summaryrefslogtreecommitdiff
path: root/arch/x86/math-emu/reg_u_sub.S
blob: f05dea7dec38f7e2ecacbaf3e5017507cc0bd48a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
/* SPDX-License-Identifier: GPL-2.0 */
	.file	"reg_u_sub.S"
/*---------------------------------------------------------------------------+
 |  reg_u_sub.S                                                              |
 |                                                                           |
 | Core floating point subtraction routine.                                  |
 |                                                                           |
 | Copyright (C) 1992,1993,1995,1997                                         |
 |                  W. Metzenthen, 22 Parker St, Ormond, Vic 3163, Australia |
 |                  E-mail   billm@suburbia.net                              |
 |                                                                           |
 | Call from C as:                                                           |
 |    int FPU_u_sub(FPU_REG *arg1, FPU_REG *arg2, FPU_REG *answ,             |
 |                                                int control_w)             |
 |    Return value is the tag of the answer, or-ed with FPU_Exception if     |
 |    one was raised, or -1 on internal error.                               |
 |                                                                           |
 +---------------------------------------------------------------------------*/

/*
 |    Kernel subtraction routine FPU_u_sub(reg *arg1, reg *arg2, reg *answ).
 |    Takes two valid reg f.p. numbers (TAG_Valid), which are
 |    treated as unsigned numbers,
 |    and returns their difference as a TAG_Valid or TAG_Zero f.p.
 |    number.
 |    The first number (arg1) must be the larger.
 |    The returned number is normalized.
 |    Basic checks are performed if PARANOID is defined.
 */

#include "exception.h"
#include "fpu_emu.h"
#include "control_w.h"

.text
ENTRY(FPU_u_sub)
	pushl	%ebp
	movl	%esp,%ebp
	pushl	%esi
	pushl	%edi
	pushl	%ebx

	movl	PARAM1,%esi	/* source 1 */
	movl	PARAM2,%edi	/* source 2 */
	
	movl	PARAM6,%ecx
	subl	PARAM7,%ecx	/* exp1 - exp2 */

#ifdef PARANOID
	/* source 2 is always smaller than source 1 */
	js	L_bugged_1

	testl	$0x80000000,SIGH(%edi)	/* The args are assumed to be be normalized */
	je	L_bugged_2

	testl	$0x80000000,SIGH(%esi)
	je	L_bugged_2
#endif /* PARANOID */

/*--------------------------------------+
 |	Form a register holding the     |
 |	smaller number                  |
 +--------------------------------------*/
	movl	SIGH(%edi),%eax	/* register ms word */
	movl	SIGL(%edi),%ebx	/* register ls word */

	movl	PARAM3,%edi	/* destination */
	movl	PARAM6,%edx
	movw	%dx,EXP(%edi)	/* Copy exponent to destination */

	xorl	%edx,%edx	/* register extension */

/*--------------------------------------+
 |	Shift the temporary register	|
 |      right the required number of	|
 |	places.				|
 +--------------------------------------*/

	cmpw	$32,%cx		/* shrd only works for 0..31 bits */
	jnc	L_more_than_31

/* less than 32 bits */
	shrd	%cl,%ebx,%edx
	shrd	%cl,%eax,%ebx
	shr	%cl,%eax
	jmp	L_shift_done

L_more_than_31:
	cmpw	$64,%cx
	jnc	L_more_than_63

	subb	$32,%cl
	jz	L_exactly_32

	shrd	%cl,%eax,%edx
	shr	%cl,%eax
	orl	%ebx,%ebx
	jz	L_more_31_no_low	/* none of the lowest bits is set */

	orl	$1,%edx			/* record the fact in the extension */

L_more_31_no_low:
	movl	%eax,%ebx
	xorl	%eax,%eax
	jmp	L_shift_done

L_exactly_32:
	movl	%ebx,%edx
	movl	%eax,%ebx
	xorl	%eax,%eax
	jmp	L_shift_done

L_more_than_63:
	cmpw	$65,%cx
	jnc	L_more_than_64

	/* Shift right by 64 bits */
	movl	%eax,%edx
	orl	%ebx,%ebx
	jz	L_more_63_no_low

	orl	$1,%edx
	jmp	L_more_63_no_low

L_more_than_64:
	jne	L_more_than_65

	/* Shift right by 65 bits */
	/* Carry is clear if we get here */
	movl	%eax,%edx
	rcrl	%edx
	jnc	L_shift_65_nc

	orl	$1,%edx
	jmp	L_more_63_no_low

L_shift_65_nc:
	orl	%ebx,%ebx
	jz	L_more_63_no_low

	orl	$1,%edx
	jmp	L_more_63_no_low

L_more_than_65:
	movl	$1,%edx		/* The shifted nr always at least one '1' */

L_more_63_no_low:
	xorl	%ebx,%ebx
	xorl	%eax,%eax

L_shift_done:
L_subtr:
/*------------------------------+
 |	Do the subtraction	|
 +------------------------------*/
	xorl	%ecx,%ecx
	subl	%edx,%ecx
	movl	%ecx,%edx
	movl	SIGL(%esi),%ecx
	sbbl	%ebx,%ecx
	movl	%ecx,%ebx
	movl	SIGH(%esi),%ecx
	sbbl	%eax,%ecx
	movl	%ecx,%eax

#ifdef PARANOID
	/* We can never get a borrow */
	jc	L_bugged
#endif /* PARANOID */

/*--------------------------------------+
 |	Normalize the result		|
 +--------------------------------------*/
	testl	$0x80000000,%eax
	jnz	L_round		/* no shifting needed */

	orl	%eax,%eax
	jnz	L_shift_1	/* shift left 1 - 31 bits */

	orl	%ebx,%ebx
	jnz	L_shift_32	/* shift left 32 - 63 bits */

/*
 *	 A rare case, the only one which is non-zero if we got here
 *         is:           1000000 .... 0000
 *                      -0111111 .... 1111 1
 *                       -------------------- 
 *                       0000000 .... 0000 1 
 */

	cmpl	$0x80000000,%edx
	jnz	L_must_be_zero

	/* Shift left 64 bits */
	subw	$64,EXP(%edi)
	xchg	%edx,%eax
	jmp	fpu_reg_round

L_must_be_zero:
#ifdef PARANOID
	orl	%edx,%edx
	jnz	L_bugged_3
#endif /* PARANOID */ 

	/* The result is zero */
	movw	$0,EXP(%edi)		/* exponent */
	movl	$0,SIGL(%edi)
	movl	$0,SIGH(%edi)
	movl	TAG_Zero,%eax
	jmp	L_exit

L_shift_32:
	movl	%ebx,%eax
	movl	%edx,%ebx
	movl	$0,%edx
	subw	$32,EXP(%edi)	/* Can get underflow here */

/* We need to shift left by 1 - 31 bits */
L_shift_1:
	bsrl	%eax,%ecx	/* get the required shift in %ecx */
	subl	$31,%ecx
	negl	%ecx
	shld	%cl,%ebx,%eax
	shld	%cl,%edx,%ebx
	shl	%cl,%edx
	subw	%cx,EXP(%edi)	/* Can get underflow here */

L_round:
	jmp	fpu_reg_round	/* Round the result */


#ifdef PARANOID
L_bugged_1:
	pushl	EX_INTERNAL|0x206
	call	EXCEPTION
	pop	%ebx
	jmp	L_error_exit

L_bugged_2:
	pushl	EX_INTERNAL|0x209
	call	EXCEPTION
	pop	%ebx
	jmp	L_error_exit

L_bugged_3:
	pushl	EX_INTERNAL|0x210
	call	EXCEPTION
	pop	%ebx
	jmp	L_error_exit

L_bugged_4:
	pushl	EX_INTERNAL|0x211
	call	EXCEPTION
	pop	%ebx
	jmp	L_error_exit

L_bugged:
	pushl	EX_INTERNAL|0x212
	call	EXCEPTION
	pop	%ebx
	jmp	L_error_exit

L_error_exit:
	movl	$-1,%eax

#endif /* PARANOID */

L_exit:
	popl	%ebx
	popl	%edi
	popl	%esi
	leave
	ret
ENDPROC(FPU_u_sub)