I recently discovered that you can declare doubles as 2 integers and convert a float into a double as 2 integers in Blitz. This was figured out ages ago for OpenGL but isn't in the code archives. I then found some WinBatch code and converted it to Blitz, so I thought I'd tidy it up and release it as a module.
Update: This module now includes 2 new functions, FloatToInt and IntToFloat. These allow 'type casting' without needing to use a bank.
Create a file in your userlibs folder called "floattodouble.decls" and copy/paste this code in.
.lib " "
;FloatToDouble Module Decls
FloatToDouble%(value#, dpart)
DoubleToFloat#(dho, dhi)
FloatToInt%(value#)
IntToFloat#(value)
Save this code entry as "floattodouble.bb" and include it in your main program. See the test code below for example usage. |
;FloatToDouble Module for B3D/B+
;Author: markcw, edited 29 Nov 08
;DLL functions with doubles for parameters can be declared
;as 2 integers instead. Then you can use FloatToDouble to convert a
;float into 2 integers which are the binary equivalent of a double
;but with the same value as the float. Finally you can pass
;the 2 integers to the 2 parameters in the DLL function.
;For example, glDepthRange takes 2 doubles so the decls would be:
;glDepthRange(zNearLo,zNearHi, zFarLo,zFarHi):"glDepthRange"
;Then a 'wrapper' function can be written to take 2 floats:
;bbglDepthRange(zNear#,zFar#)
Function FloatToDouble(value#, dpart = 0)
;Converts a float into a double as 2 integers
;dpart -> Double flag indicating which part to return, 0=dlo, 1=dhi
;Returns a low or high double integer - decimal equivalent of the float
;Site: techsupt.winbatch.com/TS/T000001034F21.html
Local integer, sign, exponent, fraction, dexp, dlo, dhi
integer = FloatToInt(value#)
sign = integer And $80000000 ;Sign bit
exponent = integer And $7F800000 ;8-bit exponent
fraction = integer And $007FFFFF ;23-bit mantissa
dexp = ((exponent Shr 23) - 127 + 1023) Shl 20 ;Double exponent
dlo = (fraction And 7) Shl 29
dhi = sign Or dexp Or (fraction Shr 3)
If dpart = 0 Then Return dlo
Return dhi
End Function
Function DoubleToFloat#(dlo, dhi)
;Converts a double as 2 integers into a float
;dlo -> Low double integer, dhi -> High double integer
;Returns a float - decimal equivalent of the double as 2 integers
;Site: techsupt.winbatch.com/TS/T000001034F21.html
Local dsgn, sign, dexp, exponent, fraction
dsgn = Abs(dhi Shr 31) ;Double sign
sign = dsgn Shl 31 ;Sign bit
dexp = Abs((dhi Shr 20) - (dsgn Shl 11)) ;Double exponent
exponent = (dexp + 127 - 1023) Shl 23 ;8-bit exponent
fraction = ((dhi And $000FFFFF) Shl 3) + (dlo Shr 29) ;23-bit mantissa
Return IntToFloat(sign Or exponent Or fraction)
End Function
Function FloatToInt(value#)
;Converts a float into a float as an integer
;Returns an integer that is the binary equivalent of the float
;Site: wiki.tcl.tk/756
Local sign, exponent, fraction#
Local f1f#, f2f#, f3f#, se1, e2f1, f1, f2, f3
If value# > 0 Then sign = 0 Else sign = 1
value# = Abs(value#)
exponent = Int(Floor(Log(value#) / 0.69314718055994529)) + 127
fraction# = (value# / (2 ^ (exponent - 127))) - 1
If exponent < 0 Then exponent = 0 : fraction# = 0.0 ;Round off to zero
If exponent > 255 Then exponent = 255 ;Outside legal range for a float
fraction# = fraction# * 128.0
f1f# = Floor(fraction#)
fraction# = (fraction# - f1f#) * 256.0
f2f# = Floor(fraction#)
fraction# = (fraction# - f2f#) * 256.0
f3f# = Floor(fraction#)
f1 = Int(f1f#) : f2 = Int(f2f#) : f3 = Int(f3f#)
se1 = (sign Shl 7) Or (exponent Shr 1) ;Sign and Exponent1
e2f1 = ((exponent And 1) Shl 7) Or f1 ;Exponent2 and Fraction1
Return (se1 Shl 24) Or (e2f1 Shl 16) Or (f2 Shl 8) Or f3
End Function
Function IntToFloat#(value)
;Converts a float as an integer into a float
;Returns a float that is the binary equivalent of the integer
;Site: www.cs.princeton.edu/introcs/91float/
Local sign, exponent, fraction
sign = (value And $80000000) Shr 31
exponent = (value And $7F800000) Shr 23
fraction = value And $007FFFFF
Return (-1 ^ sign) * (2 ^ (exponent - 127)) * (1 + (fraction / (2 ^ 23)))
End Function |
