JPS62273440A - Nuclear magnetic resonator - Google Patents

Abstract

PURPOSE:To execute a matrix computation efficiently, by employing the sum of the product between an output of a step register and one output o a plurality of address registers and outputs of the remaining address registers as memory address. CONSTITUTION:Prior to the accessing of a memory 4, the size of a line of a matrix data to be handled is specified for a matrix unit register 8 with a CPU1. For example, (m) is set in a matrix with the size of (m) lines and (n) columns as data to be handled. Then, when accessing the memory 4, the CPU1 specifies an address value for a line address register 3 or a column address register 7 and the assignment of a reading or writing is done through a memory control 2 to execute the accessing. An output Y of the register 7 and an output (m) of the register 8 are both inputted into a multiplier 9. The resulting product is inputted into an adder 10 together with an output X of the register 3 and the sum thereof will be a final memory address. In this manner, the computation of the matrix data can be simplified in the execution program as the data of line and column can be handled on equal terms thereby enabling efficient execution of the matrix computation.

Description

[Detailed Description of the Invention] 3. Detailed Description of the Invention [Field of Industrial Application] The present invention is directed to nuclear magnetic resonance 4! i:e, particularly the Fourier transform form of nuclear magnetic resonance.

[Conventional technology]

New measurement methods such as the two-dimensional Fourier transform method are being developed in Fourier transform nuclear magnetic resonance systems, and there are many opportunities to handle matrix data, and the data size is becoming extremely large.

Therefore, such a process efficiently executes L-V.
('t'1% A memory system of a type suitable for matrix operations is required.

Conventionally, it is common to have one memory 7 stem V('/"1.1.1 address register. An example of such a system is shown in FIG. 2.

1rr central fgll-fuuai, 2a memory controller,
3a address register, 4 memory, 5ri write,)
Gate buffer, 6 read gate buffer.

When accessing one memory, first the central song 1 play device 1 (
The CPU (hereinafter referred to as CPU) sets an address in an address register. CPU] i, next c memory tip a-ra 2
? A read/write command is issued through P, and access is made to line f-'4. For write operation, the write data is
Prior to the write command, the data is transferred to the memory data bus through the write gate buffer 5, and written to the external memory in response to the write command. 1 in case of read@production Data is sent via read gate buffer 6? Uke is taken.

[Problem that the invention seeks to solve]

However, when handling matrix data in the above memory system, it is necessary to create address data f in the executing program every time it is accessed, which makes the data extremely sloppy and reduces reliability. The problem was that the execution speed was also slowed down. In other words, if we consider addressing the data in the X row and Y column in a matrix theta with m rows and n columns, (the sum of (X-1-YXml) - g is calculated repeatedly every time in the execution program When calculating row data in this formula, it is simply compared in the form (X + constant), but when calculating column data, it becomes (constant + YXm).
It takes the form of

This method requires different address calculation power for row operations and column operations, so it is not possible to share arithmetic programs.

Therefore, an object of the present invention is to provide a nuclear magnetic resonance apparatus which is fx based on such a fact and is capable of efficiently performing matrix operations.

[Means for solving problems]

In order to accomplish this purpose, the present invention provides jet 1
A nuclear magnetic resonance apparatus t is equipped with a data processing system that stores No. A memory system in which the memory address is the product of the output of the register and the output of one of the above address registers, and the sum of the output of the remaining other address registers? I have worked hard to prepare for it.

〔Example〕

FIG. 1074 shows an embodiment of the nuclear magnetic resonance apparatus according to the second aspect of the present invention. Items having the same reference numerals as those in FIG. 1 have the same functions.

Configuration rtcPUx> and memory controller a different from Figure 1
- a column address register 7 and a row unit register 8 which receive inputs from λ, and a multiplier 9 for each of these outputs.
It is designed to be input to .

The output of the multiplier 9 and the output of the address register 3 are input to the memory 4 via the adder 10.

Prior to accessing the memory 7, the CPU specifies the row size of the matrix data to be handled in the row unit register 8. For example, if the data to be handled is a matrix of size m rows and n columns, set m. When accessing the memory next time, the address value is specified in the CPUIH row address register or the a column address register. Next, read or write is specified through the memory controller and access is executed.

The output of the column address register 7 and the output of the row unit register δ are both input to the multiplier 9 and multiplied. This embezzlement is input to the adder 10 and summed with the row address register. This sum value becomes the final memory address.

In other words, in matrix data of m rows and n columns, when accessing one notator in X row and Y column, i is set in row unit register 8Knm, row address register 3vcnX, and column address register
Y is set and the final memory address is (X0YXm
).

Row data? HX when accessing T, iY when accessing column data, 0, 2, 3 . ... and town fK
It becomes executable by T. Note that when dealing with one-dimensional data, 6 can be done by setting "02" in the row unit register.

The above procedure is to perform the light and matrix data using 7 stems of memory.
i, row and column data can be treated completely equally, and the execution program can be simply written as f.

As a result, the execution speed of the program is significantly improved, and the reliability is also improved.

〔Effect of the invention〕

From the above explanation, it is clear that if the nuclear magnetic resonance device invented by KL is used, matrix calculations can be performed. Be able to execute efficiently.

[Brief explanation of the drawing]

The first factor is an embodiment of the nuclear magnetic resonance apparatus according to the present invention.
FIG. 2 is a block diagram showing an example of a conventional nuclear magnetic resonance apparatus. DESCRIPTION OF SYMBOLS 1...Central I collapse device, 2...Memory controller, 3...Address register, 4...Memory, 5...
・Write gate buffer, 6...Read gate buffer, 7...Z I 1\J

Claims (1)

[Claims] 1. A nuclear magnetic resonance apparatus equipped with a data processing system that stores a detection signal in a memory and performs Fourier transform to convert it into a frequency spectrum, the data processing system comprising:
Equipped with multiple address registers and step registers,
A nuclear magnetic resonance apparatus characterized by comprising a memory system whose memory address is the product of the output of the step register and the output of one of the address registers, and the sum of the outputs of the remaining other address registers.