xref: /linux/arch/m68k/fpsp040/scale.S (revision bd628c1bed7902ec1f24ba0fe70758949146abbe)
1|
2|	scale.sa 3.3 7/30/91
3|
4|	The entry point sSCALE computes the destination operand
5|	scaled by the source operand.  If the absolute value of
6|	the source operand is (>= 2^14) an overflow or underflow
7|	is returned.
8|
9|	The entry point sscale is called from do_func to emulate
10|	the fscale unimplemented instruction.
11|
12|	Input: Double-extended destination operand in FPTEMP,
13|		double-extended source operand in ETEMP.
14|
15|	Output: The function returns scale(X,Y) to fp0.
16|
17|	Modifies: fp0.
18|
19|	Algorithm:
20|
21|		Copyright (C) Motorola, Inc. 1990
22|			All Rights Reserved
23|
24|       For details on the license for this file, please see the
25|       file, README, in this same directory.
26
27|SCALE    idnt    2,1 | Motorola 040 Floating Point Software Package
28
29	|section	8
30
31#include "fpsp.h"
32
33	|xref	t_ovfl2
34	|xref	t_unfl
35	|xref	round
36	|xref	t_resdnrm
37
38SRC_BNDS: .short	0x3fff,0x400c
39
40|
41| This entry point is used by the unimplemented instruction exception
42| handler.
43|
44|
45|
46|	FSCALE
47|
48	.global	sscale
49sscale:
50	fmovel		#0,%fpcr		|clr user enabled exc
51	clrl		%d1
52	movew		FPTEMP(%a6),%d1	|get dest exponent
53	smi		L_SCR1(%a6)	|use L_SCR1 to hold sign
54	andil		#0x7fff,%d1	|strip sign
55	movew		ETEMP(%a6),%d0	|check src bounds
56	andiw		#0x7fff,%d0	|clr sign bit
57	cmp2w		SRC_BNDS,%d0
58	bccs		src_in
59	cmpiw		#0x400c,%d0	|test for too large
60	bge		src_out
61|
62| The source input is below 1, so we check for denormalized numbers
63| and set unfl.
64|
65src_small:
66	moveb		DTAG(%a6),%d0
67	andib		#0xe0,%d0
68	tstb		%d0
69	beqs		no_denorm
70	st		STORE_FLG(%a6)	|dest already contains result
71	orl		#unfl_mask,USER_FPSR(%a6) |set UNFL
72den_done:
73	leal		FPTEMP(%a6),%a0
74	bra		t_resdnrm
75no_denorm:
76	fmovel		USER_FPCR(%a6),%FPCR
77	fmovex		FPTEMP(%a6),%fp0	|simply return dest
78	rts
79
80
81|
82| Source is within 2^14 range.  To perform the int operation,
83| move it to d0.
84|
85src_in:
86	fmovex		ETEMP(%a6),%fp0	|move in src for int
87	fmovel		#rz_mode,%fpcr	|force rz for src conversion
88	fmovel		%fp0,%d0		|int src to d0
89	fmovel		#0,%FPSR		|clr status from above
90	tstw		ETEMP(%a6)	|check src sign
91	blt		src_neg
92|
93| Source is positive.  Add the src to the dest exponent.
94| The result can be denormalized, if src = 0, or overflow,
95| if the result of the add sets a bit in the upper word.
96|
97src_pos:
98	tstw		%d1		|check for denorm
99	beq		dst_dnrm
100	addl		%d0,%d1		|add src to dest exp
101	beqs		denorm		|if zero, result is denorm
102	cmpil		#0x7fff,%d1	|test for overflow
103	bges		ovfl
104	tstb		L_SCR1(%a6)
105	beqs		spos_pos
106	orw		#0x8000,%d1
107spos_pos:
108	movew		%d1,FPTEMP(%a6)	|result in FPTEMP
109	fmovel		USER_FPCR(%a6),%FPCR
110	fmovex		FPTEMP(%a6),%fp0	|write result to fp0
111	rts
112ovfl:
113	tstb		L_SCR1(%a6)
114	beqs		sovl_pos
115	orw		#0x8000,%d1
116sovl_pos:
117	movew		FPTEMP(%a6),ETEMP(%a6)	|result in ETEMP
118	movel		FPTEMP_HI(%a6),ETEMP_HI(%a6)
119	movel		FPTEMP_LO(%a6),ETEMP_LO(%a6)
120	bra		t_ovfl2
121
122denorm:
123	tstb		L_SCR1(%a6)
124	beqs		den_pos
125	orw		#0x8000,%d1
126den_pos:
127	tstl		FPTEMP_HI(%a6)	|check j bit
128	blts		nden_exit	|if set, not denorm
129	movew		%d1,ETEMP(%a6)	|input expected in ETEMP
130	movel		FPTEMP_HI(%a6),ETEMP_HI(%a6)
131	movel		FPTEMP_LO(%a6),ETEMP_LO(%a6)
132	orl		#unfl_bit,USER_FPSR(%a6)	|set unfl
133	leal		ETEMP(%a6),%a0
134	bra		t_resdnrm
135nden_exit:
136	movew		%d1,FPTEMP(%a6)	|result in FPTEMP
137	fmovel		USER_FPCR(%a6),%FPCR
138	fmovex		FPTEMP(%a6),%fp0	|write result to fp0
139	rts
140
141|
142| Source is negative.  Add the src to the dest exponent.
143| (The result exponent will be reduced).  The result can be
144| denormalized.
145|
146src_neg:
147	addl		%d0,%d1		|add src to dest
148	beqs		denorm		|if zero, result is denorm
149	blts		fix_dnrm	|if negative, result is
150|					;needing denormalization
151	tstb		L_SCR1(%a6)
152	beqs		sneg_pos
153	orw		#0x8000,%d1
154sneg_pos:
155	movew		%d1,FPTEMP(%a6)	|result in FPTEMP
156	fmovel		USER_FPCR(%a6),%FPCR
157	fmovex		FPTEMP(%a6),%fp0	|write result to fp0
158	rts
159
160
161|
162| The result exponent is below denorm value.  Test for catastrophic
163| underflow and force zero if true.  If not, try to shift the
164| mantissa right until a zero exponent exists.
165|
166fix_dnrm:
167	cmpiw		#0xffc0,%d1	|lower bound for normalization
168	blt		fix_unfl	|if lower, catastrophic unfl
169	movew		%d1,%d0		|use d0 for exp
170	movel		%d2,-(%a7)	|free d2 for norm
171	movel		FPTEMP_HI(%a6),%d1
172	movel		FPTEMP_LO(%a6),%d2
173	clrl		L_SCR2(%a6)
174fix_loop:
175	addw		#1,%d0		|drive d0 to 0
176	lsrl		#1,%d1		|while shifting the
177	roxrl		#1,%d2		|mantissa to the right
178	bccs		no_carry
179	st		L_SCR2(%a6)	|use L_SCR2 to capture inex
180no_carry:
181	tstw		%d0		|it is finished when
182	blts		fix_loop	|d0 is zero or the mantissa
183	tstb		L_SCR2(%a6)
184	beqs		tst_zero
185	orl		#unfl_inx_mask,USER_FPSR(%a6)
186|					;set unfl, aunfl, ainex
187|
188| Test for zero. If zero, simply use fmove to return +/- zero
189| to the fpu.
190|
191tst_zero:
192	clrw		FPTEMP_EX(%a6)
193	tstb		L_SCR1(%a6)	|test for sign
194	beqs		tst_con
195	orw		#0x8000,FPTEMP_EX(%a6) |set sign bit
196tst_con:
197	movel		%d1,FPTEMP_HI(%a6)
198	movel		%d2,FPTEMP_LO(%a6)
199	movel		(%a7)+,%d2
200	tstl		%d1
201	bnes		not_zero
202	tstl		FPTEMP_LO(%a6)
203	bnes		not_zero
204|
205| Result is zero.  Check for rounding mode to set lsb.  If the
206| mode is rp, and the zero is positive, return smallest denorm.
207| If the mode is rm, and the zero is negative, return smallest
208| negative denorm.
209|
210	btstb		#5,FPCR_MODE(%a6) |test if rm or rp
211	beqs		no_dir
212	btstb		#4,FPCR_MODE(%a6) |check which one
213	beqs		zer_rm
214zer_rp:
215	tstb		L_SCR1(%a6)	|check sign
216	bnes		no_dir		|if set, neg op, no inc
217	movel		#1,FPTEMP_LO(%a6) |set lsb
218	bras		sm_dnrm
219zer_rm:
220	tstb		L_SCR1(%a6)	|check sign
221	beqs		no_dir		|if clr, neg op, no inc
222	movel		#1,FPTEMP_LO(%a6) |set lsb
223	orl		#neg_mask,USER_FPSR(%a6) |set N
224	bras		sm_dnrm
225no_dir:
226	fmovel		USER_FPCR(%a6),%FPCR
227	fmovex		FPTEMP(%a6),%fp0	|use fmove to set cc's
228	rts
229
230|
231| The rounding mode changed the zero to a smallest denorm. Call
232| t_resdnrm with exceptional operand in ETEMP.
233|
234sm_dnrm:
235	movel		FPTEMP_EX(%a6),ETEMP_EX(%a6)
236	movel		FPTEMP_HI(%a6),ETEMP_HI(%a6)
237	movel		FPTEMP_LO(%a6),ETEMP_LO(%a6)
238	leal		ETEMP(%a6),%a0
239	bra		t_resdnrm
240
241|
242| Result is still denormalized.
243|
244not_zero:
245	orl		#unfl_mask,USER_FPSR(%a6) |set unfl
246	tstb		L_SCR1(%a6)	|check for sign
247	beqs		fix_exit
248	orl		#neg_mask,USER_FPSR(%a6) |set N
249fix_exit:
250	bras		sm_dnrm
251
252
253|
254| The result has underflowed to zero. Return zero and set
255| unfl, aunfl, and ainex.
256|
257fix_unfl:
258	orl		#unfl_inx_mask,USER_FPSR(%a6)
259	btstb		#5,FPCR_MODE(%a6) |test if rm or rp
260	beqs		no_dir2
261	btstb		#4,FPCR_MODE(%a6) |check which one
262	beqs		zer_rm2
263zer_rp2:
264	tstb		L_SCR1(%a6)	|check sign
265	bnes		no_dir2		|if set, neg op, no inc
266	clrl		FPTEMP_EX(%a6)
267	clrl		FPTEMP_HI(%a6)
268	movel		#1,FPTEMP_LO(%a6) |set lsb
269	bras		sm_dnrm		|return smallest denorm
270zer_rm2:
271	tstb		L_SCR1(%a6)	|check sign
272	beqs		no_dir2		|if clr, neg op, no inc
273	movew		#0x8000,FPTEMP_EX(%a6)
274	clrl		FPTEMP_HI(%a6)
275	movel		#1,FPTEMP_LO(%a6) |set lsb
276	orl		#neg_mask,USER_FPSR(%a6) |set N
277	bra		sm_dnrm		|return smallest denorm
278
279no_dir2:
280	tstb		L_SCR1(%a6)
281	bges		pos_zero
282neg_zero:
283	clrl		FP_SCR1(%a6)	|clear the exceptional operand
284	clrl		FP_SCR1+4(%a6)	|for gen_except.
285	clrl		FP_SCR1+8(%a6)
286	fmoves		#0x80000000,%fp0
287	rts
288pos_zero:
289	clrl		FP_SCR1(%a6)	|clear the exceptional operand
290	clrl		FP_SCR1+4(%a6)	|for gen_except.
291	clrl		FP_SCR1+8(%a6)
292	fmoves		#0x00000000,%fp0
293	rts
294
295|
296| The destination is a denormalized number.  It must be handled
297| by first shifting the bits in the mantissa until it is normalized,
298| then adding the remainder of the source to the exponent.
299|
300dst_dnrm:
301	moveml		%d2/%d3,-(%a7)
302	movew		FPTEMP_EX(%a6),%d1
303	movel		FPTEMP_HI(%a6),%d2
304	movel		FPTEMP_LO(%a6),%d3
305dst_loop:
306	tstl		%d2		|test for normalized result
307	blts		dst_norm	|exit loop if so
308	tstl		%d0		|otherwise, test shift count
309	beqs		dst_fin		|if zero, shifting is done
310	subil		#1,%d0		|dec src
311	lsll		#1,%d3
312	roxll		#1,%d2
313	bras		dst_loop
314|
315| Destination became normalized.  Simply add the remaining
316| portion of the src to the exponent.
317|
318dst_norm:
319	addw		%d0,%d1		|dst is normalized; add src
320	tstb		L_SCR1(%a6)
321	beqs		dnrm_pos
322	orl		#0x8000,%d1
323dnrm_pos:
324	movemw		%d1,FPTEMP_EX(%a6)
325	moveml		%d2,FPTEMP_HI(%a6)
326	moveml		%d3,FPTEMP_LO(%a6)
327	fmovel		USER_FPCR(%a6),%FPCR
328	fmovex		FPTEMP(%a6),%fp0
329	moveml		(%a7)+,%d2/%d3
330	rts
331
332|
333| Destination remained denormalized.  Call t_excdnrm with
334| exceptional operand in ETEMP.
335|
336dst_fin:
337	tstb		L_SCR1(%a6)	|check for sign
338	beqs		dst_exit
339	orl		#neg_mask,USER_FPSR(%a6) |set N
340	orl		#0x8000,%d1
341dst_exit:
342	movemw		%d1,ETEMP_EX(%a6)
343	moveml		%d2,ETEMP_HI(%a6)
344	moveml		%d3,ETEMP_LO(%a6)
345	orl		#unfl_mask,USER_FPSR(%a6) |set unfl
346	moveml		(%a7)+,%d2/%d3
347	leal		ETEMP(%a6),%a0
348	bra		t_resdnrm
349
350|
351| Source is outside of 2^14 range.  Test the sign and branch
352| to the appropriate exception handler.
353|
354src_out:
355	tstb		L_SCR1(%a6)
356	beqs		scro_pos
357	orl		#0x8000,%d1
358scro_pos:
359	movel		FPTEMP_HI(%a6),ETEMP_HI(%a6)
360	movel		FPTEMP_LO(%a6),ETEMP_LO(%a6)
361	tstw		ETEMP(%a6)
362	blts		res_neg
363res_pos:
364	movew		%d1,ETEMP(%a6)	|result in ETEMP
365	bra		t_ovfl2
366res_neg:
367	movew		%d1,ETEMP(%a6)	|result in ETEMP
368	leal		ETEMP(%a6),%a0
369	bra		t_unfl
370	|end
371