AR999A ArchGlbPrm
Functional Design Document
For
Architecture Global Parameters
VERSION: 1.0
DATE: 05-Mar-2015
Prepared By:
Nexteer Automotive,
Saginaw, MI, USA
Revision History
| Version | Description | Author | Section Modified | Date |
| 1.0 | Initial Version | Kathleen Creager | All | 05-Mar-2015 |
Table of Contents
4.1.1 float32 Constant specification 7
Abbrevations And Acronyms
| Abbreviation | Description |
|---|
References
This section lists the title & version of all the documents that are referred for development of this document
| Sr. No. | Title | Version |
|---|---|---|
Purpose
The purpose of this document is to describe the design of the Architecture Global Parameters component including the selection criteria for items to be included in the component and the method used for specifying the values of the items.
ArchGlbPrm Design
constants
The ArchGlbPrm component contains global constants that are either “mathematical” (e.g. the value of pi) or software-oriented (e.g. the value to be used as the zero threshold for float32 comparisons) in nature. The constants are defined in AR999A_ArchGlbPrm_DataDict.m and implemented as #define statements in “ArchGlbPrm.h”.
float32 Constant specification
For each float32 constant, the number of digits was chosen based on the smallest number of digits needed to give the same float32 representation as the value calculated by Excel. Rationale: this gives the most accurate representation possible for float32, while not containing a misleading number of digits beyond the resolution actually possible with float32.
Example procedure for determining the number of digits to use:
Enter the desired constant in Excel as a formula, e.g. “=PI()/2”
Copy the cell and paste as value into another cell
Select the new cell and copy the displayed value to the clipboard
Paste this value in a IEEE floating point conversion tool; record the single precision floating point representation (in hex), both rounded and unrounded.
Val = the pasted Excel value
X = number of digits in Val
Y = X
Do:
Y = Y - 1
Val1 = Val rounded to Y digits
While ((rounded single float representation of Val == rounded single float representation of Val1) AND
(unrounded single float representation of Val == unrounded single float representation of Val1))
Val2 = Val rounded to Y + 1 digits
Use Val2 as the constant value in the .h and .m files