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
275
276
277
278
279
280
281
282
283
284
285
286
287
|
/*
* Copyright(c) 2019-2021 Qualcomm Innovation Center, Inc. All Rights Reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
#include "fpu/softfloat.h"
#include "cpu.h"
#include "fma_emu.h"
#include "arch.h"
#include "macros.h"
#define SF_BIAS 127
#define SF_MAXEXP 254
#define SF_MANTBITS 23
#define float32_nan make_float32(0xffffffff)
#define BITS_MASK_8 0x5555555555555555ULL
#define PAIR_MASK_8 0x3333333333333333ULL
#define NYBL_MASK_8 0x0f0f0f0f0f0f0f0fULL
#define BYTE_MASK_8 0x00ff00ff00ff00ffULL
#define HALF_MASK_8 0x0000ffff0000ffffULL
#define WORD_MASK_8 0x00000000ffffffffULL
uint64_t interleave(uint32_t odd, uint32_t even)
{
/* Convert to long long */
uint64_t myodd = odd;
uint64_t myeven = even;
/* First, spread bits out */
myodd = (myodd | (myodd << 16)) & HALF_MASK_8;
myeven = (myeven | (myeven << 16)) & HALF_MASK_8;
myodd = (myodd | (myodd << 8)) & BYTE_MASK_8;
myeven = (myeven | (myeven << 8)) & BYTE_MASK_8;
myodd = (myodd | (myodd << 4)) & NYBL_MASK_8;
myeven = (myeven | (myeven << 4)) & NYBL_MASK_8;
myodd = (myodd | (myodd << 2)) & PAIR_MASK_8;
myeven = (myeven | (myeven << 2)) & PAIR_MASK_8;
myodd = (myodd | (myodd << 1)) & BITS_MASK_8;
myeven = (myeven | (myeven << 1)) & BITS_MASK_8;
/* Now OR together */
return myeven | (myodd << 1);
}
uint64_t deinterleave(uint64_t src)
{
/* Get odd and even bits */
uint64_t myodd = ((src >> 1) & BITS_MASK_8);
uint64_t myeven = (src & BITS_MASK_8);
/* Unspread bits */
myeven = (myeven | (myeven >> 1)) & PAIR_MASK_8;
myodd = (myodd | (myodd >> 1)) & PAIR_MASK_8;
myeven = (myeven | (myeven >> 2)) & NYBL_MASK_8;
myodd = (myodd | (myodd >> 2)) & NYBL_MASK_8;
myeven = (myeven | (myeven >> 4)) & BYTE_MASK_8;
myodd = (myodd | (myodd >> 4)) & BYTE_MASK_8;
myeven = (myeven | (myeven >> 8)) & HALF_MASK_8;
myodd = (myodd | (myodd >> 8)) & HALF_MASK_8;
myeven = (myeven | (myeven >> 16)) & WORD_MASK_8;
myodd = (myodd | (myodd >> 16)) & WORD_MASK_8;
/* Return odd bits in upper half */
return myeven | (myodd << 32);
}
int32_t conv_round(int32_t a, int n)
{
int64_t val;
if (n == 0) {
val = a;
} else if ((a & ((1 << (n - 1)) - 1)) == 0) { /* N-1..0 all zero? */
/* Add LSB from int part */
val = ((fSE32_64(a)) + (int64_t) (((uint32_t) ((1 << n) & a)) >> 1));
} else {
val = ((fSE32_64(a)) + (1 << (n - 1)));
}
val = val >> n;
return (int32_t)val;
}
/* Floating Point Stuff */
static const FloatRoundMode softfloat_roundingmodes[] = {
float_round_nearest_even,
float_round_to_zero,
float_round_down,
float_round_up,
};
void arch_fpop_start(CPUHexagonState *env)
{
set_float_exception_flags(0, &env->fp_status);
set_float_rounding_mode(
softfloat_roundingmodes[fREAD_REG_FIELD(USR, USR_FPRND)],
&env->fp_status);
}
#ifdef CONFIG_USER_ONLY
/*
* Hexagon Linux kernel only sets the relevant bits in USR (user status
* register). The exception isn't raised to user mode, so we don't
* model it in qemu user mode.
*/
#define RAISE_FP_EXCEPTION do {} while (0)
#endif
#define SOFTFLOAT_TEST_FLAG(FLAG, MYF, MYE) \
do { \
if (flags & FLAG) { \
if (GET_USR_FIELD(USR_##MYF) == 0) { \
SET_USR_FIELD(USR_##MYF, 1); \
if (GET_USR_FIELD(USR_##MYE)) { \
RAISE_FP_EXCEPTION; \
} \
} \
} \
} while (0)
void arch_fpop_end(CPUHexagonState *env)
{
int flags = get_float_exception_flags(&env->fp_status);
if (flags != 0) {
SOFTFLOAT_TEST_FLAG(float_flag_inexact, FPINPF, FPINPE);
SOFTFLOAT_TEST_FLAG(float_flag_divbyzero, FPDBZF, FPDBZE);
SOFTFLOAT_TEST_FLAG(float_flag_invalid, FPINVF, FPINVE);
SOFTFLOAT_TEST_FLAG(float_flag_overflow, FPOVFF, FPOVFE);
SOFTFLOAT_TEST_FLAG(float_flag_underflow, FPUNFF, FPUNFE);
}
}
static float32 float32_mul_pow2(float32 a, uint32_t p, float_status *fp_status)
{
float32 b = make_float32((SF_BIAS + p) << SF_MANTBITS);
return float32_mul(a, b, fp_status);
}
int arch_sf_recip_common(float32 *Rs, float32 *Rt, float32 *Rd, int *adjust,
float_status *fp_status)
{
int n_exp;
int d_exp;
int ret = 0;
float32 RsV, RtV, RdV;
int PeV = 0;
RsV = *Rs;
RtV = *Rt;
if (float32_is_any_nan(RsV) && float32_is_any_nan(RtV)) {
if (extract32(RsV & RtV, 22, 1) == 0) {
float_raise(float_flag_invalid, fp_status);
}
RdV = RsV = RtV = float32_nan;
} else if (float32_is_any_nan(RsV)) {
if (extract32(RsV, 22, 1) == 0) {
float_raise(float_flag_invalid, fp_status);
}
RdV = RsV = RtV = float32_nan;
} else if (float32_is_any_nan(RtV)) {
/* or put NaN in num/den fixup? */
if (extract32(RtV, 22, 1) == 0) {
float_raise(float_flag_invalid, fp_status);
}
RdV = RsV = RtV = float32_nan;
} else if (float32_is_infinity(RsV) && float32_is_infinity(RtV)) {
/* or put Inf in num fixup? */
RdV = RsV = RtV = float32_nan;
float_raise(float_flag_invalid, fp_status);
} else if (float32_is_zero(RsV) && float32_is_zero(RtV)) {
/* or put zero in num fixup? */
RdV = RsV = RtV = float32_nan;
float_raise(float_flag_invalid, fp_status);
} else if (float32_is_zero(RtV)) {
/* or put Inf in num fixup? */
uint8_t RsV_sign = float32_is_neg(RsV);
uint8_t RtV_sign = float32_is_neg(RtV);
RsV = infinite_float32(RsV_sign ^ RtV_sign);
RtV = float32_one;
RdV = float32_one;
if (float32_is_infinity(RsV)) {
float_raise(float_flag_divbyzero, fp_status);
}
} else if (float32_is_infinity(RtV)) {
RsV = make_float32(0x80000000 & (RsV ^ RtV));
RtV = float32_one;
RdV = float32_one;
} else if (float32_is_zero(RsV)) {
/* Does this just work itself out? */
/* No, 0/Inf causes problems. */
RsV = make_float32(0x80000000 & (RsV ^ RtV));
RtV = float32_one;
RdV = float32_one;
} else if (float32_is_infinity(RsV)) {
uint8_t RsV_sign = float32_is_neg(RsV);
uint8_t RtV_sign = float32_is_neg(RtV);
RsV = infinite_float32(RsV_sign ^ RtV_sign);
RtV = float32_one;
RdV = float32_one;
} else {
PeV = 0x00;
/* Basic checks passed */
n_exp = float32_getexp(RsV);
d_exp = float32_getexp(RtV);
if ((n_exp - d_exp + SF_BIAS) <= SF_MANTBITS) {
/* Near quotient underflow / inexact Q */
PeV = 0x80;
RtV = float32_mul_pow2(RtV, -64, fp_status);
RsV = float32_mul_pow2(RsV, 64, fp_status);
} else if ((n_exp - d_exp + SF_BIAS) > (SF_MAXEXP - 24)) {
/* Near quotient overflow */
PeV = 0x40;
RtV = float32_mul_pow2(RtV, 32, fp_status);
RsV = float32_mul_pow2(RsV, -32, fp_status);
} else if (n_exp <= SF_MANTBITS + 2) {
RtV = float32_mul_pow2(RtV, 64, fp_status);
RsV = float32_mul_pow2(RsV, 64, fp_status);
} else if (d_exp <= 1) {
RtV = float32_mul_pow2(RtV, 32, fp_status);
RsV = float32_mul_pow2(RsV, 32, fp_status);
} else if (d_exp > 252) {
RtV = float32_mul_pow2(RtV, -32, fp_status);
RsV = float32_mul_pow2(RsV, -32, fp_status);
}
RdV = 0;
ret = 1;
}
*Rs = RsV;
*Rt = RtV;
*Rd = RdV;
*adjust = PeV;
return ret;
}
int arch_sf_invsqrt_common(float32 *Rs, float32 *Rd, int *adjust,
float_status *fp_status)
{
float32 RsV, RdV;
int PeV = 0;
int r_exp;
int ret = 0;
RsV = *Rs;
if (float32_is_infinity(RsV)) {
if (extract32(RsV, 22, 1) == 0) {
float_raise(float_flag_invalid, fp_status);
}
RdV = RsV = float32_nan;
} else if (float32_lt(RsV, float32_zero, fp_status)) {
/* Negative nonzero values are NaN */
float_raise(float_flag_invalid, fp_status);
RsV = float32_nan;
RdV = float32_nan;
} else if (float32_is_infinity(RsV)) {
/* or put Inf in num fixup? */
RsV = infinite_float32(1);
RdV = infinite_float32(1);
} else if (float32_is_zero(RsV)) {
/* or put zero in num fixup? */
RdV = float32_one;
} else {
PeV = 0x00;
/* Basic checks passed */
r_exp = float32_getexp(RsV);
if (r_exp <= 24) {
RsV = float32_mul_pow2(RsV, 64, fp_status);
PeV = 0xe0;
}
RdV = 0;
ret = 1;
}
*Rs = RsV;
*Rd = RdV;
*adjust = PeV;
return ret;
}
|
