How to Change a SINGLE Floating Point Number to a Winbatch Floating Point Number

Keywords: 	 floating point

This article is outdated. Use the new DataCast function found in WinBatch versions 2006B or newer.

WinBatch, internally, keeps floating post numbers in a 64-bit format that matches the IEEE 64-bit double precision standard. The math chips and compilers et al on PC tend to support this standard. Those numbers have to be stored somehow.

There is also a 32-bit IEEE standard also supported by math hardware and the such. It is possible to get a binary file that has the floating point numbers stored in the 32-bit format.

WinBatch has no built-in support for the 32-bit format. Sometimes one would like to read a floating point number in the 32 bit format and see it as a floating point number. To do this in WinBatch requires converting the 32-bit floating point number to a 64-bit floating point number

This web page has a little discussion these two floating point formats....


 http://www.csrd.uiuc.edu/~ece412/cpu/r4400/R4000.book_133.html

Good Luck




;Single precision 32-bit formst              1-8-23
; 1 sign bit, 8 exponent bits and 32 fraction bits
;S EEEEEEEE FFFFFFFFFFFFFFFFFFFFFFF

;Double precision format        1-11-52  OR   1-11-20 x-x-32
; 1 sign bit 11 exponent bits and 52 fraction bits
;S EEEEEEEEEEE FFFFFFFFFFFFFFFFFFFF-FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF

; S is easy to preserve.
; E is the tricky part
; in S32 format subtract 127 to get the true exponent
; in D64 format subtract 1023 to get the true exponent
; F gets copied, left adjusted zero filled on the end


;So now we need a test case

;So ummm I think THREE in single format is
;0 10000000 100000000000000000000
; breaking into nibbles (4 bit chunks)
; 0100 0000 0100 0000 0000 0000 0000 0000
; 40400000 in hex
; 1077936128 in decimal

;So set test case up.  Assume we did a BinaryPeek4 from a binary
;buffer to retrieve a 32 bit number.  Now it it were an integer
;it would be 1077936128.  But its not.  So a whole bunch of bit
;bashing has to take place....
three=1077936128 

;Define some hex constants
X80000000=8 << 28
X7F800000=32640 << 16
X007FFFFF=8388607

;Figure out sign, exponent, and fraction of 32 bit number
Ssign =three & X80000000
Sexp  =three & X7F800000
Sfract=three & X007FFFFF


;Scoot the single precis exponent to make an integer
Dexp=Sexp >> 23
; Subtract 127 to get the true exponent, then
; add 1023 to get a double precision exponent
Dexp=Dexp-127+1023
; and scoot it back to where it is supposed to be
Dexp=Dexp << 20

; Now for the fraction.  The single precision number has
; 23 bits of fraction.  The double precision number has
; more, but they are not all in the first word, so we
; have to split the fraction into two parts.  The top word
; of the double precision fraction has 20 bits, leaving 3
; bits to move to the next word

; For the high word just scoot 3 bits off the end to get
; 23 bits down to 20
SFH=Sfract >> 3

;for the low word save the bottom 3 bits and scoot them up
; to the top of the word
SFL=(Sfract & 7) << 29

;Build high and low double precision words as two integers
DH=Ssign | Dexp | SFH
DL=SFL

; Now store those two integers into a binary buffer as 8 bytes
bb=BinaryAlloc(8)
BinaryPoke4(bb,0,DL)
BinaryPoke4(bb,4,DH)
;then read those 8 bytes as a WinBatch compatible floating point number.
ans=BinaryPeekFlt(bb,0)
BinaryFree(bb)

;Display answer.  Hope its 3.0
Message("And the answer is...",ans)

Article ID:   W14227

Filename:   Change a Single Floating Point to WB Floating.txt

File Created: 2006:03:07:13:11:32

Last Updated: 2006:03:07:13:11:32

Database Search

TechHome

How to Change a SINGLE Floating Point Number to a Winbatch Floating Point Number