summaryrefslogtreecommitdiff
path: root/Source/Core/Common/x64CPUDetect.cpp
blob: e64cd9d1ee141b8c682dcdf4247ff915295d2f3b (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
275
// Copyright 2013 Dolphin Emulator Project
// Licensed under GPLv2
// Refer to the license.txt file included.

#include <cstring>
#include <string>

#include "Common/Common.h"
#include "Common/CPUDetect.h"

#ifdef _WIN32
#define _interlockedbittestandset workaround_ms_header_bug_platform_sdk6_set
#define _interlockedbittestandreset workaround_ms_header_bug_platform_sdk6_reset
#define _interlockedbittestandset64 workaround_ms_header_bug_platform_sdk6_set64
#define _interlockedbittestandreset64 workaround_ms_header_bug_platform_sdk6_reset64
#include <intrin.h>
#undef _interlockedbittestandset
#undef _interlockedbittestandreset
#undef _interlockedbittestandset64
#undef _interlockedbittestandreset64
#else

//#include <config/i386/cpuid.h>
#ifndef _M_GENERIC
#include <xmmintrin.h>
#endif

#if defined __FreeBSD__
#include <sys/types.h>
#include <machine/cpufunc.h>
#else
static inline void do_cpuid(unsigned int *eax, unsigned int *ebx,
							unsigned int *ecx, unsigned int *edx)
{
#if defined _M_GENERIC
	(*eax) = (*ebx) = (*ecx) = (*edx) = 0;
#elif defined _LP64
	// Note: EBX is reserved on Mac OS X and in PIC on Linux, so it has to
	// restored at the end of the asm block.
	__asm__ (
		"cpuid;"
		"movl  %%ebx,%1;"
		: "=a" (*eax),
		  "=S" (*ebx),
		  "=c" (*ecx),
		  "=d" (*edx)
		: "a"  (*eax)
		: "rbx"
		);
#else
	__asm__ (
		"cpuid;"
		"movl  %%ebx,%1;"
		: "=a" (*eax),
		  "=S" (*ebx),
		  "=c" (*ecx),
		  "=d" (*edx)
		: "a"  (*eax)
		: "ebx"
		);
#endif
}
#endif /* defined __FreeBSD__ */

static void __cpuid(int info[4], int x)
{
#if defined __FreeBSD__
	do_cpuid((unsigned int)x, (unsigned int*)info);
#else
	unsigned int eax = x, ebx = 0, ecx = 0, edx = 0;
	do_cpuid(&eax, &ebx, &ecx, &edx);
	info[0] = eax;
	info[1] = ebx;
	info[2] = ecx;
	info[3] = edx;
#endif
}

#define _XCR_XFEATURE_ENABLED_MASK 0
static unsigned long long _xgetbv(unsigned int index)
{
#ifndef _M_GENERIC
	unsigned int eax, edx;
	__asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(index));
	return ((unsigned long long)edx << 32) | eax;
#endif
}

#endif

CPUInfo cpu_info;

CPUInfo::CPUInfo() {
	Detect();
}

// Detects the various cpu features
void CPUInfo::Detect()
{
	memset(this, 0, sizeof(*this));
#if _M_X86_32
	Mode64bit = false;
#elif _M_X86_64
	Mode64bit = true;
	OS64bit = true;
#endif
	num_cores = 1;

#ifdef _WIN32
#if _M_X86_32
	BOOL f64 = false;
	IsWow64Process(GetCurrentProcess(), &f64);
	OS64bit = (f64 == TRUE) ? true : false;
#endif
#endif

	// Set obvious defaults, for extra safety
	if (Mode64bit) {
		bSSE = true;
		bSSE2 = true;
		bLongMode = true;
	}

	// Assume CPU supports the CPUID instruction. Those that don't can barely
	// boot modern OS:es anyway.
	int cpu_id[4];
	memset(cpu_string, 0, sizeof(cpu_string));

	// Detect CPU's CPUID capabilities, and grab cpu string
	__cpuid(cpu_id, 0x00000000);
	u32 max_std_fn = cpu_id[0];  // EAX
	*((int *)cpu_string) = cpu_id[1];
	*((int *)(cpu_string + 4)) = cpu_id[3];
	*((int *)(cpu_string + 8)) = cpu_id[2];
	__cpuid(cpu_id, 0x80000000);
	u32 max_ex_fn = cpu_id[0];
	if (!strcmp(cpu_string, "GenuineIntel"))
		vendor = VENDOR_INTEL;
	else if (!strcmp(cpu_string, "AuthenticAMD"))
		vendor = VENDOR_AMD;
	else
		vendor = VENDOR_OTHER;

	// Set reasonable default brand string even if brand string not available.
	strcpy(brand_string, cpu_string);

	// Detect family and other misc stuff.
	bool ht = false;
	HTT = ht;
	logical_cpu_count = 1;
	if (max_std_fn >= 1) {
		__cpuid(cpu_id, 0x00000001);
		logical_cpu_count = (cpu_id[1] >> 16) & 0xFF;
		ht = (cpu_id[3] >> 28) & 1;

		if ((cpu_id[3] >> 25) & 1) bSSE = true;
		if ((cpu_id[3] >> 26) & 1) bSSE2 = true;
		if ((cpu_id[2])       & 1) bSSE3 = true;
		if ((cpu_id[2] >> 9)  & 1) bSSSE3 = true;
		if ((cpu_id[2] >> 19) & 1) bSSE4_1 = true;
		if ((cpu_id[2] >> 20) & 1) bSSE4_2 = true;
		if ((cpu_id[2] >> 22) & 1) bMOVBE = true;
		if ((cpu_id[2] >> 25) & 1) bAES = true;

		// To check DAZ support, we first need to check FXSAVE support.
		if ((cpu_id[3] >> 24) & 1)
		{
			// We can use FXSAVE.
			bFXSR = true;

			GC_ALIGNED16(u8 fx_state[512]);
			memset(fx_state, 0, sizeof(fx_state));
#ifdef _WIN32
#if _M_X86_32
			_fxsave(fx_state);
#elif _M_X86_64
			_fxsave64(fx_state);
#endif
#else
			__asm__("fxsave %0" : "=m" (fx_state));
#endif

			// lowest byte of MXCSR_MASK
			if ((fx_state[0x1C] >> 6) & 1)
			{
				// On x86, the FTZ field (supported since SSE1) only flushes denormal _outputs_ to zero,
				// now that we checked DAZ support (flushing denormal _inputs_ to zero),
				// we can set our generic flag.
				bFlushToZero = true;
			}
		}

		// AVX support requires 3 separate checks:
		//  - Is the AVX bit set in CPUID?
		//  - Is the XSAVE bit set in CPUID?
		//  - XGETBV result has the XCR bit set.
		if (((cpu_id[2] >> 28) & 1) && ((cpu_id[2] >> 27) & 1))
		{
			if ((_xgetbv(_XCR_XFEATURE_ENABLED_MASK) & 0x6) == 0x6)
			{
				bAVX = true;
				if ((cpu_id[2] >> 12) & 1)
					bFMA = true;
			}
		}
	}
	if (max_ex_fn >= 0x80000004) {
		// Extract brand string
		__cpuid(cpu_id, 0x80000002);
		memcpy(brand_string, cpu_id, sizeof(cpu_id));
		__cpuid(cpu_id, 0x80000003);
		memcpy(brand_string + 16, cpu_id, sizeof(cpu_id));
		__cpuid(cpu_id, 0x80000004);
		memcpy(brand_string + 32, cpu_id, sizeof(cpu_id));
	}
	if (max_ex_fn >= 0x80000001) {
		// Check for more features.
		__cpuid(cpu_id, 0x80000001);
		if (cpu_id[2] & 1) bLAHFSAHF64 = true;
		if ((cpu_id[3] >> 29) & 1) bLongMode = true;
	}

	num_cores = (logical_cpu_count == 0) ? 1 : logical_cpu_count;

	if (max_ex_fn >= 0x80000008) {
		// Get number of cores. This is a bit complicated. Following AMD manual here.
		__cpuid(cpu_id, 0x80000008);
		int apic_id_core_id_size = (cpu_id[2] >> 12) & 0xF;
		if (apic_id_core_id_size == 0) {
			if (ht) {
				// New mechanism for modern Intel CPUs.
				if (vendor == VENDOR_INTEL) {
					__cpuid(cpu_id, 0x00000004);
					int cores_x_package = ((cpu_id[0] >> 26) & 0x3F) + 1;
					HTT = (cores_x_package < logical_cpu_count);
					cores_x_package = ((logical_cpu_count % cores_x_package) == 0) ? cores_x_package : 1;
					num_cores = (cores_x_package > 1) ? cores_x_package : num_cores;
					logical_cpu_count /= cores_x_package;
				}
			}
		} else {
			// Use AMD's new method.
			num_cores = (cpu_id[2] & 0xFF) + 1;
		}
	}
}

// Turn the cpu info into a string we can show
std::string CPUInfo::Summarize()
{
	std::string sum(cpu_string);
	if (bSSE) sum += ", SSE";
	if (bSSE2)
	{
		sum += ", SSE2";
		if (!bFlushToZero)
			sum += " (but not DAZ!)";
	}
	if (bSSE3) sum += ", SSE3";
	if (bSSSE3) sum += ", SSSE3";
	if (bSSE4_1) sum += ", SSE4.1";
	if (bSSE4_2) sum += ", SSE4.2";
	if (HTT) sum += ", HTT";
	if (bAVX) sum += ", AVX";
	if (bFMA) sum += ", FMA";
	if (bAES) sum += ", AES";
	if (bMOVBE) sum += ", MOVBE";
	if (bLongMode) sum += ", 64-bit support";
	return sum;
}

bool CPUInfo::IsUnsafe()
{
	return !bFlushToZero;
}