Message-Id: <199908051936.PAA05056@delorie.com>
From: "Batchex" <thedark1 AT kubrick DOT hotpop DOT com>
Organization: Digital Arts Creatives
To: djgpp AT delorie DOT com
Date: Fri, 6 Aug 1999 02:39:07 +0000
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: Forming a float from 4 seperate bytes
In-reply-to: <7nku8s$hkl$1@nnrp1.deja.com>
Reply-To: djgpp AT delorie DOT com
X-Mailing-List: djgpp AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> From:          jsc AT lds DOT co DOT uk
> Subject:       Forming a float from 4 seperate bytes
> Date:          Tue, 27 Jul 1999 18:37:16 GMT
> Organization:  Deja.com - Share what you know. Learn what you don't.
> To:            djgpp AT delorie DOT com
> Reply-to:      djgpp AT delorie DOT com

> Hello all, sorry for the post that isn't specifically related to DJGPP, but
> it's the only C news group I'm on.
> 
> I'm writing an (old) 3DStudio4 file parser and found a problem in the way
> that the vertex coordinates are listed. each floating point value is
> represented by 4 bytes which I can access but can't combine to form the
> original float.
> 
> Well, I sort of can. If I do it directly from the file and use
> 
> fread (&Xvalue,4,1,fp);
> 
> where fp is my file pointer, 4 = the sizeof(float) , 1 = I want to do this
> once and Xvalue = the final float value I want to end up with, it works. I
>
I guess it's because that fread() reads the 4 bytes as one package, 
and the ordering of each bytes are not changed.

> don't know why - I just took it from a doc on the 3ds format. However I want
> to unload the file to a buffer and then start parsing. I've done the reading
> into the buffer, but now I don't know how to reform my float from the data, I
> tried  taking the 4 values and &-ing them together, but that doesn't work as
> the final value I'm after is a float and I get an operand error. Any ideas on
> how to take 4 integers and form the float? I'm surre it's easy but I can't
> get it to work. Thanks Chris  jsc AT lds DOT co DOT uk
> 
Taking the 4 values then ANDing them together won't work because of 
the storing format of int and float differs greatly. The IEEE 
standard for storing a 32 bit float is :
	bit 0 - 22	significand
	bit 23 - 30	exponent
	bit 31	sign

As you can see, ANDing or any other bitwise operation with the 4 
integers won't produce a float as integers use all of it bits to 
represent the integer value, unlike float.

I have a similar problem once. The problem is how to read & combine a 
32 bit integer that is consists of 4 bytes.  My approach to the 
problem is to use a bit assembly. 

Let's assume that BytePtr is the pointer to the series of 4 bytes 
that makes the integer in the memory buffer.

;------------------------------------------------------------
; prototype : int combineInt(char *BytePtr)
;------------------------------------------------------------
combineInt:
	push ebp
	mov ebp,esp

	sub eax,eax
	mov esi,[ebp+BYTEPTR]
	mov ah,[esi+3]
	mov al,[esi+2]
	shl eax,16
	mov ah,[esi+1]
	mov al,[esi]

	mov esp,ebp
	pop ebp
	ret

All the above function do is taking each byte from the memory buffer 
and store it at the proper place in a 32 bit register.

To modify it to combine a float instead, then just modify the 
prototype so that it returns a float instead of an int.

prototype : float CombineInt(char *BytePtr);

> 
> Sent via Deja.com http://www.deja.com/
> Share what you know. Learn what you don't.
> 

Hope that helps.

Batchex
thedark1 AT Phreaker DOT net