-
公开(公告)号:KR101662829B1
公开(公告)日:2016-10-05
申请号:KR1020090112096
申请日:2009-11-19
Applicant: 삼성전자주식회사
CPC classification number: G06F12/0831 , G06F2212/1024 , G06F2212/502
Abstract: 본발명에의한캐시일관성관리장치는, 복수의프로세서들사이에서발생되는핑퐁마이그레이션시퀀스를모니터링하는핑퐁모니터링부, 상기모니터링결과에응답해서상기핑퐁마이그레이션시퀀스의연속발생횟수를카운트하는카운팅부, 그리고상기카운트결과를근거로하여마이그레이션리퀘스트를비-마이그레이토리쉐어링방식의리퀘스트로변경하는리퀘스트변경부를포함한다.
-
公开(公告)号:KR1020120101761A
公开(公告)日:2012-09-17
申请号:KR1020110019766
申请日:2011-03-07
Applicant: 삼성전자주식회사
CPC classification number: G06F12/0895 , Y02D10/13
Abstract: PURPOSE: A cache phase detector and a processor core are provided to effectively utilize critical section cache, thereby reducing power consumption. CONSTITUTION: A counting unit(110) enables a tag corresponding to a request generated in a processor core to generate a critical section miss count by counting a tag miss request which includes an effective cache line based on a tag miss signal which includes critical section cache and a cache line valid signal which includes availability of the cache line corresponding to the request. A signal generation unit(130) compares the critical section miss count with a standard value.
Abstract translation: 目的:提供缓存相位检测器和处理器内核,以有效利用关键部分缓存,从而降低功耗。 构成:计数单元(110)使得与处理器核心中产生的请求对应的标签通过对标签未命中请求进行计数来产生关键部分未命中计数,所述标签未命中包括基于标签未命中信号的有效高速缓存行,所述标签未命中信号包括关键部分高速缓存 以及包括对应于该请求的高速缓存行的可用性的高速缓存行有效信号。 信号生成单元(130)将临界区域错过次数与标准值进行比较。
-
公开(公告)号:KR1020110055179A
公开(公告)日:2011-05-25
申请号:KR1020090112096
申请日:2009-11-19
Applicant: 삼성전자주식회사
CPC classification number: G06F12/0831 , G06F2212/1024 , G06F2212/502 , G06F11/3017
Abstract: PURPOSE: A multiprocessor, cache coherence management apparatus for the same, and method thereof are provided to prevent the generation of unnecessary transaction between multiprocessors. CONSTITUTION: A ping pong monitoring unit(110) monitors a ping pong migration sequence between processors. A counting unit(130) counts the number of the ping pong migration sequences in response to a monitoring result. A request modifying unit(150) changes a migration request into a non-migration sharing method request based on a count result.
Abstract translation: 目的:提供一种用于其的多处理器,高速缓存一致性管理装置及其方法,以防止在多处理器之间产生不必要的事务。 构成:乒乓监视单元(110)监视处理器之间的乒乓移动顺序。 计数单元(130)响应于监视结果对乒乓移动序列的数量进行计数。 请求修改单元(150)基于计数结果将迁移请求改变为非迁移共享方法请求。
-
-