JPH0279104A - Numerical arithmetic unit using dsp - Google Patents

Abstract

PURPOSE:To eliminate the overhead for data delivery and reception so as to reduce the overall time for arithmetic operations by providing a means which can simultaneously execute the data setting of a host CPU and numerical operation performed by using a DSP. CONSTITUTION:A DSP 2 accesses a data delivery and reception area 3 and fetches external data after informing a host CPU 1 of the access and executes an arithmetic operation after informing the CPU 1 that the DSP 2 does not access the area 3. Upon receiving the information, the CPU 1 sets data necessary for the next arithmetic operation in the area 3. The DSP 2 places the results of the arithmetic operations in the area 3 after informing the CPU 1 of the access to the area 3 when the arithmetic operations are completed and executes the next arithmetic operation on the already set data after informing the CPU 1 that the DSP 2 does not access the area 3. The CPU 1 fetches the results of the previous arithmetic operation and sets data for the next arithmetic operation. Therefore, the time for the arithmetic operation can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a numerical calculation device using a DSP for a control device such as an industrial robot.

[Conventional technology]

Numerical calculation devices using conventional DSP are described in the 30th Automatic Control Association Conference Preprint (1987), pages 307 to 310.
Discussed on page.

[Problem to be solved by the invention]

In the above conventional technology, on page 308, lines 11 to 13, it is stated that the overhead involved in inputting and outputting external data is not a major problem, but the area for transferring external data is an 8-bit wide memory element. Considering the process of calculating the 484 matrix product between external data using , and returning the result to external memory, 57.
It takes 6μs. As stated in lines 9 to 1o on page 309, if the calculation time of the 4×4 matrix is 40 μs or less, there is a problem that the overhead may be sufficiently large.

SUMMARY OF THE INVENTION An object of the present invention is to provide a numerical calculation device that can reduce the data transfer overhead and shorten the overall calculation time even in the case of the calculation processing described above.

[Means to solve the problem]

In order to achieve the above object, a means is provided for indicating whether or not a numerical calculation device using a DSP is accessing a data transfer area.

[Effect]

First, a numerical arithmetic device using a DSP notifies the host CPU that the data transfer area will be accessed by means of indicating whether or not the data transfer area is being accessed, and takes in external data.

Next, the host CPU is notified that the data transfer area will not be accessed and an operation is executed. At this time host C
The PU sets the data necessary for the next calculation in the data transfer area. Numerical calculation equipment.

When the calculation is completed, notify the host CPU that the data transfer area will be accessed, place the result in the data transfer area, notify the host CPU that the data transfer area will not be accessed, and perform the next calculation using the data that has already been set. Execute. At this time.

The host CPU takes out the previous calculation result and sets the data necessary for the next calculation in the data transfer area.

As mentioned above, the data settings of the host CPU and the
Since the calculations of the numerical calculation device using P are executed simultaneously, there is no overhead for data exchange, and the overall calculation time can be shortened.

〔Example〕

An embodiment of the present invention will be described below with reference to FIGS. 1 and 2. This embodiment uses a DSP manufactured by NEC Corporation.

This is an example configured using μPD77230.

As shown in Fig. 1, this control device has a host CPU,
Numerical calculation device 2 using DSP, data transfer area 3 for exchanging data between 1 and 2, and a data transfer area 3 for requesting calculation execution from the host CPU to numerical calculation device 2, and for data transfer by the host CPU. Control 4IC1゜12 for notifying the DSP2 whether or not area 3 is being accessed; control aC for notifying the host CPUI from the DSP2 whether or not an operation is being executed;
2,.

, L3. It is composed of a control line C3°[4 that indicates whether or not the DSP2 is accessing the data transfer area:3. DSP2 uses the first memory as an internal memory resource.
RAM 4, a second RAM 5, and a data ROM 8, an arithmetic unit 9 that executes arithmetic operations, and a host CPU
Control unit 7 to connect the control line with
, and a calculation unit 9 and a control unit 7
It consists of an instruction ROM 6 that stores a control program for operating the . In addition, the host C: PU1 and the data transfer area 3 are connected by a 16-bit data bus 10, and the DSP 2 and the data transfer area 3 are connected by an 8b
It is connected by a data bus 11 of it. The operation will be explained below with reference to FIG. FIG. 2 shows a timing diagram of the control lines C1, C2, and C3 for one cycle of calculation, and the operations of the host CPUI and DSP2.

First, in the host CPtJ1, the DSP2 is not in operation (
After confirming that C2 is Low), it issues a calculation execution request to DSP2 (CI High). At this time, DSP2 notifies the host CPU that it has accepted the calculation request (C2 goes High).
gh) with.

Notify that data transfer area 3 will be used (C3 to H
igh)L, data is stored in the first RAM4 or the second R.
AM58 transfer. In response to this, the host CPU
After confirming that SP2 has accepted the calculation (C1 is Low), it waits until the data transfer area 3 becomes available (C3 is Low). When the DSP2 has finished transferring the data internally, it transfers the data to the data transfer area. is released (C3 is set low) and the execution of the operation is started. In response to this, host CP
The UI indicates that the DSP2 uses the data passing area.
(C1 is set High), and the data necessary for retrieving the previous calculation result and the primary calculation is set. Then, the data transfer area 3 is released (C1 is set low), and the DSP
Wait until operation 2 is completed (C2 is low). on the other hand,
When the execution of the operation is completed, the DSP2
is not using data transfer area 3 (C1 is LOW), and then notifies the host CPU that data transfer area 3 will be used (C3 is HIGH).
and transfers the calculation results to the data transfer area. When the transfer is completed, the host CPU 1 is notified of the end of the calculation (C
2 is set to LOW), the data transfer area is released (C3 is set to LOW), and the next calculation request is waited for.

It goes without saying that when making the first calculation request, necessary data must be set in the data transfer area in advance.

As described above, according to this embodiment, when the DSP 2 is made to perform an operation, the data for the second and subsequent operations are as follows:
Since the settings can be made in parallel while the previous calculation is being executed, the overhead due to data transfer can be largely eliminated and the final calculation time can be shortened.

〔Effect of the invention〕

According to the present invention, data necessary for the next calculation can be set from the host CPU in the data transfer area during calculation by a numerical calculation device using a DSP, so the overhead of data transfer can be significantly reduced. can. This allows the overall calculation time to be significantly reduced.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of an embodiment of the present invention, and FIG. 2 is a timing chart of control lines in the configuration of FIG. 1. DESCRIPTION OF SYMBOLS 1... Host CPU, 2... DSP, 3... Data transfer area, 4... First RAM, 5... Second RAM, 6... Instruction ROM, 7...
...Control unit, 8...Data ROM, 9
... Arithmetic unit h, 10, 1.1... Data bus,
12, 13, 14...Control 1-roll line.

Claims (1)

[Claims] 1. In a numerical calculation device using a host CPU and a DSP of an industrial robot control device that exchanges data with the host CPU, a means is provided for indicating whether or not the means for exchanging data is being accessed. A numerical calculation device using a DSP characterized by: