Intel 80287, 80286 manual DoeF

NUMERIC PROGRAMMING EXAMPLES

iAPX2B6 MACRO ASSEMBLER

Loe OBJ

0083 09F4

0085 80FAFB

0088 7240

OOBA eOEAFB

0080 09E8

DoeF

DoeF DCC1 0091 DEE9

0093 D9F4

0095 D9C9

0097 OF15

0099 DEC;!

Q09Il F71D

009D 7529

009F D9C9

aOA1 OFID

aOA3 80EAF8

OOA6 E9A400

OOA9

OOA9 BOEAFC

()OAC E99EOO

aOAF

DOAF 09E8

OOBl DC'1C9

00133 D9FB

00135 D9F4

aaB7 09£5

0099 9BDFEO

COBe 09C9

DaDE D9CA

aOCD BOEAFA

aOC::3 1090044

aOC6 74C7

Dace ODDS

aOCA

aOCA

OOCA CB7EFO

OOCD OF56FC

0000 D9Ee

0002 DEe9

0004 DF5EFA

80287	Floating-Point to	IS-Digit ASCII Conversion		10: 12: 89 09/25/83 PAGE
LINE	SOURCE
228	fxtratt		J Separate exponent f'T'om significand
22.	cmp	dl,UNNDRMAL	,	Test for unnormal value
230	Jb	normal_value
231	sub	d 1, UNNORMAL-NORMAL		Return normal status with correct Sign
232

233

234Normalize the fraction. adJust the power of two in ST(1) and 1t.et

235the denormal count value.

236	Assert: 0 <'"ST(O) ( 1.0
237
238	-Fld 1	, Load constant to normalize fraction
239
240

241norma 1 i z e_frac t ion:

243	fadd	st(1).1t.t			Set integer bit in fraction
24.	hub			j	Form normalized fraction in ST(O)
245				J	Power of two Held will be negative
246	f xc h				of denormal count
247				j	Put denormal count in ST(O)
248	fi1t.t	word ptr	CdiJ	J	Put negative of dl!'normal count in memorl,l
24'	faddp	st (2). st			Form correct power of two in stt!)
250					OK to use word ptr CdiJ now
251	ne-g	word ph	CdiJ		Form positive denormt!ll count
:?52	Jnz	not_psuedo_zero
253

254A psuedo zero will appear an unnormal numbeor. When attemptinll

255to normalia it. the resultant fraction field will be zero. Perfo~minll

256	an f'xtrBct on zero will yield a		zero exponent value.	.
257			) Put power of tllJO value in !OttO)
258	f~ch
259	Fistp	word ptr CdiJ	Set denorm<ll count to power of' two value
260			Word ph (diJ is not u!;ed by convert
261			integer. OK to letJve	running
262	!:tub	dl, NDRMAL-P5UEDO_ZERD	,Set return value saving thf! sign bit
263	Jmp	convert_i.nteger	) Put zero value into memory
264

265The number is a real zero. 'iet the return value and setup for

266conversion to BCD

267

268
269	d 1, ZERO-NORMAL	,	Convert <;t.3tlIS to	normal v<'Jlu!'
270
~71	convert_integer	i	Treat the: l~ro as	an i.ntE!geT'

27.

273Th. nUmbllr is a dllnoTmal. FXTRACT will not wOTk correctly in tht.

27.elutt. To corrttetlu slIparate the eKponllnt and fraction, add a fixed

275	J	constant to tht' exponent to guarantee thll rellult il not a denormal.
276	,

277found_d enorma 1:

27.	fld 1	j PTeparl! to. bump axponent
280	heh	ForeD denormal to smallelit represllntable
281	fpram
28.		utendlld raal -Format exponent
283	htract	) This will work corrllctly now
28.	The power of the original dllnoT'mal valu. n.n baen safely isohtad.
28.

286Check if the fraction value is an unnormal.

287				SII' if the fraction is an unnormal
288	ham
28.	htsw			Save 80287 statu. In CPU AX rIg for later
2'0	hch		I	Put	uponent	in 9T(O)
291	heh	atC:!)	,	Put	1. ~ into	SHO),	uponant in 9T(2)
2'2	"b	dl. DENORMAL-NORMAI..	j	Return normal status with correct lign
:293	t!!!!t	8K,4400H	j	See	if! C:3-C:!"O imp ling unnormal aT' NAN
2 ••	J'	norma 1 i :re_fTac t i on	j	Jump if -Fraction is an unnormal
29'
	fstp	stCO)		Remov. unnecessary			1. 0 from lIteO)
2'6			j

2'7

2..8, Calculatt the d.eimal magnitude a"oelated with this numb Ill' to

lIIithin one order. Tnt. error lIIi1l allllays bll inevihbh due to

300rounding and lost pr.ttlion. A. a rlSult, we will delibarfttwly fail

301to considar the LOOIO of the fraction valua in calculating thll oT'dn.

302Since th hattion will alwaYI be 1 <- F <:!, its L.DOIO will not change

303thll baSic accurilcl/ oft the function. To get thll decimal order of magnitude.

30.limply multipl1J tha power of two bl,l L.OQ10(2) and truncate thll ruult to

30' , .an integar. 306

307normal_valu.:

308notJlIu.do_:rero:

30,

310	fltp	fTacUon		tlsve	,;;nll Tr.1;;~I.I;U' ii.."";	; ... , ~
311	fiat	POWtT'_two		Save	power of two
312	Udlg2			Qet	L.OQ10(2)
313				Pow.r_two is noUl saf. to us.
31.	fmul		I	Form LOG10(of exponent		of number)
31.	fhtp	power_tan	j	AnI,!	rounding moda will work harll
316

317Check if the magnitude of thll number rule. out treating it al

318integllr.

31.

320CX ha. the maximum number of decimal digits allowed.

Figure 4-6. Floating-Point to ASCII Conversion Routine (Cont'd.)

4-10