Browse Prior Art Database

Method of Forced Bus Interleaving

IP.com Disclosure Number: IPCOM000117293D
Original Publication Date: 1996-Jan-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 2 page(s) / 63K

Publishing Venue

IBM

Related People

Getzlaff, KJ: AUTHOR [+5]

Abstract

Disclosed is a bus protocol for multiprocessor computer systems, optimizing the data transfer throughput of the memory bank(s) of the system by allowing a command gap between two line transfers, thus increasing the utilization of the bus and improving the overall system's throughput.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 77% of the total text.

Method of Forced Bus Interleaving

      Disclosed is a bus protocol for multiprocessor computer
systems, optimizing the data transfer throughput of the memory
bank(s) of the system by allowing a command gap between two line
transfers, thus increasing the utilization of the bus and improving
the overall system's throughput.

      A known method of bus interleaving disclosed in (*), increases
the bus performance of a single or a multiprocessor computer system
by accessing different independent memory banks, like dynamic or
static random access memories, simultaneously.  Most main storages
are organized in two or more memory banks which can be accessed
nearly simultaneously by a memory controller.  So it is possible to
start a data access to the first bank, and after a few cycles, before
the requested data is available, to start a second data access to the
second bank.

      Fig. 1 shows the advantage of the known bus interleaving scheme
for a second line-fetch operation.  With no interleaving, a second
line-fetch would start in cycle 22 instead of cycle 7.  Due to the
data transfer of the first fetch operation, the second has an initial
access which is only two cycles longer than the fastest possible
access to the memory.  Fig. 1 shows also that additional memory
requests of other processors have to wait until both fetch operations
are completed.  A new fetch command then starts again with the normal
access time to the memory.

      According to Fig....