专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明授权

US09519586B2 Methods and apparatus to reduce cache pollution caused by data prefetching 有权
标题翻译：减少数据预取引起的缓存污染的方法和装置
公开(公告)号：US09519586B2
公开(公告)日：2016-12-13
申请号：US13746000
申请日：2013-01-21
申请人： QUALCOMM Incorporated
发明人： Matthew M. Gilbert
IPC分类号： G06F12/10 , G06F12/08 , G06F9/30 , G06F9/32 , G06F9/345 , G06F9/38
CPC分类号： G06F12/0862 , G06F9/30065 , G06F9/325 , G06F9/3455 , G06F9/383 , G06F9/3832
摘要： Efficient techniques are described for reducing cache pollution by use of a prefetch logic that recognizes exits from software loops or function returns to cancel any pending prefetch request operations. The prefetch logic includes a loop data address monitor to determine a data access stride based on repeated execution of a memory access instruction in a program loop. Data prefetch logic then speculatively issues prefetch requests according to the data access stride. A stop prefetch circuit is used to cancel pending prefetch requests in response to an identified loop exit. The prefetch logic may also recognize a return from a called function and cancel any pending prefetch request operations associated with the called function. When prefetch requests are canceled, demand requests, such as based on load instructions, are not canceled. This approach to reduce cache pollution uses program flow information to throttle data cache prefetching.
摘要翻译：描述了用于通过使用识别从软件循环退出或函数返回的预取逻辑来减少高速缓存污染以消除任何未决预取请求操作的有效技术。预取逻辑包括循环数据地址监视器，用于基于程序循环中的存储器访问指令的重复执行来确定数据访问步幅。然后，数据预取逻辑根据数据访问步幅推测地发出预取请求。停止预取电路用于响应于所识别的循环出口取消等待的预取请求。预取逻辑还可以识别来自被调用函数的返回并且取消与被调用函数相关联的任何等待预取请求操作。取消预取请求时，不会取消基于加载指令的请求请求。这种减少缓存污染的方法使用程序流信息来限制数据缓存预取。

2. 发明申请

US20140208039A1 METHODS AND APPARATUS TO REDUCE CACHE POLLUTION CASUED BY DATA PREFETCHING 有权
标题翻译：减少数据预处理引起的高速缓存污染的方法和设备
公开(公告)号：US20140208039A1
公开(公告)日：2014-07-24
申请号：US13746000
申请日：2013-01-21
申请人： QUALCOMM INCORPORATED
发明人： Matthew M. Gilbert
IPC分类号： G06F12/08
CPC分类号： G06F12/0862 , G06F9/30065 , G06F9/325 , G06F9/3455 , G06F9/383 , G06F9/3832
摘要： Efficient techniques are described for reducing cache pollution by use of a prefetch logic that recognizes exits from software loops or function returns to cancel any pending prefetch request operations. The prefetch logic includes a loop data address monitor to determine a data access stride based on repeated execution of a memory access instruction in a program loop. Data prefetch logic then speculatively issues prefetch requests according to the data access stride. A stop prefetch circuit is used to cancel pending prefetch requests in response to an identified loop exit. The prefetch logic may also recognize a return from a called function and cancel any pending prefetch request operations associated with the called function. When prefetch requests are canceled, demand requests, such as based on load instructions, are not canceled. This approach to reduce cache pollution uses program flow information to throttle data cache prefetching.
摘要翻译：描述了通过使用识别从软件循环退出或函数返回来取消任何未决预取请求操作的预取逻辑来减少高速缓存污染的高效技术。预取逻辑包括循环数据地址监视器，用于基于程序循环中的存储器访问指令的重复执行来确定数据访问步幅。然后，数据预取逻辑根据数据访问步幅推测地发出预取请求。停止预取电路用于响应于所识别的循环出口取消等待的预取请求。预取逻辑还可以识别来自被调用函数的返回并且取消与被调用函数相关联的任何等待预取请求操作。取消预取请求时，不会取消基于加载指令的请求请求。这种减少缓存污染的方法使用程序流信息来限制数据缓存预取。

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式