Matrix operation apparatus and digital signal processor...

Details Matrix operation apparatus and digital signal processor... Matrix operation apparatus and digital signal processor...

2000-08-16

2003-01-07

Ellis, Kevin L. (Department: 2186)

Electrical computers and digital processing systems: memory

Address formation

Combining two or more values to create address

C711S217000, C711S218000

Reexamination Certificate

active

06505288

ABSTRACT:

This application claims priority under 35 U.S.C. §§119 and/or 365 to 99-58763 filed in REPUBLIC OF KOREA on Dec. 17, 1999; the entire content of which is hereby incorporated by reference.
BACKGROUND OF THE INVENTION
1 Field of the Invention
The present invention relates to an apparatus for operating on a large amount of data such as a digital signal processor or an operation apparatus for an electronic system, and more particularly, to an operation apparatus of a digital signal processor, capable of processing the instructions represented in a matrix form in order to process the large amount of data and a digital signal processor capable of performing matrix operations.
2. Description of the Related Art
Due to the development of the semiconductor industry, systems whose sizes are extremely small and which can be used for a long time with batteries of small capacity have appeared. In particular, power consumption and the sizes of the systems are the center of interest in a technical field such as personal radio communication systems and adhesive medical instruments, which includes hearing aids.
Digital signal processors, which process a large amount of data, have been significantly developed in the electronics industry, keeping pace with an information-oriented society. However, the power consumption of systems has been rapidly increased since the development has been achieved with priority to the improvement of the performance.
In general, data items are processed one by one in the digital signal processor. For example, when a 2×2 matrix multiplication represented in Equation 1 is done, a general micro processor processes the multiplications one by one.
[
R1
R2
R3
R4
]
=
[
R5
R6
R7
R8
]
×
[
R9
R10
R11
R12
]
t
(
1
)
Equation 1 is represented by a program performed by a general micro processor as follows.
CLR ACC
MULT R
5
R
9
ADD ACC R
5
MULT R
6
R
10
ADD ACC R
6
MV R
1
ACC
CLR ACC
MULT R
5
R
11
ADD ACC R
5
MULT R
6
R
12
ADD ACC R
6
MV R
2
ACC
CLR ACC
MULT R
7
R
9
ADD ACC R
5
MULT R
8
R
10
ADD ACC R
6
MV R
3
ACC
CLR ACC
MULT R
7
R
11
ADD ACC R
5
MULT R
8
R
12
ADD ACC R
6
MV R
4
ACC
In order to reduce the size of the above program, a “MAC” function of performing a multiplication and an addition in one clock period is added to a lot of digital signal processors. The program performed by the digital signal processors to which the “MAC” function is added is represented as follows. It is possible to reduce the time for performing the program and the size of the program by adding the “MAC” function.
CLR ACC
MAC R
5
R
9
MAC R
6
R
10
ADD ACC temp
MV R
1
ACC
CLR ACC
MAC R
5
R
11
MAC R
6
R
12
ADD ACC temp
MV R
2
ACC
CLR ACC
MAC R
7
R
9
MAC R
8
R
10
ADD ACC temp
MV R
3
ACC
CLR ACC
MAC R
7
R
11
MAC R
8
R
12
ADD ACC temp
MV R
4
ACC
It is noted from the above program that only addresses increase and the “MAC” function is periodically repeated. In order to effectively process the program, a repeat statement, “RPT”, and an automatic address increase function (for example, R
5
++) are used by conventional digital signal processors. Such functions can be effectively used for programs for operations such as matrix operations in which a plurality of repeat statements are required. The program to which the repeat statements and the automatic address increase functions are provided can be written as follows.
CLR ACC
RPT
2
MAC R
5
++ R
9
++
ADD ACC temp
MV R
1
ACC
CLR ACC
RPT
2
MAC R
5
++ R
11
++
ADD ACC temp
MV R
2
ACC
CLR ACC
RPT
2
MAC R
7
++ R
9
++
ADD ACC temp
MV R
3
ACC
CLR ACC
RPT
2
MAC R
7
++ R
11
++
ADD ACC temp
MV R
4
ACC
The size of the program performed by conventional digital signal processors is still large in spite of the above-mentioned improvements.
However, when the matrix is expressed using variables, the calculation of Equation 1 can be performed by only the following codes.
SETD
2
2
2
MULT X
1
X
2
X
3
When a certain algorithm or a function is expressed by various instructions, a memory having a large capacity is required. Accordingly, a processor such as the digital signal processor consumes much power. This is because a significant amount of power consumed by the processor is consumed by the memory for reading the instructions.
Also, in a system using a built-in processor where the capacity of the instruction memory is limited and functions expressed by various instructions must be performed, all the required instructions may not be loaded in a chip. In this case, an external memory must be added and used for the system. Accordingly, more power is consumed. Therefore, when a certain function is expressed by various instructions, a time for which products can be used without recharge is reduced, the sizes of the products are enlarged and the prices of the products are raised, since various chips are used in portable electronic systems.
In a system which performs a vector operation such as a systolic array operation, it is possible to process a plurality of operations. In such a system, data items are processed at one time by including various operation apparatuses such as a multiplexer and an adder inside the chip. Such a method can improve the performance of the system, however, does not reduce the power consumption and the size of the chip. Also, the vector operation is effectively used for only some digital signal processing algorithms.
SUMMARY OF THE INVENTION
The present invention relates to a technology of guaranteeing a long operation time with a small amount of power consumption, which is essential for portability of electronic products. A large amount (between 50 and 80%) of power consumed by a processor such as a digital signal processor is consumed by a program memory for reading instructions. Since the power consumption by the memory is proportional to the size of the memory and the number of times that data is read from and written to the memory, it is necessary to reduce the number of times that data is read from the memory and the size of the memory in order to reduce the power consumption.
Since many signal processing algorithms can be simply expressed using a method of matrix representation, it is an object of the present invention to provide a matrix operation apparatus in a digital signal processor capable of reducing the size of the program memory and the number of times that instructions are read from the memory by applying the method of matrix representation for a method of data representation.
It is another object of the present invention to provide a digital signal processor capable of performing matrix operations.
Accordingly, to achieve the first object, there is provided an apparatus included in a digital signal processor, for performing matrix operations, comprising a data storage means for storing operand data comprising matrix data in the form of a circular linked list and operation result data, address generating means for sequentially generating addresses required for performing matrix operations, the addresses including a series of addresses of first operand data, a series of addresses of second operand data, and a series of stored addresses of operation result data, whereby the addresses are sequentially generated according to the contents of the instruction words performed by the digital signal processor, and an operation means for reading data positioned in the address generated by the data storage means and performing operations according to the contents of the instruction words.
The address generating means comprises a matrix address storage means for storing the addresses of the matrix data items in the form of the circular linked list and outputting a first matrix data address, a second matrix data address, and an operation result matrix data address according to the contents of the instruction words, an A address generator for receiving the first matrix data address from the matrix address storage means and sequentially generating the series of addresses of the first operand data according to the instruct

Affiliated with

Also associated with

Rating

Matrix operation apparatus and digital signal processor... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Matrix operation apparatus and digital signal processor..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Matrix operation apparatus and digital signal processor... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3047623