Electrical computers and digital processing systems: memory – Storage accessing and control – Memory configuring
Reexamination Certificate
1999-01-29
2002-06-18
Hudspeth, David (Department: 2651)
Electrical computers and digital processing systems: memory
Storage accessing and control
Memory configuring
C711S163000, C711S159000
Reexamination Certificate
active
06408371
ABSTRACT:
BACKGROUND
The invention relates to computer system memory architectures and, more particularly, to the control of memory access operations based on a memory page cache configured to provide a programmable number of open pages.
Many current computer system memory architectures use synchronous random access memories (synchronous RAM) such as synchronous dynamic random access memory (SDRAM), SyncLink dynamic random access memory (SLDRAM), and Rambus dynamic random access memory (RDRAM). The SyncLink standard has been assigned the tentative designation of IEEE-1596.7 by the Microprocessor & Microcomputer Standards Committee (MMSC) of the Institute of Electrical and Electronics Engineers (IEEE). The Rambus® standard is published by Rambus, Incorporated of Mountain View, Calif.
In addition to providing inherently faster operation than previous types of memories, synchronous RAM may generally be organized into banks. Banks represent a physical compartmentalization of memory space, where each bank may correspond to a unit or array of physical memory. A bank may be further divided into pages, where a page is typically defined in terms of a row address. All those memory locations in a bank having a common row address are said to be on the same page of memory.
One feature of banked memory systems is that consecutive memory access operations to a common page may be performed faster than consecutive memory access operations directed to different pages within the same bank. As shown in
FIG. 1
, the time to perform first access
100
(directed to a first page in a first bank) includes the time needed to select the target page
102
and the time to select the uniquely targeted memory location
104
. If second access
106
is directed to another memory location in the same page, the time required to complete the memory transfer includes that needed to select the target location
108
; no time is required for page selection. If a subsequent, third access
110
is directed to a different page in the same bank however, the previously selected (open) page must be closed (an operation referred to as precharging
112
) before access
110
may proceed. Following precharge operation
112
, access
110
continues through page selection
114
and data selection
116
phases. Because precharge operations require some time to complete, they generally limit the speed with which a sequence of memory access operations may be performed.
As indicated above, by leaving a page open after completing a memory access operation the precharge time penalty may be avoided when a subsequent bank access is directed to that same page (a page hit). Conversely, when a subsequent bank access is to a different page (a page miss), the open page must be closed and the precharge operation performed before the memory access operation may proceed. Thus, while there exists benefits to leaving a page open in the event there are frequent page hits, there also exists time penalties associated with a large number of page misses when pages are kept/left open.
As the number of banks in a memory system increases, the potential for improved memory access bandwidth increases. The potential improvement may generally be attributed to two factors: (1) the ability to avoid precharge when making successive accesses to a common page of memory; and (2) the ability to hide precharge by interleaving memory accesses between different banks. Actual improvement in memory system performance, however, may be less than expected. For example, as the number of memory banks increase so does the amount of hardware needed to track each open page. Thus, the hardware overhead associated with maintaining a large number of pages in the open state may set a practical upper limit on the number of pages that may be simultaneously open. Further, some applications and devices are known to exhibit low degrees of memory access locality—memory access operations by these types of entities do not generally benefit, and may actually slow overall access operations due to frequent page misses. Thus, there is a need for memory systems having improved performance.
SUMMARY
In one embodiment, the invention provides a computer system comprising a memory (having a plurality of pages), a memory controller adapted to maintain a programmable number of the plurality of pages in an open state at one time, and a page limit value (representing the programmable number of pages that can be kept open at one time) accessible to the memory controller. The page limit value may be a user specified value, a function of the total number of banks in the memory, a function of the type of processing performed by the computer system, or it may periodically change during the course of system operations.
REFERENCES:
patent: 4745541 (1988-05-01), Vaniglia et al.
patent: 4937791 (1990-06-01), Steele et al.
patent: 5414658 (1995-05-01), Challa
patent: 5727178 (1998-03-01), Pletcher et al.
patent: 5748551 (1998-05-01), Ryan et al.
patent: 5784699 (1998-07-01), McMahon et al.
patent: 5825878 (1998-10-01), Takahashi et al.
patent: 5848025 (1998-12-01), Marietta et al.
patent: 5889714 (1999-03-01), Schumann et al.
patent: 6009019 (1999-12-01), Fang
patent: 6052134 (2000-04-01), Foster
patent: 6052756 (2000-04-01), Barnaby et al.
patent: 6078994 (2000-06-01), Carey
patent: 6182089 (2001-01-01), Ganapathy et al.
patent: 6189062 (2001-02-01), Jander et al.
patent: 6247107 (2001-06-01), Christie
Intel, “Intel 440BX AGP set: 82443BX Host Bridge/Controller” Datasheet, entirety. (Apr., 1998).
David A. Patterson and John L. Hennessy, “Computer Architecture A Quantitative Approach” Second Addition, Morgan Kaufmann Publishers, pp. 439-466. (1996).
Inventor, Joseph M. Jeddeloh, U.S. Patent Application No. 09/239,911; Filed Jan. 29, 1999, entitled “Method to Access Memory Based on a Programmable Page Limit”.
Inventor, Joseph M. Jeddeloh, U.S. Patent Application No. 09/239,911; Office Action Mailed Nov. 6, 2001.
Keith Diefendorff, Jalapeno Powers Cyrix's M3, Microdesign Resources, Microprocessor Report, pp. 24-30, Nov. 1998.
Hudspeth David
Trop Pruner & Hu P.C.
Tzeng Fred F.
LandOfFree
Device to access memory based on a programmable page limit does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Device to access memory based on a programmable page limit, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Device to access memory based on a programmable page limit will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2893438