）对同步集合操作的影响最小化。

公开(公告)号：US20050078605A1

公开(公告)日：2005-04-14

申请号：US10685144

申请日：2003-10-14

申请人： Su-Hsuan Huang ,  William Tuel

发明人： Su-Hsuan Huang ,  William Tuel

IPC分类号： H04L29/06 ,  H04L29/08 ,  H04L12/26

CPC分类号： H04L29/06 ,  H04L69/164 ,  H04L69/326

摘要： Messages arriving at a receiver are managed to ensure proper ordering of the messages. To facilitate proper ordering, a message sequence number is used, as well as matching criteria to match a correctly sequenced message with a posted receive. In response to processing a message, a check is made as to whether previously out of order messages can now be processed.

摘要翻译： 到达接收机的消息被管理以确保消息的正确排序。 为了方便正确的排序，使用消息序列号，以及匹配标准以将正确排序的消息与发布的接收匹配。 响应于处理消息，检查是否可以处理先前不按顺序的消息。