Electronic control system expediting floating point processing

Details Electronic control system expediting floating point processing Electronic control system expediting floating point processing

2003-01-08

2004-11-09

Wolfe, Willis R. (Department: 3747)

Data processing: vehicles, navigation, and relative location

Vehicle control, guidance, operation, or indication

With indicator or control of power plant

C073S116070, C073S117020, C701S102000, C701S110000, C701S111000, C701S115000

Reexamination Certificate

active

06816777

ABSTRACT:

CROSS REFERENCE TO RELATED APPLICATIONS
This application is based on and incorporates herein by reference Japanese Patent Applications No. 2002-6057 filed on Jan. 15, 2002 and No. 2002-329588 filed on Nov. 13, 2002.
FIELD OF THE INVENTION
The present invention relates to an electronic control system equipped with floating point processing function, in which each computation result is replaced with a default value for expediting the floating point processing when it is determined to be ‘non-numeral’ or ‘infinity.’
BACKGROUND OF THE INVENTION
Recent electronic control units (ECUs) for vehicles or the like are equipped with floating point processing units (FPUs). Use of the FPU enables computation of a floating point mode datum to obtain a higher precision computation result in comparison with use of a fixed point mode datum.
The floating point computation is well known and its datum format is specified in various standards. For instance, a datum format of a single-precision floating point mode compliant with IEEE 754 is composed of a 1-bit sign part, 8-bit exponent part, and 23-bit mantissa part. The total four byte datum possesses a seven digit resolution. The floating point mode datum can represent, as a ‘non-numeral,’ a computation result that cannot be represented as a numeral such as a ‘0/0’ and ‘+∞-∞.’ For instance, in the above standard, if a datum has a mantissa part of other than ‘0’ and an exponent part of all eight bits to be ‘1’, the datum is treated as a ‘non-numeral.’ This kind of ‘non-numeral’ datum format may appear due to noise. Namely, when the noise forces all the exponent part to be changed to ‘1’ with the mantissa part being other than ‘0’, the datum is treated as a ‘non-numeral.’
Once the ‘non-numeral’ appears, a computation result using the ‘non-numeral’ naturally becomes ‘non-numeral.’ The ‘non-numeral’ must be therefore rewritten to a normal value. To solve this issue, the following two well-known methods are used.
Method (1): An entire area storing floating point mode data are searched for a ‘non-numeral’ using timer interruption and a default value is substituted for the ‘non-numeral’ when the ‘non-numeral’ is found.
Method (2): Similarly to the above method (1), an entire area storing floating point mode data are searched for a ‘non-numeral’ and a CPU is reset to initialize all the floating point mode data using an initializing routine when the ‘non-numeral’ is found.
The above method (1) adversely affects a memory area. Specifically, reasons will be explained by referring to the following computation types.
Computation (1): A+B→C, where the sum of A and B is substituted for C.
When a computation result C becomes ‘non-numeral,’ other computation results using C become ‘non-numeral.’ However, when neither A nor B becomes ‘non-numeral’ thereafter, C returns to a normal numeral in the following computation. Hence, C and a computation result using C naturally return to the normal value for a short interval.
Computation (2): A+B→A, where the sum of A and B is substituted for A.
When a computation result A becomes ‘non-numeral,’ other computation results using A become ‘non-numeral.’ Here, in a subsequent computation using A, a result thereby becomes ‘non-numeral’ and never returns to a normal value. Hence, A and a computation result using A never returns to the normal value.
As explained in the computation (2), where a preceding result is used for a current computation, appearance of a ‘non-numeral’ poses issues on the computation.
The method (1) that substitutes a default value involves a sizable memory area for storing default values of floating point mode data to be substituted for ‘non-numeral’ computation results. The default values must be prepared and stored even for computation results that will naturally return to the normal values in the computation (1).
The method (2) that resets a CPU adversely affects reliability of controlling a system due to suspending operation of a running system even if it is suspended for any short period.
The above issues result from not only a ‘non-numeral’ but also an ‘infinity.’ For instance, in the IEEE 754 specification, if a four byte datum has a mantissa part to be ‘0’ and an exponent part of all eight bits to be ‘1’, the datum is treated as an ‘infinity.’
SUMMARY OF THE INVENTION
It is an object of the present invention to provide an electronic control system capable of shortly returning a non-numeral or an infinity value of a floating point mode datum to a normal value.
To achieve the above and other objects, an electronic control system is programmed as follows: Repetitive computation is executed in the control program. In the computation, a floating point mode datum is used and a preceding computation result of the floating point mode datum is used for computing a current computation result. Whether the computation result is non-numeral is determined. When the computation result is determined to be non-numeral, a default value is substituted for the computation result of the non-numeral.
In one embodiment, whether a computation result is infinity is determined. When the result is determined to be infinity, a default value is substituted for the computation result of the infinity value.
The above processing enables the electronic control system to return a non-numeral or an infinity value of a floating point mode datum to a normal value with no impact on a memory area and no suspension of operation of a running system. This thereby ensures high reliability on controlling the system.


REFERENCES:
patent: 5201296 (1993-04-01), Wunning et al.
patent: 6082187 (2000-07-01), Schricker et al.
patent: 6175911 (2001-01-01), Oberman et al.
patent: 6199007 (2001-03-01), Zavarehi et al.
patent: 6216061 (2001-04-01), Franke et al.
patent: 6393554 (2002-05-01), Oberman et al.
patent: 6603481 (2003-08-01), Kawai et al.
patent: 6714197 (2004-03-01), Thekkath et al.
patent: 2001/0005808 (2001-06-01), Kawai et al.
patent: B2-H6-19740 (1994-03-01), None
patent: A-2001-138758 (1999-11-01), None

Affiliated with

Also associated with

Rating

Electronic control system expediting floating point processing does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Electronic control system expediting floating point processing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Electronic control system expediting floating point processing will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3284388