FreeBasic SDK: extended data type

**ErosOlmi** · 22-08-2008, 02:22

I'm working on implementing thinBasic SDK for FreeBasic compiler but I'm stuck with EXTENDED data type.
Most of the thinBasic SDK functions handling numbers are designed to work with EXTENDED numbers but EXTENDED are not supported natively by FreeBasic.

So, for example, if a SDK function expect an EXT passed BYREF, how to handle in FreeBasic?

Any help is appreciated.

Thanks a lot
Eros

**jack** · 22-08-2008, 03:05

someone wrote a real10 library and it's included in the examples folder, but I must warn you that the output routine is not 100% solid, I will see if I can work on it on my days off (a week from now)
you have to compile the library yourself from the command line.

**ErosOlmi** · 22-08-2008, 03:18

Thanks a lot.
Do you know how to invoke FBC to compile the library? I can see a make file but no idea how to use it.

**jack** · 22-08-2008, 03:25

actually it's very simple, just launch the freeBasic command prompt, cd to real10\libreal10 and type make

**ErosOlmi** · 22-08-2008, 03:34

It doesn't work here. I have a lot of compilers installed and when I invoke MAKE it starts the Borland make utility.
I need to check my path precedences and FreeBasic installation. I usually compile FreeBasic sources from inside FBEdit making a FBEdit project.

Do not worry, I'll find a way.
Thanks
Eros

**jack** · 22-08-2008, 03:45

you may need to have Mingw32 installed, anyway attached is the one I just compiled with the latest FB
I believe my system to be clean but recommend you check for virus.

**ErosOlmi** · 22-08-2008, 03:55

Thanks a lot. Compiled test example and is working fine.

I will now work on thinBasic SDK interface. I do not need to make complex operations. I just need to declare some EXT numbers, pass them to thinBasic SDK functions, return values and convert to Double. All FreeBasic calculations will be performed with doubles.

**jack** · 22-08-2008, 04:03

one thing you need to change is the ext structure, as it is defined in the real10 library it's 16 bytes, posibly for speed reasons.

**Charles Pegge** · 22-08-2008, 07:28

Converting between 64 bit and 80 bit floating point is very simple with a little machine code

We know that varptr always returns the variable address in the eax register so, all we have to do is formulate fld and fstp pairs which use the eax register as a simple pointer.

I would suggest managing the conversion entirely on the FreeBasic side:

FreeBasic[code=thinbasic]

type extended
dbl as double
xtn as long ' assuming alignment to 4 bytes
end type

dim myvar as extended
dim p as any ptr

'convert from double to extended
p=varptr(myVar)
asm .byte &hdd, &h00, &hdb,&h38

'convert from extended to double
p=varptr(myVar)
asm .byte &hdb, &h28, &hdd,&h18

[/code]

**ErosOlmi** · 22-08-2008, 12:22

Thanks to all.

Charles, yes, I will do in FreeBasic all the stuff as you suggested.
My target is not to duplicate thinBasic SDK just to be compatible with EXT and DOUBLE. I want all functions as they are and study some simple transparent way for FreeBasic programmers.

Ciao
Eros