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
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
|
/* Initialize CPU feature data.
This file is part of the GNU C Library.
Copyright (C) 2008-2017 Free Software Foundation, Inc.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library 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
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
#include <cpuid.h>
#include <cpu-features.h>
#include <dl-hwcap.h>
static void
get_common_indeces (struct cpu_features *cpu_features,
unsigned int *family, unsigned int *model,
unsigned int *extended_model, unsigned int *stepping)
{
if (family)
{
unsigned int eax;
__cpuid (1, eax, cpu_features->cpuid[COMMON_CPUID_INDEX_1].ebx,
cpu_features->cpuid[COMMON_CPUID_INDEX_1].ecx,
cpu_features->cpuid[COMMON_CPUID_INDEX_1].edx);
cpu_features->cpuid[COMMON_CPUID_INDEX_1].eax = eax;
*family = (eax >> 8) & 0x0f;
*model = (eax >> 4) & 0x0f;
*extended_model = (eax >> 12) & 0xf0;
*stepping = eax & 0x0f;
if (*family == 0x0f)
{
*family += (eax >> 20) & 0xff;
*model += *extended_model;
}
}
if (cpu_features->max_cpuid >= 7)
__cpuid_count (7, 0,
cpu_features->cpuid[COMMON_CPUID_INDEX_7].eax,
cpu_features->cpuid[COMMON_CPUID_INDEX_7].ebx,
cpu_features->cpuid[COMMON_CPUID_INDEX_7].ecx,
cpu_features->cpuid[COMMON_CPUID_INDEX_7].edx);
/* Can we call xgetbv? */
if (CPU_FEATURES_CPU_P (cpu_features, OSXSAVE))
{
unsigned int xcrlow;
unsigned int xcrhigh;
asm ("xgetbv" : "=a" (xcrlow), "=d" (xcrhigh) : "c" (0));
/* Is YMM and XMM state usable? */
if ((xcrlow & (bit_YMM_state | bit_XMM_state)) ==
(bit_YMM_state | bit_XMM_state))
{
/* Determine if AVX is usable. */
if (CPU_FEATURES_CPU_P (cpu_features, AVX))
{
cpu_features->feature[index_arch_AVX_Usable]
|= bit_arch_AVX_Usable;
/* The following features depend on AVX being usable. */
/* Determine if AVX2 is usable. */
if (CPU_FEATURES_CPU_P (cpu_features, AVX2))
cpu_features->feature[index_arch_AVX2_Usable]
|= bit_arch_AVX2_Usable;
/* Determine if FMA is usable. */
if (CPU_FEATURES_CPU_P (cpu_features, FMA))
cpu_features->feature[index_arch_FMA_Usable]
|= bit_arch_FMA_Usable;
}
/* Check if OPMASK state, upper 256-bit of ZMM0-ZMM15 and
ZMM16-ZMM31 state are enabled. */
if ((xcrlow & (bit_Opmask_state | bit_ZMM0_15_state
| bit_ZMM16_31_state)) ==
(bit_Opmask_state | bit_ZMM0_15_state | bit_ZMM16_31_state))
{
/* Determine if AVX512F is usable. */
if (CPU_FEATURES_CPU_P (cpu_features, AVX512F))
{
cpu_features->feature[index_arch_AVX512F_Usable]
|= bit_arch_AVX512F_Usable;
/* Determine if AVX512DQ is usable. */
if (CPU_FEATURES_CPU_P (cpu_features, AVX512DQ))
cpu_features->feature[index_arch_AVX512DQ_Usable]
|= bit_arch_AVX512DQ_Usable;
}
}
}
}
}
static inline void
init_cpu_features (struct cpu_features *cpu_features)
{
unsigned int ebx, ecx, edx;
unsigned int family = 0;
unsigned int model = 0;
enum cpu_features_kind kind;
#if !HAS_CPUID
if (__get_cpuid_max (0, 0) == 0)
{
kind = arch_kind_other;
goto no_cpuid;
}
#endif
__cpuid (0, cpu_features->max_cpuid, ebx, ecx, edx);
/* This spells out "GenuineIntel". */
if (ebx == 0x756e6547 && ecx == 0x6c65746e && edx == 0x49656e69)
{
unsigned int extended_model, stepping;
kind = arch_kind_intel;
get_common_indeces (cpu_features, &family, &model, &extended_model,
&stepping);
if (family == 0x06)
{
model += extended_model;
switch (model)
{
case 0x1c:
case 0x26:
/* BSF is slow on Atom. */
cpu_features->feature[index_arch_Slow_BSF]
|= bit_arch_Slow_BSF;
break;
case 0x57:
/* Knights Landing. Enable Silvermont optimizations. */
cpu_features->feature[index_arch_Prefer_No_VZEROUPPER]
|= bit_arch_Prefer_No_VZEROUPPER;
case 0x5c:
case 0x5f:
/* Unaligned load versions are faster than SSSE3
on Goldmont. */
case 0x4c:
/* Airmont is a die shrink of Silvermont. */
case 0x37:
case 0x4a:
case 0x4d:
case 0x5a:
case 0x5d:
/* Unaligned load versions are faster than SSSE3
on Silvermont. */
#if index_arch_Fast_Unaligned_Load != index_arch_Prefer_PMINUB_for_stringop
# error index_arch_Fast_Unaligned_Load != index_arch_Prefer_PMINUB_for_stringop
#endif
#if index_arch_Fast_Unaligned_Load != index_arch_Slow_SSE4_2
# error index_arch_Fast_Unaligned_Load != index_arch_Slow_SSE4_2
#endif
#if index_arch_Fast_Unaligned_Load != index_arch_Fast_Unaligned_Copy
# error index_arch_Fast_Unaligned_Load != index_arch_Fast_Unaligned_Copy
#endif
cpu_features->feature[index_arch_Fast_Unaligned_Load]
|= (bit_arch_Fast_Unaligned_Load
| bit_arch_Fast_Unaligned_Copy
| bit_arch_Prefer_PMINUB_for_stringop
| bit_arch_Slow_SSE4_2);
break;
default:
/* Unknown family 0x06 processors. Assuming this is one
of Core i3/i5/i7 processors if AVX is available. */
if (!CPU_FEATURES_CPU_P (cpu_features, AVX))
break;
case 0x1a:
case 0x1e:
case 0x1f:
case 0x25:
case 0x2c:
case 0x2e:
case 0x2f:
/* Rep string instructions, unaligned load, unaligned copy,
and pminub are fast on Intel Core i3, i5 and i7. */
#if index_arch_Fast_Rep_String != index_arch_Fast_Unaligned_Load
# error index_arch_Fast_Rep_String != index_arch_Fast_Unaligned_Load
#endif
#if index_arch_Fast_Rep_String != index_arch_Prefer_PMINUB_for_stringop
# error index_arch_Fast_Rep_String != index_arch_Prefer_PMINUB_for_stringop
#endif
#if index_arch_Fast_Rep_String != index_arch_Fast_Unaligned_Copy
# error index_arch_Fast_Rep_String != index_arch_Fast_Unaligned_Copy
#endif
cpu_features->feature[index_arch_Fast_Rep_String]
|= (bit_arch_Fast_Rep_String
| bit_arch_Fast_Unaligned_Load
| bit_arch_Fast_Unaligned_Copy
| bit_arch_Prefer_PMINUB_for_stringop);
break;
case 0x3f:
/* Xeon E7 v3 with stepping >= 4 has working TSX. */
if (stepping >= 4)
break;
case 0x3c:
case 0x45:
case 0x46:
/* Disable Intel TSX on Haswell processors (except Xeon E7 v3
with stepping >= 4) to avoid TSX on kernels that weren't
updated with the latest microcode package (which disables
broken feature by default). */
cpu_features->cpuid[index_cpu_RTM].reg_RTM &= ~bit_cpu_RTM;
break;
}
}
/* Unaligned load with 256-bit AVX registers are faster on
Intel processors with AVX2. */
if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable))
cpu_features->feature[index_arch_AVX_Fast_Unaligned_Load]
|= bit_arch_AVX_Fast_Unaligned_Load;
/* To avoid SSE transition penalty, use _dl_runtime_resolve_slow.
If XGETBV suports ECX == 1, use _dl_runtime_resolve_opt. */
cpu_features->feature[index_arch_Use_dl_runtime_resolve_slow]
|= bit_arch_Use_dl_runtime_resolve_slow;
if (cpu_features->max_cpuid >= 0xd)
{
unsigned int eax;
__cpuid_count (0xd, 1, eax, ebx, ecx, edx);
if ((eax & (1 << 2)) != 0)
cpu_features->feature[index_arch_Use_dl_runtime_resolve_opt]
|= bit_arch_Use_dl_runtime_resolve_opt;
}
}
/* This spells out "AuthenticAMD". */
else if (ebx == 0x68747541 && ecx == 0x444d4163 && edx == 0x69746e65)
{
unsigned int extended_model, stepping;
kind = arch_kind_amd;
get_common_indeces (cpu_features, &family, &model, &extended_model,
&stepping);
ecx = cpu_features->cpuid[COMMON_CPUID_INDEX_1].ecx;
unsigned int eax;
__cpuid (0x80000000, eax, ebx, ecx, edx);
if (eax >= 0x80000001)
__cpuid (0x80000001,
cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].eax,
cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].ebx,
cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].ecx,
cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].edx);
if (HAS_ARCH_FEATURE (AVX_Usable))
{
/* Since the FMA4 bit is in COMMON_CPUID_INDEX_80000001 and
FMA4 requires AVX, determine if FMA4 is usable here. */
if (CPU_FEATURES_CPU_P (cpu_features, FMA4))
cpu_features->feature[index_arch_FMA4_Usable]
|= bit_arch_FMA4_Usable;
}
if (family == 0x15)
{
#if index_arch_Fast_Unaligned_Load != index_arch_Fast_Copy_Backward
# error index_arch_Fast_Unaligned_Load != index_arch_Fast_Copy_Backward
#endif
/* "Excavator" */
if (model >= 0x60 && model <= 0x7f)
cpu_features->feature[index_arch_Fast_Unaligned_Load]
|= (bit_arch_Fast_Unaligned_Load
| bit_arch_Fast_Copy_Backward);
}
}
else
{
kind = arch_kind_other;
get_common_indeces (cpu_features, NULL, NULL, NULL, NULL);
}
/* Support i586 if CX8 is available. */
if (CPU_FEATURES_CPU_P (cpu_features, CX8))
cpu_features->feature[index_arch_I586] |= bit_arch_I586;
/* Support i686 if CMOV is available. */
if (CPU_FEATURES_CPU_P (cpu_features, CMOV))
cpu_features->feature[index_arch_I686] |= bit_arch_I686;
#if !HAS_CPUID
no_cpuid:
#endif
cpu_features->family = family;
cpu_features->model = model;
cpu_features->kind = kind;
/* Reuse dl_hwcap and dl_hwcap_mask for x86. */
GLRO(dl_hwcap) = 0;
if (CPU_FEATURES_CPU_P (cpu_features, SSE2))
GLRO(dl_hwcap) |= HWCAP_X86_SSE2;
if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable))
GLRO(dl_hwcap) |= HWCAP_X86_AVX2;
GLRO(dl_hwcap_mask) = HWCAP_IMPORTANT;
}
|
