Programmable multi-tasking memory management system

Details Programmable multi-tasking memory management system Programmable multi-tasking memory management system

1999-11-19

2003-02-25

Myers, Paul R. (Department: 2781)

Electrical computers and digital data processing systems: input/

Access arbitrating

Decentralized arbitrating

C710S027000

Reexamination Certificate

active

06526462

ABSTRACT:

FIELD OF THE INVENTION
The present invention relates to a programmable multi-tasking memory management system, and more particularly, to a system and method for managing requests for memory associated with a system on chip (SOC) architecture. The multi-tasking memory management system of the present invention includes a routing controller or central processing unit (RCPU) that is used for routing/switching high rate stream data between communication cores and digital signal processors with minimum reliance and demand on a virtual central processing unit (VCPU) running the application software and the system bus. The present invention is further directed to a system and method for partitioning tasks between the RCPU and the VCPU within the SOC architecture.
BACKGROUND OF THE INVENTION
A system on chip (SOC) product provides many advantages and benefits over a traditional, separate component integrated circuit (IC) product. The separate IC product generally includes components that are connected to each other on a printed circuit board. Alternatively, an SOC product is designed such that an entire system (processors, memory, logic, clock, I/O control unit, etc.) can be implemented or embedded on a single chip, thereby producing a product that is smaller, faster, and more efficient than the separate IC product. Each SOC product has at least the following three components in common: embedded processor (e.g., ARM, LEXRA, MIPS, ARC, DSP core); memory; and logic.
Using SOC technology, the overall size of the end product is reduced because manufacturers can put major system functions on a single chip, as opposed to putting them on multiple chips. This reduces the total number of chips needed for the end product.
In addition, SOC products provide faster chip speeds due to the integration of the components/functions into one chip. Many applications such as high-speed communication devices (VoIP, MoIP, wireless) require chip speeds that may be unattainable with separate IC products. This is primarily due to the physical limitations of moving data from one chip to another, through bonding pads, wires, buses, etc. Integrating chip components/functions into one chip eliminates the need to physically move data from one chip to another, thereby producing faster chip speeds. Further, the SOC product consumes less power than the separate IC product since data do not need to be moved from one chip to another.
Another advantage of using the SOC product is that it less expensive for the manufacturer because of the reduced number of chips used in the end product. Packaging costs, which can be significant, are likewise reduced as a result of having fewer chips. Thus, SOC products are becoming ever more popular and are widely used in many applications such as in Internet products.
FIG. 1
illustrates a block diagram of a conventional SOC architecture. In the conventional SOC architecture, there may be multiple processors such as DSP
2
and CPU
4
connected to a system bus
24
. Only two such processors are illustrated herein, but it is understood that multiple DSPs, CPUs, or any other kinds of processors can be used, which are also connected to the system bus
24
. Other functions/devices that are connected to the system bus
24
include DMA (Direct Memory Access)
6
, GPIO (general purpose I/O unit)
8
, arbiter
10
, interrupt controller
12
, and internal/external memory
16
. Other conventional devices, which are not illustrated herein, may also be connected to the system bus
24
. A bridge
14
can further be used to connect the system bus
24
to a peripheral bus
26
. The peripheral bus
26
connects lower rate stream data communication cores such as MAC 10/100 Ethernet 18, AC97 20, and USB 2.0 UDC 22, and the like for concurrent and independent operation from the devices that are directly connected to the system bus
24
.
During operation, the peripheral devices (MAC 10/100 Ethernet 18, AC 97 20, USB 2.0 UDC 22) using the bridge
14
will interrupt the processor (i.e., CPU
4
) and attempt to become “masters” of the system bus
24
in order to access the internal/external memory
16
using the DMA
6
. The DMA
6
is a direct memory access device that allows a peripheral device (master) to access the internal/external memory
16
without requiring the assistance of the processor (i.e., CPU
4
) on the system bus
24
. The DMA
6
will generally use an internal 32 bit FIFO for temporary storage of the DMA data. Source and destination addresses can be aligned on any byte address boundary using this method. When the peripheral master occupies the system bus
24
and interfaces with the memory
16
for an extended period of time, a time-out feature can be used to break off the connection between the peripheral master and the system bus
24
to allow the processors
2
,
4
to access the system bus
24
.
Alternatively, the CPU
4
, in the case of a cache line-miss, will read the cache line from the internal/external memory
16
. In the case where multiple masters (e.g., DSP
2
and CPU
4
) attempt to access the internal/external memory
16
simultaneously, there will likely be conflicts and so-called “bottleneck” problems. Such problems occur because the requests from the masters will be transmitted simultaneously to the internal/external memory
16
, and conventional systems will not be able to process such requests at the same time. The arbiter
10
will essentially control the arbitration and scheduling scheme of the masters so that priority is given to a particular master on the system bus
24
. In other words, the arbiter
10
will decide which master will control the system bus
24
at a given time. One particular master will have control over the system bus
24
at the given time and will prevent other masters, including peripheral masters, from accessing the system bus
24
.
One major problem associated with such a conventional SOC system is that many masters would be required to read/write from/to the internal/external memory
16
during the same clock cycle, which is not possible and thus will cause delays and conflicts among them. At times, the masters will attempt to read/read from/to a different location (i.e., memory banks) on the memory, while at other times, the masters may attempt to read/write from/to the same location (i.e., memory bank).
Another problem associated with the conventional SOC system is that there may be multiple peripheral masters or communication cores (e.g., 30 or more) in the SOC architecture requesting access to the memory
16
. For each memory request from the peripheral master, the CPU
4
will need to process an interrupt, read/write data length and availability registers and connect the peripheral bus
26
to the system bus
24
. As a result, the CPU
4
may not be able to efficiently perform general system/application tasks since each interrupt consumes tens or hundreds of CPU
4
clock cycles.
Thus, there is a need for a highly intelligent memory management system that allows the main or virtual CPU (VCPU) to perform system/application tasks without having to perform memory routing/switching tasks requested by the peripheral devices. There is also a need to keep all the masters in the SOC architecture satisfied by processing and/or predicting the memory requests and enabling masters to update their internal data once new data is written in the memory.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a programmable multi-tasking memory management system.
It is another object of the present invention to provide a system and method for managing requests for internal/external memory associated with a system on chip architecture.
It is a further object of the present invention to provide a system and method for simultaneously processing multiple memory requests from multiple masters using the multi-tasking memory management system.
It is yet another object of the present invention to provide a system and method for transmitting memory requests from multiple masters to the multi-tasking memory management system using dedicated m

Affiliated with

Also associated with

Rating

Programmable multi-tasking memory management system does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Programmable multi-tasking memory management system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Programmable multi-tasking memory management system will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3175423