src/CPUFeatures_Win32.cpp

   1 ///////////////////////////////////////////////////////////////////////////////
   2 // MuldeR's Utilities for Qt
   3 // Copyright (C) 2004-2023 LoRd_MuldeR <MuldeR2@GMX.de>
   4 //
   5 // This library is free software; you can redistribute it and/or
   6 // modify it under the terms of the GNU Lesser General Public
   7 // License as published by the Free Software Foundation; either
   8 // version 2.1 of the License, or (at your option) any later version.
   9 //
  10 // This library is distributed in the hope that it will be useful,
  11 // but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13 // Lesser General Public License for more details.
  14 //
  15 // You should have received a copy of the GNU Lesser General Public
  16 // License along with this library; if not, write to the Free Software
  17 // Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  18 //
  19 // http://www.gnu.org/licenses/lgpl-2.1.txt
  20 //////////////////////////////////////////////////////////////////////////////////
  21
  22 //Win32 API
  23 #define WIN32_LEAN_AND_MEAN 1
  24 #include <Windows.h>
  25
  26 //ASM
  27 #include <intrin.h>
  28
  29 //MUtils
  30 #include <MUtils/CPUFeatures.h>
  31 #include <MUtils/OSSupport.h>
  32 #include "Utils_Win32.h"
  33
  34 #define MY_CPUID(X,Y) __cpuid(((int*)(X)), ((int)(Y)))
  35 #define CHECK_VENDOR(X,Y,Z) (_stricmp((X), (Y)) ? 0U : (Z));
  36 #define CHECK_FLAG(X,Y,Z) (((X) & (Y)) ? (Z) : 0U)
  37
  38 MUtils::CPUFetaures::cpu_info_t MUtils::CPUFetaures::detect(void)
  39 {
  40         const OS::ArgumentMap &args = OS::arguments();
  41         typedef BOOL (WINAPI *IsWow64ProcessFun)(__in HANDLE hProcess, __out PBOOL Wow64Process);
  42         static const quint32 FLAGS_X64 = (FLAG_MMX | FLAG_SSE | FLAG_SSE2);
  43
  44         cpu_info_t  features;
  45         SYSTEM_INFO systemInfo;
  46         uint32_t    cpuInfo[4];
  47
  48         //Initialize variables to zero
  49         memset(&features,   0, sizeof(cpu_info_t));
  50         memset(&systemInfo, 0, sizeof(SYSTEM_INFO));
  51         memset(&cpuInfo[0], 0, sizeof(cpuInfo));
  52
  53         //Detect the CPU identifier string
  54         MY_CPUID(&cpuInfo[0], 0);
  55         const uint32_t max_basic_cap = cpuInfo[0];
  56         memcpy(&features.idstr[0U * sizeof(uint32_t)], &cpuInfo[1], sizeof(uint32_t));
  57         memcpy(&features.idstr[1U * sizeof(uint32_t)], &cpuInfo[3], sizeof(uint32_t));
  58         memcpy(&features.idstr[2U * sizeof(uint32_t)], &cpuInfo[2], sizeof(uint32_t));
  59         features.idstr[3U * sizeof(uint32_t)] = '\0';
  60         features.vendor |= CHECK_VENDOR(features.idstr, "GenuineIntel", VENDOR_INTEL);
  61         features.vendor |= CHECK_VENDOR(features.idstr, "AuthenticAMD", VENDOR_AMD);
  62
  63         //Detect the CPU model and feature flags
  64         if(max_basic_cap >= 1)
  65         {
  66                 MY_CPUID(&cpuInfo[0], 1);
  67                 features.features |= CHECK_FLAG(cpuInfo[3], 0x00008000, FLAG_CMOV);
  68                 features.features |= CHECK_FLAG(cpuInfo[3], 0x00800000, FLAG_MMX);
  69                 features.features |= CHECK_FLAG(cpuInfo[3], 0x02000000, FLAG_SSE);
  70                 features.features |= CHECK_FLAG(cpuInfo[3], 0x04000000, FLAG_SSE2);
  71                 features.features |= CHECK_FLAG(cpuInfo[2], 0x00000001, FLAG_SSE3);
  72                 features.features |= CHECK_FLAG(cpuInfo[2], 0x00000200, FLAG_SSSE3);
  73                 features.features |= CHECK_FLAG(cpuInfo[2], 0x00080000, FLAG_SSE41);
  74                 features.features |= CHECK_FLAG(cpuInfo[2], 0x00100000, FLAG_SSE42);
  75
  76                 //Check for AVX
  77                 if ((cpuInfo[2] & 0x18000000) == 0x18000000)
  78                 {
  79                         if((_xgetbv(0) & 0x6ui64) == 0x6ui64) /*AVX requires OS support!*/
  80                         {
  81                                 features.features |= FLAG_AVX;
  82                                 features.features |= CHECK_FLAG(cpuInfo[2], 0x00001000, FLAG_FMA3);
  83                         }
  84                 }
  85
  86                 //Compute the CPU stepping, model and family
  87                 features.stepping = cpuInfo[0] & 0xf;
  88                 features.model    = ((cpuInfo[0] >> 4) & 0xf) + (((cpuInfo[0] >> 16) & 0xf) << 4);
  89                 features.family   = ((cpuInfo[0] >> 8) & 0xf) + ((cpuInfo[0] >> 20) & 0xff);
  90         }
  91
  92         //Detect extended feature flags
  93         if (max_basic_cap >= 7)
  94         {
  95                 MY_CPUID(&cpuInfo[0], 7);
  96                 if (features.features & FLAG_AVX)
  97                 {
  98                         features.features |= CHECK_FLAG(cpuInfo[1], 0x00000020, FLAG_AVX2);
  99                 }
 100         }
 101
 102         //Read the CPU "brand" string
 103         if (max_basic_cap > 0)
 104         {
 105                 MY_CPUID(&cpuInfo[0], 0x80000000);
 106                 const uint32_t max_extended_cap = qBound(0x80000000, cpuInfo[0], 0x80000004);
 107                 if (max_extended_cap >= 0x80000001)
 108                 {
 109                         MY_CPUID(&cpuInfo[0], 0x80000001);
 110                         features.features |= CHECK_FLAG(cpuInfo[2], 0x00000020, FLAG_LZCNT);
 111                         for (uint32_t i = 0x80000002; i <= max_extended_cap; ++i)
 112                         {
 113                                 MY_CPUID(&cpuInfo[0], i);
 114                                 memcpy(&features.brand[(i - 0x80000002) * sizeof(cpuInfo)], &cpuInfo[0], sizeof(cpuInfo));
 115                         }
 116                         features.brand[sizeof(features.brand) - 1] = '\0';
 117                 }
 118         }
 119
 120         //Detect 64-Bit processors
 121         if (OS::os_architecture() == OS::ARCH_X64)
 122         {
 123                 features.x64 = true;
 124                 features.features |= FLAGS_X64; /*x86_64 implies SSE2*/
 125         }
 126
 127         //Make sure that (at least) the MMX flag has been set!
 128         if (!(features.features & FLAG_MMX))
 129         {
 130                 qWarning("Warning: CPU does not seem to support MMX. Take care!\n");
 131                 features.features = 0;
 132         }
 133
 134         //Count the number of available(!) CPU cores
 135         DWORD_PTR procAffinity, sysAffinity;
 136         if(GetProcessAffinityMask(GetCurrentProcess(), &procAffinity, &sysAffinity))
 137         {
 138                 for(DWORD_PTR mask = 1; mask; mask <<= 1)
 139                 {
 140                         features.count += ((sysAffinity & mask) ? (1) : (0));
 141                 }
 142         }
 143         if(features.count < 1)
 144         {
 145                 GetNativeSystemInfo(&systemInfo);
 146                 features.count = qBound(1UL, systemInfo.dwNumberOfProcessors, 64UL);
 147         }
 148
 149         //Apply manual CPU overwrites
 150         bool userFlag = false;
 151         if (args.contains(QLatin1String("cpu-no-simd")))   { userFlag = true; features.features = 0U; }
 152         if (args.contains(QLatin1String("cpu-no-vendor"))) { userFlag = true; features.vendor   = 0U; }
 153         if (args.contains(QLatin1String("cpu-no-x64")))    { userFlag = true; features.x64      = 0U; }
 154         if(userFlag)
 155         {
 156                 qWarning("CPU flags overwritten by user-defined parameters. Take care!\n");
 157         }
 158
 159         return features;
 160 }