displaying floating point problem - Forums

Home » Forums » MAIN » displaying floating point problem

I do not know if such a question appeared. So sorry if "YES".
I'm just starting with asm so ... ... . . . . .. . . . . . . . . . . .. ...................................thx.

.date
X dd 1.5
format db "%d",0
text db 10 dup (?)

invoke wsprintf,ADDR text,ADDR format,X
invoke TextOut,hdc,50,50,addr text,11

;and a strange things appears in the window

it is very short version as u can see

Posted on 2002-11-15 07:25:13 by drakoforma

displaying floating point problem

You show floating points with '%f' not '%d'. '%d' shows the parameter as a DWORD.

Posted on 2002-11-15 08:42:00 by gliptic

displaying floating point problem

wsprintf doesn't handle floating points

h.
:rolleyes:

Posted on 2002-11-15 15:25:16 by hitchhikr

displaying floating point problem

OK thx but it not solve my problem.

"" BDW %d does not mean a DWORD but "decimal" ""

Posted on 2002-11-16 10:47:30 by drakoforma

displaying floating point problem

The masm32 package come with several floating point to string conversion routines.
have a look at them....

Posted on 2002-11-16 15:52:24 by MArtial_Code

displaying floating point problem

Those functions take QWORDS.
Unfortunately, most people are using FLOATS which are REAL4 (DWORDS).

To convert a DWORD to a QWORD you can use the CDQ opcode.
It extends 32bit EAX into 64bit EDX:EAX, keeping the sign true.
What a pain !!
We want better support in masmlib for larger data types !!

Posted on 2002-12-04 00:21:33 by Homer

displaying floating point problem

We want better support in masmlib for larger data types !!

Write your own routine and submit it, im sure hutch will be happy to consider it for everyones use. Cheers, thanks a lot for volunteering:rolleyes:

Posted on 2002-12-04 03:17:10 by huh

displaying floating point problem

I have cut out this snip of code, it takes a 80 bit float (lpdvFloat), a maximum precision in digits (dwPrec) and returns a string in lpszOut.
The float parameter can be easily changed from tbyte to qword or dword by changing the fld command.

store	macro	data

	local	stoVoid



	test	edi, edi

	je	stoVoid



	mov	byte ptr [edi + edx], data

stoVoid:	inc	edx



	endm



fjl	MACRO Marke

	fstsw	ax

	sahf

	ja	Marke

	ENDM



fjge	MACRO Marke

	fstsw	ax

	sahf

	jbe	Marke

	ENDM



;***********************************************************************

; does:	converts float to string			*

; last update:	07-12-2001 - Scholz			*

; parameters:	dwLength FloatToString (lpdvFloat, dwPrec, lpszOut)		*

; returns:	length of string or zero			*

;***********************************************************************



FloatToString	PROC	USES edi esi, lpdvFloat:DWORD, dwPrec:DWORD, lpszOut:DWORD

	LOCAL	dvTenth:QWORD, dvCalc:QWORD, dvTen:QWORD, dvFloat:QWORD

	LOCAL	wFPU:DWORD, wOrg:DWORD, dwNumber:DWORD, fSign:DWORD



	finit

	fstcw	word ptr wFPU



	mov	eax, wFPU

	mov	wOrg, eax



	or	word ptr wFPU, 0F00h

	fldcw	word ptr wFPU



	mov	dword ptr dvTen, 10

	fild	dword ptr dvTen

	fst	dvTenth

	fstp	dvTen



	xor	eax, eax

	mov	fSign, eax



	mov	eax, lpdvFloat

	fld	tbyte ptr [eax]

	fldz

	fcomp	st(1)

	fjge	fstPos

	fchs



	mov	fSign, TRUE



fstPos:	fstp	dvFloat





;----------------- calc maximum decimal base



fstLength:	fld	dvFloat

	fdiv	dvTenth

	frndint



	fld1

	fcomp	st(1)

	fjl	fstConInt



	ffree	st

	fld	dvTenth

	fmul	dvTen

	fstp	dvTenth

	jmp	fstLength





;----------------- convert integer part



fstConInt:	xor	edx, edx

	mov	edi, lpszOut



	mov	eax, fSign

	test	eax, eax

	je	fstStart



	store	"-"



fstStart:	fld	dvTenth

	fdiv	dvTen

	fstp	dvTenth



	fld	dvFloat

	fstp	dvCalc



fstIntLoop:	fld1

	fcomp	dvTenth

	fjl	fstConFrac



	fld	dvCalc

	fld	dvCalc

	fdiv	dvTenth

	frndint

	fist	dword ptr dwNumber

	fmul	dvTenth

	fsubp	st(1), st



	mov	ecx, dwNumber

	add	cl, "0"

	store	cl



	fld	dvTenth

	fdiv	dvTen

	fstp	dvTenth



	fstp	dvCalc

	jmp	fstIntLoop





;----------------- convert fractional part



fstConFrac:	store	"."



	xor	ebx, ebx

	mov	esi, dwPrec

fstFracLoop:	cmp	ebx, esi

	jae	fstZeros



	fld	dvCalc

	fld	dvCalc

	fdiv	dvTenth

	frndint

	fist	dword ptr dwNumber

	fmul	dvTenth

	fsubp	st(1), st



	mov	ecx, dwNumber

	add	cl, "0"

	store	cl

	inc	ebx



	fld	dvTenth

	fdiv	dvTen

	fstp	dvTenth



	fstp	dvCalc

	jmp	fstFracLoop





;----------------- remove least zeros



fstZeros:	add	edx, dwPrec

	test	edi, edi

	je	fstDone



	sub	edx, dwPrec

	dec	edx



	cmp	byte ptr [edi + edx], "."

	je	fstDot

	cmp	byte ptr [edi + edx], "0"

	jne	fstZero

	jmp	fstZeros



fstDot:	dec	edi

fstZero:	mov	byte ptr [edi + edx + 1], 0



fstDone:	mov	eax, edx



	finit

	fldcw	word ptr wOrg

	ret



FloatToString	ENDP

Posted on 2002-12-04 04:26:17 by beaster

displaying floating point problem

Unfortunately, most people are using FLOATS which are REAL4 (DWORDS).

To convert a DWORD to a QWORD you can use the CDQ opcode.
It extends 32bit EAX into 64bit EDX:EAX, keeping the sign true.

You do not convert a REAL4 to REAL8 by using CDQ. You are better off just loading an FP register, which automatically extends a REAL4 number to 80-bit format. And then storing a REAL8 number from that register. I don't use the FP registers, so I'm guessing the opcodes are FLD and FST.

Hint: FP does not use two's complement arithmetic.

Posted on 2002-12-04 12:17:26 by tenkey

ASM Community