JPH056392A - System for selecting fourier transformation computing method - Google Patents

Abstract

PURPOSE:To automatically select an arithmetic method having the shortest calculation time to the combination of inputted data points. CONSTITUTION:A switching point table storing the multiplicity of one- dimensional complex Fourier transformation corresponding to the data length of input data is prepared and the multiplicity of the inputted data is compared with that obtained by referring the switching point table while referring the table upon the data length of data inputted at the time of operating multiple one-dimensional complex Fourier transformation. When the former is smaller or equal than/to the latter, a multiple one-dimensional complex Fourier transformation computing part based upon an one-dimensional scanning method is selected, and when the former is larger than the latter, a multiple one- dimensional complex Fourier transformation computing part incorporating the multiplication of one-dimensional complex Fourier transformation in the interior of one-dimensional Fourier transformation is selected, as the feature of this system.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for selecting a Fourier transform calculation method, and more particularly to a method for selecting a Fourier transform calculation method in calculation of multiple one-dimensional complex Fourier transform using a vector computer.

[0002]

2. Description of the Related Art Conventionally, this kind of Fourier transform operation has been performed by using a multiple one-dimensional complex Fourier transform by a one-dimensional scanning method, or by multiplexing a one-dimensional complex Fourier transform, regardless of the combination of the number of input data. The method was fixed to one of the multiple first-order complex Fourier transforms incorporated in.

[0003]

In the above-mentioned conventional Fourier transform calculation method selection method, since the calculation method is fixed, the calculation time is the shortest for the combination of the data points, and the optimum calculation is performed. There was no possibility.

Therefore, the user uses a vector computer to improve the program for the combination of the number of input data, measures the calculation time in advance, and determines the calculation method according to the measurement result in order to perform the optimum calculation. However, there is a drawback that it requires a long time and a lot of labor.

[0005]

According to the Fourier transform calculation method selection method of the present invention, a switching point table storing the multiplicity of the one-dimensional complex Fourier transform is provided corresponding to the data length of the input data, and the multiple one-dimensional complex Fourier transform is provided. By referring to the switching point table according to the data length of the data input at the time of calculation, the multiplicity of the input data is compared with the multiplicity obtained by the reference, and if the former is less than or equal to the latter, A multiple 1-dimensional complex Fourier transform operation unit by the 1-dimensional scanning method, if the former is larger than the latter,
The present invention is characterized in that each of the multiple 1-dimensional complex Fourier transform operation units in which the multiplexing of the one-dimensional complex Fourier transform is incorporated into the one-dimensional Fourier transform is selected.

[0006]

The present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention.

Reference numeral 2 is a multiplex one-dimensional complex Fourier transform operation unit by the one-dimensional scan method, 3 is a multiplex one-dimensional complex Fourier transform operation unit incorporating multiplexing inside the one-dimensional complex Fourier transform, and 1 is the number of input data points. 1 by the combination of
Optimal calculation method for automatically selecting the multiplex one-dimensional complex Fourier transform calculation unit 2 by the three-dimensional scanning method and the multiplex one-dimensional complex Fourier transform calculation unit 3 in which multiplexing of the one-dimensional complex Fourier transform is incorporated in the Fourier transform It is an automatic selection unit.

The functions of the respective parts will be described below in order.

First, data is input to the optimum calculation method automatic selection unit 1 of FIG. The input data has a format as shown in FIG. Each data of the one-dimensional complex Fourier transform is stored in INCN and each data string has an interval I
It is stored in each NCM and multiplexed. INCN,
INCM is a value input by the user.

The optimum calculation method automatic selection unit 1 analyzes the input data set. Here, if the input 1
The data length of the dimensional complex Fourier transform is N, and the number of multiplexing of the one-dimensional complex Fourier transform is M.

The optimum calculation method automatic selection section 1 refers to the switching point table according to the values of the input data length n and the multiplicity m, and the multiple one-dimensional complex Fourier transform calculation section 2 by the one-dimensional scanning method One of the multiplex one-dimensional complex Fourier transform operation units 3 in which multiplexing is incorporated in the dimensional complex Fourier transform is selected, and the input data is passed to the selected operation unit.

The switching point table is shown in FIG.
FIG. 4 is a graph showing the relationship between the values of and the calculation method selected at that time. When the data length of the input data is n and the multiplicity of the one-dimensional complex Fourier transform is m,
Multiplicity M when the data length of the switching point table is N = n
When the input multiplicity m is smaller than or equal to the value of, the input data is passed to the multiplex one-dimensional complex Fourier transform operation unit 2 by the one-dimensional scan method, and the multiplex one-dimensional complex by the one-dimensional scan method is given. The operation is calculated by the Fourier transform, and the result is output.

On the contrary, the multiplicity M of the switching point table
When the value of is smaller than the input value of the multiplicity m, the input data is passed to the multiplex one-dimensional complex Fourier transform calculation unit 3 in which the multiplexing of the one-dimensional complex Fourier transform is incorporated in the Fourier transform, and the one-dimensional complex Fourier transform is performed. The operation is calculated by the multiple one-dimensional complex Fourier transform in which the multiplex of the Fourier transform is incorporated into the Fourier transform, and the result is output.

For example, when n = 4 and m = 3, the value of M in the switching point table when N = 4 is 1, so that the multiplicity of the one-dimensional complex Fourier transform is obtained from the input multiplicity m = 3> 1. The input data (n = 4, m = 3) is passed to the multiplex one-dimensional complex Fourier transform operation unit 3 in which the multiplexing is incorporated in the Fourier transform, and the multiplex 1 in which the multiplexing of the one-dimensional complex Fourier transform is incorporated in the Fourier transform is performed. The operation is calculated by the dimensional complex Fourier transform.

In the multiple one-dimensional complex Fourier transform calculation unit 2 by the one-dimensional scan method, the data input as continuous data is taken out as INCN and one-dimensional data by the one-dimensional scan method, and the one-dimensional complex Fourier transform is calculated by one.
Multiplex 1 times by the method of calculating the number of times of multiplexing one by one
Compute the dimensional complex Fourier transform.

In the multiplex one-dimensional complex Fourier transform calculation unit 3 in which the one-dimensional complex Fourier transform multiplexing is incorporated into the Fourier transform, the data input as continuous data is taken out and the one-dimensional complex Fourier transform multiplexing is repeated. 1
The calculation in which the repetition for multiplexing the operation of the one-dimensional complex Fourier transform and the repetition for multiplexing the operation of the one-dimensional complex Fourier transform are exchanged is performed.

The results calculated by the arithmetic units 2 and 3 are output respectively.

[0019]

As described above, according to the present invention, in the multiple one-dimensional complex Fourier transform, the optimum calculation method for the vector computer is automatically calculated by the combination of the input data. By making a selective selection, it is possible to measure and compare the calculation time for each combination of data, thereby making it unnecessary to search for a calculation method with the shortest calculation time.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is a structural diagram of input multiple 1-dimensional complex Fourier transform data.

FIG. 3 is a diagram showing an example of a switching point table according to the present invention.

FIG. 4 is a graph showing a Fourier transform switching value according to the present invention.

[Explanation of symbols]

1 Optimal calculation method automatic selection generation unit 2 Multiple 1-dimensional complex Fourier transform calculation unit by 1-dimensional scanning method 3 Multiplexed 1-dimensional complex Fourier transform calculation unit incorporating 1-dimensional complex Fourier transform multiplexing inside 1-dimensional complex Fourier transform

Claims (1)

Claim: What is claimed is: 1. A switching point table storing the multiplicity of a one-dimensional complex Fourier transform is provided corresponding to the data length of input data, and the data length of data input during a multiple one-dimensional complex Fourier transform operation. By referring to the switching point table, the multiplicity of the input data is compared with the multiplicity obtained by the reference, and if the former is smaller than or equal to the latter, the multiple one-dimensional scanning method is performed. A complex Fourier transform operation unit, and when the former is larger than the latter, a Fourier transform characterized by selecting a multiplex one-dimensional complex Fourier transform operation unit in which multiplexing of one-dimensional complex Fourier transform is incorporated into the one-dimensional Fourier transform Calculation method selection method.