首页> 外国专利> FORWARD PROGRESS ON RETRIED SNOOP HITS

FORWARD PROGRESS ON RETRIED SNOOP HITS

机译:重试鼻快照的前瞻性进展

摘要

The present invention relates to re-execute the snoop forward progress performing the method and system of the heat upon. When the apparatus for snooping the system bus of a multiprocessor system detects an operation requesting the data present in the local memory in a predetermined consistency condition that requires that data is provided from the device, the device attempts to intervene. When disturbed by a second device that the intervention of re claims, the apparatus sets the flag to provide historical information about the failed intervention. In the same cache if a subsequent snoop hit during device for a location which is again assert intervention and rerun the snooping operation again, the device is required cache entries towards the end coherency state is expected to be corresponding to the result of the initial operation requiring the cache entry the start of the operation to correct the inconsistent state. When requesting cache entry including the modified data existing in the local memory of the device, the operation may include a push operation to record the request cache entry to the system memory. This push operation may be snooping by another device from the system bus to update the local memory of the other device. If you are not a modification request cache entry states, including the data in a predetermined consistent state, the operation can simply be included to modify the coherency state to share or disabled. When the initial operation is a read operation, the state transition of the cache proceeds in accordance with a sequence that ends in a shared state. If it contains an indication of the intent to modify the request cache entry as read in the initial operation intended to be modified, consistent state transition proceeds in accordance with a sequence that ends in an invalid state. May be used to determine the time when a single failure involved (single failed intervention) flag instead of the reflection counter is sustainable lock the threshold number failed intervention indicating that generation of the excess.
机译:本发明涉及重新执行探听前进进程,以执行加热的方法和系统。当用于监听多处理器系统的系统总线的设备在要求要求从设备提供数据的预定一致性条件下检测到请求存在于本地存储器中的数据的操作时,设备尝试进行干预。当受到请求干预的第二设备的干扰时,该设备设置标记以提供关于失败干预的历史信息。在同一个高速缓存中,如果在设备期间针对某个位置的后续侦听命中再次断言介入并再次重新运行侦听操作,则要求设备朝向最终一致性状态的高速缓存条目预计与初始操作要求的结果相对应高速缓存条目纠正错误状态的操作的开始。当请求包括存在于设备的本地存储器中的已修改数据的高速缓存条目时,该操作可以包括将请求高速缓存条目记录到系统存储器中的推入操作。此推送操作可能正在被另一个设备从系统总线监听,以更新另一个设备的本地内存。如果您不是修改请求缓存项状态,包括处于预定一致状态的数据,则可以简单地包括该操作以修改一致性状态以共享或禁用。当初始操作是读取操作时,高速缓存的状态转换根据以共享状态结束的序列进行。如果它包含意图在要修改的初始操作中读取的修改请求缓存项的意图的指示,则一致状态转换将按照以无效状态结束的序列进行。可用于确定涉及单个故障(单个故障干预)标志而不是反射计数器的可持续锁定时间,该阈值数目指示失败干预的发生,表明发生了过多的故障。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号