Merge branch 'Preview4_0' into feature-vfs-security

This commit is contained in:
Robert MacGregor 2022-06-13 08:05:26 -04:00
commit 161ffc62fe
3013 changed files with 348715 additions and 182470 deletions

View file

@ -55,50 +55,11 @@
/// @note These enums must be globally scoped so that they work with the inline assembly
enum ProcessorType
{
// x86
CPU_X86Compatible,
CPU_Intel_Unknown,
CPU_Intel_486,
CPU_Intel_Pentium,
CPU_Intel_PentiumMMX,
CPU_Intel_PentiumPro,
CPU_Intel_PentiumII,
CPU_Intel_PentiumCeleron,
CPU_Intel_PentiumIII,
CPU_Intel_Pentium4,
CPU_Intel_PentiumM,
CPU_Intel_Core,
CPU_Intel_Core2,
CPU_Intel_Corei7Xeon, // Core i7 or Xeon
CPU_AMD_K6,
CPU_AMD_K6_2,
CPU_AMD_K6_3,
CPU_AMD_Athlon,
CPU_AMD_Phenom,
CPU_AMD_PhenomII,
CPU_AMD_Bulldozer,
CPU_AMD_Unknown,
CPU_Cyrix_6x86,
CPU_Cyrix_MediaGX,
CPU_Cyrix_6x86MX,
CPU_Cyrix_GXm, ///< Media GX w/ MMX
CPU_Cyrix_Unknown,
// PowerPC
CPU_PowerPC_Unknown,
CPU_PowerPC_601,
CPU_PowerPC_603,
CPU_PowerPC_603e,
CPU_PowerPC_603ev,
CPU_PowerPC_604,
CPU_PowerPC_604e,
CPU_PowerPC_604ev,
CPU_PowerPC_G3,
CPU_PowerPC_G4,
CPU_PowerPC_G4_7450,
CPU_PowerPC_G4_7455,
CPU_PowerPC_G4_7447,
CPU_PowerPC_G5,
CPU_ArmCompatible,
CPU_Intel,
CPU_AMD,
CPU_Apple
};
/// Properties for CPU.
@ -107,17 +68,17 @@ enum ProcessorProperties
CPU_PROP_C = (1<<0), ///< We should use C fallback math functions.
CPU_PROP_FPU = (1<<1), ///< Has an FPU. (It better!)
CPU_PROP_MMX = (1<<2), ///< Supports MMX instruction set extension.
CPU_PROP_3DNOW = (1<<3), ///< Supports AMD 3dNow! instruction set extension.
CPU_PROP_SSE = (1<<4), ///< Supports SSE instruction set extension.
CPU_PROP_RDTSC = (1<<5), ///< Supports Read Time Stamp Counter op.
CPU_PROP_SSE2 = (1<<6), ///< Supports SSE2 instruction set extension.
CPU_PROP_SSE3 = (1<<7), ///< Supports SSE3 instruction set extension.
CPU_PROP_SSE3xt = (1<<8), ///< Supports extended SSE3 instruction set
CPU_PROP_SSE4_1 = (1<<9), ///< Supports SSE4_1 instruction set extension.
CPU_PROP_SSE4_2 = (1<<10), ///< Supports SSE4_2 instruction set extension.
CPU_PROP_MP = (1<<11), ///< This is a multi-processor system.
CPU_PROP_LE = (1<<12), ///< This processor is LITTLE ENDIAN.
CPU_PROP_64bit = (1<<13), ///< This processor is 64-bit capable
CPU_PROP_SSE = (1<<3), ///< Supports SSE instruction set extension.
CPU_PROP_SSE2 = (1<<4), ///< Supports SSE2 instruction set extension.
CPU_PROP_SSE3 = (1<<5), ///< Supports SSE3 instruction set extension.
CPU_PROP_SSE3ex = (1<<6), ///< Supports Supplemental SSE3 instruction set
CPU_PROP_SSE4_1 = (1<<7), ///< Supports SSE4_1 instruction set extension.
CPU_PROP_SSE4_2 = (1<<8), ///< Supports SSE4_2 instruction set extension.
CPU_PROP_AVX = (1<<9), ///< Supports AVX256 instruction set extension.
CPU_PROP_MP = (1<<10), ///< This is a multi-processor system.
CPU_PROP_LE = (1<<11), ///< This processor is LITTLE ENDIAN.
CPU_PROP_64bit = (1<<12), ///< This processor is 64-bit capable
CPU_PROP_NEON = (1<<13), ///< Supports the Arm Neon instruction set extension.
};
/// Processor info manager.
@ -336,7 +297,6 @@ namespace Platform
bool isHyperThreaded;
U32 numLogicalProcessors;
U32 numPhysicalProcessors;
U32 numAvailableCores;
U32 properties; // CPU type specific enum
} processor;
};

View file

@ -28,264 +28,47 @@
Signal<void(void)> Platform::SystemInfoReady;
enum CPUFlags
{
// EDX Register flags
BIT_FPU = BIT(0),
BIT_RDTSC = BIT(4),
BIT_MMX = BIT(23),
BIT_SSE = BIT(25),
BIT_SSE2 = BIT(26),
BIT_3DNOW = BIT(31),
// These use a different value for comparison than the above flags (ECX Register)
BIT_SSE3 = BIT(0),
BIT_SSE3xt = BIT(9),
BIT_SSE4_1 = BIT(19),
BIT_SSE4_2 = BIT(20),
};
// fill the specified structure with information obtained from asm code
void SetProcessorInfo(Platform::SystemInfo_struct::Processor& pInfo,
char* vendor, U32 processor, U32 properties, U32 properties2)
void SetProcessorInfo(Platform::SystemInfo_struct::Processor& pInfo, const char* vendor, const char* brand)
{
Platform::SystemInfo.processor.properties |= (properties & BIT_FPU) ? CPU_PROP_FPU : 0;
Platform::SystemInfo.processor.properties |= (properties & BIT_RDTSC) ? CPU_PROP_RDTSC : 0;
Platform::SystemInfo.processor.properties |= (properties & BIT_MMX) ? CPU_PROP_MMX : 0;
if (dStricmp(vendor, "GenuineIntel") == 0)
{
pInfo.properties |= (properties & BIT_SSE) ? CPU_PROP_SSE : 0;
pInfo.properties |= (properties & BIT_SSE2) ? CPU_PROP_SSE2 : 0;
pInfo.properties |= (properties2 & BIT_SSE3) ? CPU_PROP_SSE3 : 0;
pInfo.properties |= (properties2 & BIT_SSE3xt) ? CPU_PROP_SSE3xt : 0;
pInfo.properties |= (properties2 & BIT_SSE4_1) ? CPU_PROP_SSE4_1 : 0;
pInfo.properties |= (properties2 & BIT_SSE4_2) ? CPU_PROP_SSE4_2 : 0;
pInfo.type = CPU_Intel_Unknown;
// switch on processor family code
switch ((processor >> 8) & 0x0f)
{
case 4:
pInfo.type = CPU_Intel_486;
pInfo.name = StringTable->insert("Intel 486 class");
break;
// Pentium Family
case 5:
// switch on processor model code
switch ((processor >> 4) & 0xf)
{
case 1:
case 2:
case 3:
pInfo.type = CPU_Intel_Pentium;
pInfo.name = StringTable->insert("Intel Pentium");
break;
case 4:
pInfo.type = CPU_Intel_PentiumMMX;
pInfo.name = StringTable->insert("Intel Pentium MMX");
break;
default:
pInfo.type = CPU_Intel_Pentium;
pInfo.name = StringTable->insert( "Intel (unknown)" );
break;
}
break;
// Pentium Pro/II/II family
case 6:
{
U32 extendedModel = ( processor & 0xf0000 ) >> 16;
// switch on processor model code
switch ((processor >> 4) & 0xf)
{
case 1:
pInfo.type = CPU_Intel_PentiumPro;
pInfo.name = StringTable->insert("Intel Pentium Pro");
break;
case 3:
case 5:
pInfo.type = CPU_Intel_PentiumII;
pInfo.name = StringTable->insert("Intel Pentium II");
break;
case 6:
pInfo.type = CPU_Intel_PentiumCeleron;
pInfo.name = StringTable->insert("Intel Pentium Celeron");
break;
case 7:
case 8:
case 11:
pInfo.type = CPU_Intel_PentiumIII;
pInfo.name = StringTable->insert("Intel Pentium III");
break;
case 0xA:
if( extendedModel == 1)
{
pInfo.type = CPU_Intel_Corei7Xeon;
pInfo.name = StringTable->insert( "Intel Core i7 / Xeon" );
}
else
{
pInfo.type = CPU_Intel_PentiumIII;
pInfo.name = StringTable->insert( "Intel Pentium III Xeon" );
}
break;
case 0xD:
if( extendedModel == 1 )
{
pInfo.type = CPU_Intel_Corei7Xeon;
pInfo.name = StringTable->insert( "Intel Core i7 / Xeon" );
}
else
{
pInfo.type = CPU_Intel_PentiumM;
pInfo.name = StringTable->insert( "Intel Pentium/Celeron M" );
}
break;
case 0xE:
pInfo.type = CPU_Intel_Core;
pInfo.name = StringTable->insert( "Intel Core" );
break;
case 0xF:
pInfo.type = CPU_Intel_Core2;
pInfo.name = StringTable->insert( "Intel Core 2" );
break;
default:
pInfo.type = CPU_Intel_PentiumPro;
pInfo.name = StringTable->insert( "Intel (unknown)" );
break;
}
break;
}
// Pentium4 Family
case 0xf:
pInfo.type = CPU_Intel_Pentium4;
pInfo.name = StringTable->insert( "Intel Pentium 4" );
break;
default:
pInfo.type = CPU_Intel_Unknown;
pInfo.name = StringTable->insert( "Intel (unknown)" );
break;
}
pInfo.type = CPU_Intel;
pInfo.name = StringTable->insert(brand ? brand : "Intel (Unknown)");
}
//--------------------------------------
else if (dStricmp(vendor, "AuthenticAMD") == 0)
{
pInfo.name = StringTable->insert(brand ? brand : "AMD (unknown)");
pInfo.type = CPU_AMD;
}
else if (dStricmp(vendor, "Apple") == 0)
{
pInfo.name = StringTable->insert(brand ? brand : "Apple (unknown)");
pInfo.type = CPU_Apple;
}
else
if (dStricmp(vendor, "AuthenticAMD") == 0)
{
// AthlonXP processors support SSE
pInfo.properties |= (properties & BIT_SSE) ? CPU_PROP_SSE : 0;
pInfo.properties |= ( properties & BIT_SSE2 ) ? CPU_PROP_SSE2 : 0;
pInfo.properties |= (properties & BIT_3DNOW) ? CPU_PROP_3DNOW : 0;
// Phenom and PhenomII support SSE3, SSE4a
pInfo.properties |= ( properties2 & BIT_SSE3 ) ? CPU_PROP_SSE3 : 0;
pInfo.properties |= ( properties2 & BIT_SSE4_1 ) ? CPU_PROP_SSE4_1 : 0;
// switch on processor family code
switch ((processor >> 8) & 0xf)
{
// K6 Family
case 5:
// switch on processor model code
switch ((processor >> 4) & 0xf)
{
case 0:
case 1:
case 2:
case 3:
pInfo.type = CPU_AMD_K6_3;
pInfo.name = StringTable->insert("AMD K5");
break;
case 4:
case 5:
case 6:
case 7:
pInfo.type = CPU_AMD_K6;
pInfo.name = StringTable->insert("AMD K6");
break;
case 8:
pInfo.type = CPU_AMD_K6_2;
pInfo.name = StringTable->insert("AMD K6-2");
break;
case 9:
case 10:
case 11:
case 12:
case 13:
case 14:
case 15:
pInfo.type = CPU_AMD_K6_3;
pInfo.name = StringTable->insert("AMD K6-3");
break;
}
break;
// Athlon Family
case 6:
pInfo.type = CPU_AMD_Athlon;
pInfo.name = StringTable->insert("AMD Athlon");
break;
// Phenom Family
case 15:
pInfo.type = CPU_AMD_Phenom;
pInfo.name = StringTable->insert("AMD Phenom");
break;
// Phenom II Family
case 16:
pInfo.type = CPU_AMD_PhenomII;
pInfo.name = StringTable->insert("AMD Phenom II");
break;
// Bulldozer Family
case 17:
pInfo.type = CPU_AMD_Bulldozer;
pInfo.name = StringTable->insert("AMD Bulldozer");
break;
default:
pInfo.type = CPU_AMD_Unknown;
pInfo.name = StringTable->insert("AMD (unknown)");
break;
}
}
//--------------------------------------
else
if (dStricmp(vendor, "CyrixInstead") == 0)
{
switch (processor)
{
case 0x520:
pInfo.type = CPU_Cyrix_6x86;
pInfo.name = StringTable->insert("Cyrix 6x86");
break;
case 0x440:
pInfo.type = CPU_Cyrix_MediaGX;
pInfo.name = StringTable->insert("Cyrix Media GX");
break;
case 0x600:
pInfo.type = CPU_Cyrix_6x86MX;
pInfo.name = StringTable->insert("Cyrix 6x86mx/MII");
break;
case 0x540:
pInfo.type = CPU_Cyrix_GXm;
pInfo.name = StringTable->insert("Cyrix GXm");
break;
default:
pInfo.type = CPU_Cyrix_Unknown;
pInfo.name = StringTable->insert("Cyrix (unknown)");
break;
}
}
{
#if defined(TORQUE_CPU_X86) || defined(TORQUE_CPU_X64)
pInfo.name = StringTable->insert(brand ? brand : "x86 Compatible (unknown)");
pInfo.type = CPU_X86Compatible;
#elif defined(TORQUE_CPU_ARM64)
pInfo.name = StringTable->insert(brand ? brand : "Arm Compatible (unknown)");
pInfo.type = CPU_ArmCompatible;
#else
#error "Unknown CPU Architecture"
#endif
}
// Get multithreading caps.
CPUInfo::EConfig config = CPUInfo::CPUCount( pInfo.numLogicalProcessors, pInfo.numAvailableCores, pInfo.numPhysicalProcessors );
CPUInfo::EConfig config = CPUInfo::CPUCount( pInfo.numLogicalProcessors, pInfo.numPhysicalProcessors );
pInfo.isHyperThreaded = CPUInfo::isHyperThreaded( config );
pInfo.isMultiCore = CPUInfo::isMultiCore( config );
// Trigger the signal
Platform::SystemInfoReady.trigger();
}
}

View file

@ -1,657 +0,0 @@
// Original code is:
// Copyright (c) 2005 Intel Corporation
// All Rights Reserved
//
// CPUCount.cpp : Detects three forms of hardware multi-threading support across IA-32 platform
// The three forms of HW multithreading are: Multi-processor, Multi-core, and
// HyperThreading Technology.
// This application enumerates all the logical processors enabled by OS and BIOS,
// determine the HW topology of these enabled logical processors in the system
// using information provided by CPUID instruction.
// A multi-processing system can support any combination of the three forms of HW
// multi-threading support. The relevant topology can be identified using a
// three level decomposition of the "initial APIC ID" into
// Package_id, core_id, and SMT_id. Such decomposition provides a three-level map of
// the topology of hardware resources and
// allow multi-threaded software to manage shared hardware resources in
// the platform to reduce resource contention
// Multicore detection algorithm for processor and cache topology requires
// all leaf functions of CPUID instructions be available. System administrator
// must ensure BIOS settings is not configured to restrict CPUID functionalities.
//-------------------------------------------------------------------------------------------------
#if defined(TORQUE_OS_LINUX) || defined(LINUX)
// TODO GCC code don't compile on Release with optimizations, mover code to platform layer
#else
#include "platform/platform.h"
#include "platform/platformCPUCount.h"
#if defined(TORQUE_OS_LINUX) || defined(TORQUE_OS_OSX)
#ifdef TORQUE_OS_LINUX
// The Linux source code listing can be compiled using Linux kernel verison 2.6
// or higher (e.g. RH 4AS-2.8 using GCC 3.4.4).
// Due to syntax variances of Linux affinity APIs with earlier kernel versions
// and dependence on glibc library versions, compilation on Linux environment
// with older kernels and compilers may require kernel patches or compiler upgrades.
#include <stdlib.h>
#include <unistd.h>
#include <string.h>
#include <sched.h>
#define DWORD unsigned long
#elif defined( TORQUE_OS_WIN )
#include <windows.h>
#elif defined( TORQUE_OS_MAC )
# include <sys/types.h>
# include <sys/sysctl.h>
#else
#error Not implemented on platform.
#endif
#include <stdio.h>
#include <assert.h>
namespace CPUInfo {
#define HWD_MT_BIT 0x10000000 // EDX[28] Bit 28 is set if HT or multi-core is supported
#define NUM_LOGICAL_BITS 0x00FF0000 // EBX[23:16] Bit 16-23 in ebx contains the number of logical
// processors per physical processor when execute cpuid with
// eax set to 1
#define NUM_CORE_BITS 0xFC000000 // EAX[31:26] Bit 26-31 in eax contains the number of cores minus one
// per physical processor when execute cpuid with
// eax set to 4.
#define INITIAL_APIC_ID_BITS 0xFF000000 // EBX[31:24] Bits 24-31 (8 bits) return the 8-bit unique
// initial APIC ID for the processor this code is running on.
#ifndef TORQUE_OS_MAC
static U32 CpuIDSupported(void);
static U32 find_maskwidth(unsigned int);
static U32 HWD_MTSupported(void);
static U32 MaxLogicalProcPerPhysicalProc(void);
static U32 MaxCorePerPhysicalProc(void);
static U8 GetAPIC_ID(void);
static U8 GetNzbSubID(U8, U8, U8);
#endif
static char g_s3Levels[2048];
#ifndef TORQUE_OS_MAC
//
// CpuIDSupported will return 0 if CPUID instruction is unavailable. Otherwise, it will return
// the maximum supported standard function.
//
static U32 CpuIDSupported(void)
{
U32 maxInputValue = 0;
// If CPUID instruction is supported
#ifdef TORQUE_COMPILER_GCC
try
{
// call cpuid with eax = 0
asm
(
"pushl %%ebx\n\t"
"xorl %%eax,%%eax\n\t"
"cpuid\n\t"
"popl %%ebx\n\t"
: "=a" (maxInputValue)
:
: "%ecx", "%edx"
);
}
catch (...)
{
return(0); // cpuid instruction is unavailable
}
#elif defined( TORQUE_COMPILER_VISUALC )
try
{
// call cpuid with eax = 0
__asm
{
xor eax, eax
cpuid
mov maxInputValue, eax
}
}
catch (...)
{
// cpuid instruction is unavailable
}
#else
# error Not implemented.
#endif
return maxInputValue;
}
//
// Function returns the maximum cores per physical package. Note that the number of
// AVAILABLE cores per physical to be used by an application might be less than this
// maximum value.
//
static U32 MaxCorePerPhysicalProc(void)
{
U32 Regeax = 0;
if (!HWD_MTSupported()) return (U32) 1; // Single core
#ifdef TORQUE_COMPILER_GCC
{
asm
(
"pushl %ebx\n\t"
"xorl %eax, %eax\n\t"
"cpuid\n\t"
"cmpl $4, %eax\n\t" // check if cpuid supports leaf 4
"jl .single_core\n\t" // Single core
"movl $4, %eax\n\t"
"movl $0, %ecx\n\t" // start with index = 0; Leaf 4 reports
"popl %ebx\n\t"
); // at least one valid cache level
asm
(
"cpuid"
: "=a" (Regeax)
:
: "%ecx", "%edx"
);
asm
(
"jmp .multi_core\n"
".single_core:\n\t"
"xor %eax, %eax\n"
".multi_core:"
);
}
#elif defined( TORQUE_COMPILER_VISUALC )
__asm
{
xor eax, eax
cpuid
cmp eax, 4 // check if cpuid supports leaf 4
jl single_core // Single core
mov eax, 4
mov ecx, 0 // start with index = 0; Leaf 4 reports
cpuid // at least one valid cache level
mov Regeax, eax
jmp multi_core
single_core:
xor eax, eax
multi_core:
}
#else
# error Not implemented.
#endif
return (U32)((Regeax & NUM_CORE_BITS) >> 26)+1;
}
//
// The function returns 0 when the hardware multi-threaded bit is not set.
//
static U32 HWD_MTSupported(void)
{
U32 Regedx = 0;
if ((CpuIDSupported() >= 1))
{
#ifdef TORQUE_COMPILER_GCC
asm
(
"pushl %%ebx\n\t"
"movl $1,%%eax\n\t"
"cpuid\n\t"
"popl %%ebx\n\t"
: "=d" (Regedx)
:
: "%eax","%ecx"
);
#elif defined( TORQUE_COMPILER_VISUALC )
__asm
{
mov eax, 1
cpuid
mov Regedx, edx
}
#else
# error Not implemented.
#endif
}
return (Regedx & HWD_MT_BIT);
}
//
// Function returns the maximum logical processors per physical package. Note that the number of
// AVAILABLE logical processors per physical to be used by an application might be less than this
// maximum value.
//
static U32 MaxLogicalProcPerPhysicalProc(void)
{
U32 Regebx = 0;
if (!HWD_MTSupported()) return (U32) 1;
#ifdef TORQUE_COMPILER_GCC
asm
(
"movl $1,%%eax\n\t"
"cpuid"
: "=b" (Regebx)
:
: "%eax","%ecx","%edx"
);
#elif defined( TORQUE_COMPILER_VISUALC )
__asm
{
mov eax, 1
cpuid
mov Regebx, ebx
}
#else
# error Not implemented.
#endif
return (unsigned int) ((Regebx & NUM_LOGICAL_BITS) >> 16);
}
static U8 GetAPIC_ID(void)
{
U32 Regebx = 0;
#ifdef TORQUE_COMPILER_GCC
asm
(
"movl $1, %%eax\n\t"
"cpuid"
: "=b" (Regebx)
:
: "%eax","%ecx","%edx"
);
#elif defined( TORQUE_COMPILER_VISUALC )
__asm
{
mov eax, 1
cpuid
mov Regebx, ebx
}
#else
# error Not implemented.
#endif
return (unsigned char) ((Regebx & INITIAL_APIC_ID_BITS) >> 24);
}
//
// Determine the width of the bit field that can represent the value count_item.
//
U32 find_maskwidth(U32 CountItem)
{
U32 MaskWidth,
count = CountItem;
#ifdef TORQUE_COMPILER_GCC
asm
(
#ifdef __x86_64__ // define constant to compile
"push %%rcx\n\t" // under 64-bit Linux
"push %%rax\n\t"
#else
"pushl %%ecx\n\t"
"pushl %%eax\n\t"
#endif
// "movl $count, %%eax\n\t" //done by Assembler below
"xorl %%ecx, %%ecx"
// "movl %%ecx, MaskWidth\n\t" //done by Assembler below
: "=c" (MaskWidth)
: "a" (count)
// : "%ecx", "%eax" We don't list these as clobbered because we don't want the assembler
//to put them back when we are done
);
asm
(
"decl %%eax\n\t"
"bsrw %%ax,%%cx\n\t"
"jz next\n\t"
"incw %%cx\n\t"
// "movl %%ecx, MaskWidth\n" //done by Assembler below
: "=c" (MaskWidth)
:
);
asm
(
"next:\n\t"
#ifdef __x86_64__
"pop %rax\n\t"
"pop %rcx"
#else
"popl %eax\n\t"
"popl %ecx"
#endif
);
#elif defined( TORQUE_COMPILER_VISUALC )
__asm
{
mov eax, count
mov ecx, 0
mov MaskWidth, ecx
dec eax
bsr cx, ax
jz next
inc cx
mov MaskWidth, ecx
next:
}
#else
# error Not implemented.
#endif
return MaskWidth;
}
//
// Extract the subset of bit field from the 8-bit value FullID. It returns the 8-bit sub ID value
//
static U8 GetNzbSubID(U8 FullID,
U8 MaxSubIDValue,
U8 ShiftCount)
{
U32 MaskWidth;
U8 MaskBits;
MaskWidth = find_maskwidth((U32) MaxSubIDValue);
MaskBits = (0xff << ShiftCount) ^
((U8) (0xff << (ShiftCount + MaskWidth)));
return (FullID & MaskBits);
}
#endif
//
//
//
EConfig CPUCount(U32& TotAvailLogical, U32& TotAvailCore, U32& PhysicalNum)
{
EConfig StatusFlag = CONFIG_UserConfigIssue;
g_s3Levels[0] = 0;
TotAvailCore = 1;
PhysicalNum = 1;
U32 numLPEnabled = 0;
S32 MaxLPPerCore = 1;
#ifdef TORQUE_OS_MAC
//FIXME: This isn't a proper port but more or less just some sneaky cheating
// to get around having to mess with yet another crap UNIX-style API. Seems
// like there isn't a way to do this that's working across all OSX incarnations
// and machine configurations anyway.
S32 numCPUs;
S32 numPackages;
// Get the number of CPUs.
size_t len = sizeof( numCPUs );
if( sysctlbyname( "hw.ncpu", &numCPUs, &len, 0, 0 ) == -1 )
return CONFIG_UserConfigIssue;
// Get the number of packages.
len = sizeof( numPackages );
if( sysctlbyname( "hw.packages", &numPackages, &len, 0, 0 ) == -1 )
return CONFIG_UserConfigIssue;
TotAvailCore = numCPUs;
TotAvailLogical = numCPUs;
PhysicalNum = numPackages;
#else
U32 dwAffinityMask;
S32 j = 0;
U8 apicID, PackageIDMask;
U8 tblPkgID[256], tblCoreID[256], tblSMTID[256];
char tmp[256];
#ifdef TORQUE_OS_LINUX
//we need to make sure that this process is allowed to run on
//all of the logical processors that the OS itself can run on.
//A process could acquire/inherit affinity settings that restricts the
// current process to run on a subset of all logical processor visible to OS.
// Linux doesn't easily allow us to look at the Affinity Bitmask directly,
// but it does provide an API to test affinity maskbits of the current process
// against each logical processor visible under OS.
S32 sysNumProcs = sysconf(_SC_NPROCESSORS_CONF); //This will tell us how many
//CPUs are currently enabled.
//this will tell us which processors this process can run on.
cpu_set_t allowedCPUs;
sched_getaffinity(0, sizeof(allowedCPUs), &allowedCPUs);
for (S32 i = 0; i < sysNumProcs; i++ )
{
if ( CPU_ISSET(i, &allowedCPUs) == 0 )
return CONFIG_UserConfigIssue;
}
#elif defined( TORQUE_OS_WIN )
DWORD dwProcessAffinity, dwSystemAffinity;
GetProcessAffinityMask(GetCurrentProcess(),
&dwProcessAffinity,
&dwSystemAffinity);
if (dwProcessAffinity != dwSystemAffinity) // not all CPUs are enabled
return CONFIG_UserConfigIssue;
#else
# error Not implemented.
#endif
// Assume that cores within a package have the SAME number of
// logical processors. Also, values returned by
// MaxLogicalProcPerPhysicalProc and MaxCorePerPhysicalProc do not have
// to be power of 2.
MaxLPPerCore = MaxLogicalProcPerPhysicalProc() / MaxCorePerPhysicalProc();
dwAffinityMask = 1;
#ifdef TORQUE_OS_LINUX
cpu_set_t currentCPU;
while ( j < sysNumProcs )
{
CPU_ZERO(&currentCPU);
CPU_SET(j, &currentCPU);
if ( sched_setaffinity (0, sizeof(currentCPU), &currentCPU) == 0 )
{
sleep(0); // Ensure system to switch to the right CPU
#elif defined( TORQUE_OS_WIN )
while (dwAffinityMask && dwAffinityMask <= dwSystemAffinity)
{
if (SetThreadAffinityMask(GetCurrentThread(), dwAffinityMask))
{
Sleep(0); // Ensure system to switch to the right CPU
#else
# error Not implemented.
#endif
apicID = GetAPIC_ID();
// Store SMT ID and core ID of each logical processor
// Shift vlaue for SMT ID is 0
// Shift value for core ID is the mask width for maximum logical
// processors per core
tblSMTID[j] = GetNzbSubID(apicID, MaxLPPerCore, 0);
U8 maxCorePPP = MaxCorePerPhysicalProc();
U8 maskWidth = find_maskwidth(MaxLPPerCore);
tblCoreID[j] = GetNzbSubID(apicID, maxCorePPP, maskWidth);
// Extract package ID, assume single cluster.
// Shift value is the mask width for max Logical per package
PackageIDMask = (unsigned char) (0xff <<
find_maskwidth(MaxLogicalProcPerPhysicalProc()));
tblPkgID[j] = apicID & PackageIDMask;
sprintf(tmp," AffinityMask = %d; Initial APIC = %d; Physical ID = %d, Core ID = %d, SMT ID = %d\n",
dwAffinityMask, apicID, tblPkgID[j], tblCoreID[j], tblSMTID[j]);
dStrcat(g_s3Levels, tmp, 2048);
numLPEnabled ++; // Number of available logical processors in the system.
} // if
j++;
dwAffinityMask = 1 << j;
} // while
// restore the affinity setting to its original state
#ifdef TORQUE_OS_LINUX
sched_setaffinity (0, sizeof(allowedCPUs), &allowedCPUs);
sleep(0);
#elif defined( TORQUE_OS_WIN )
SetThreadAffinityMask(GetCurrentThread(), dwProcessAffinity);
Sleep(0);
#else
# error Not implemented.
#endif
TotAvailLogical = numLPEnabled;
//
// Count available cores (TotAvailCore) in the system
//
U8 CoreIDBucket[256];
DWORD ProcessorMask, pCoreMask[256];
U32 i, ProcessorNum;
CoreIDBucket[0] = tblPkgID[0] | tblCoreID[0];
ProcessorMask = 1;
pCoreMask[0] = ProcessorMask;
for (ProcessorNum = 1; ProcessorNum < numLPEnabled; ProcessorNum++)
{
ProcessorMask <<= 1;
for (i = 0; i < TotAvailCore; i++)
{
// Comparing bit-fields of logical processors residing in different packages
// Assuming the bit-masks are the same on all processors in the system.
if ((tblPkgID[ProcessorNum] | tblCoreID[ProcessorNum]) == CoreIDBucket[i])
{
pCoreMask[i] |= ProcessorMask;
break;
}
} // for i
if (i == TotAvailCore) // did not match any bucket. Start a new one.
{
CoreIDBucket[i] = tblPkgID[ProcessorNum] | tblCoreID[ProcessorNum];
pCoreMask[i] = ProcessorMask;
TotAvailCore++; // Number of available cores in the system
}
} // for ProcessorNum
//
// Count physical processor (PhysicalNum) in the system
//
U8 PackageIDBucket[256];
DWORD pPackageMask[256];
PackageIDBucket[0] = tblPkgID[0];
ProcessorMask = 1;
pPackageMask[0] = ProcessorMask;
for (ProcessorNum = 1; ProcessorNum < numLPEnabled; ProcessorNum++)
{
ProcessorMask <<= 1;
for (i = 0; i < PhysicalNum; i++)
{
// Comparing bit-fields of logical processors residing in different packages
// Assuming the bit-masks are the same on all processors in the system.
if (tblPkgID[ProcessorNum]== PackageIDBucket[i])
{
pPackageMask[i] |= ProcessorMask;
break;
}
} // for i
if (i == PhysicalNum) // did not match any bucket. Start a new one.
{
PackageIDBucket[i] = tblPkgID[ProcessorNum];
pPackageMask[i] = ProcessorMask;
PhysicalNum++; // Total number of physical processors in the system
}
} // for ProcessorNum
#endif
//
// Check to see if the system is multi-core
// Check if the system is hyper-threading
//
if (TotAvailCore > PhysicalNum)
{
// Multi-core
if (MaxLPPerCore == 1)
StatusFlag = CONFIG_MultiCoreAndHTNotCapable;
else if (numLPEnabled > TotAvailCore)
StatusFlag = CONFIG_MultiCoreAndHTEnabled;
else StatusFlag = CONFIG_MultiCoreAndHTDisabled;
}
else
{
// Single-core
if (MaxLPPerCore == 1)
StatusFlag = CONFIG_SingleCoreAndHTNotCapable;
else if (numLPEnabled > TotAvailCore)
StatusFlag = CONFIG_SingleCoreHTEnabled;
else StatusFlag = CONFIG_SingleCoreHTDisabled;
}
return StatusFlag;
}
} // namespace CPUInfo
#endif
#endif

View file

@ -29,13 +29,10 @@ namespace CPUInfo
{
enum EConfig
{
CONFIG_UserConfigIssue,
CONFIG_SingleCoreHTEnabled,
CONFIG_SingleCoreHTDisabled,
CONFIG_SingleCoreAndHTNotCapable,
CONFIG_MultiCoreAndHTNotCapable,
CONFIG_MultiCoreAndHTEnabled,
CONFIG_MultiCoreAndHTDisabled,
};
inline bool isMultiCore( EConfig config )
@ -44,7 +41,6 @@ namespace CPUInfo
{
case CONFIG_MultiCoreAndHTNotCapable:
case CONFIG_MultiCoreAndHTEnabled:
case CONFIG_MultiCoreAndHTDisabled:
return true;
default:
@ -65,11 +61,10 @@ namespace CPUInfo
}
}
EConfig CPUCount( U32& totalAvailableLogical,
U32& totalAvailableCores,
U32& numPhysical );
EConfig CPUCount( U32& totalAvailableLogical, U32& totalAvailableCores );
} // namespace CPUInfo
void SetProcessorInfo(Platform::SystemInfo_struct::Processor& pInfo, const char* vendor, const char* brand);
#endif // _TORQUE_PLATFORM_PLATFORMCOUNT_H_

View file

@ -1,128 +0,0 @@
;-----------------------------------------------------------------------------
; Copyright (c) 2012 GarageGames, LLC
;
; Permission is hereby granted, free of charge, to any person obtaining a copy
; of this software and associated documentation files (the "Software"), to
; deal in the Software without restriction, including without limitation the
; rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
; sell copies of the Software, and to permit persons to whom the Software is
; furnished to do so, subject to the following conditions:
;
; The above copyright notice and this permission notice shall be included in
; all copies or substantial portions of the Software.
;
; THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
; IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
; FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
; AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
; LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
; FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
; IN THE SOFTWARE.
;-----------------------------------------------------------------------------
segment .text
; syntax: export_fn <function name>
%macro export_fn 1
%ifidn __OUTPUT_FORMAT__, elf
; No underscore needed for ELF object files
global %1
%1:
%else
global _%1
_%1:
%endif
%endmacro
; push registers
%macro pushreg 0
; pushad
push ebx
push ebp
push esi
push edi
%endmacro
; pop registers
%macro popreg 0
pop edi
pop esi
pop ebp
pop ebx
; popad
%endmacro
; void detectX86CPUInfo(char *vendor, U32 *processor, U32 *properties);
export_fn detectX86CPUInfo
push ebp
mov ebp, esp
pushreg
push edx
push ecx
pushfd
pushfd ; save EFLAGS to stack
pop eax ; move EFLAGS into EAX
mov ebx, eax
xor eax, 0x200000 ; flip bit 21
push eax
popfd ; restore EFLAGS
pushfd
pop eax
cmp eax, ebx
jz EXIT ; doesn't support CPUID instruction
;
; get vendor information using CPUID eax == 0
xor eax, eax
cpuid
; store the vendor tag (12 bytes in ebx, edx, ecx) in the first parameter,
; which should be a char[13]
push eax ; save eax
mov eax, [ebp+8] ; store the char* address in eax
mov [eax], ebx ; move ebx into the first 4 bytes
add eax, 4 ; advance the char* 4 bytes
mov [eax], edx ; move edx into the next 4 bytes
add eax, 4 ; advance the char* 4 bytes
mov [eax], ecx ; move ecx into the last 4 bytes
pop eax ; restore eax
; get generic extended CPUID info
mov eax, 1
cpuid ; eax=1, so cpuid queries feature information
and eax, 0x0fff3fff
push ecx
mov ecx, [ebp+12]
mov [ecx], eax ; just store the model bits in processor param
mov ecx, [ebp+16]
mov [ecx], edx ; set properties param
pop ecx
; want to check for 3DNow(tm).
; need to see if extended cpuid functions present.
mov eax, 0x80000000
cpuid
cmp eax, 0x80000000
jbe MAYBE_3DLATER
mov eax, 0x80000001
cpuid
; 3DNow if bit 31 set -> put bit in our properties
and edx, 0x80000000
push eax
mov eax, [ebp+16]
or [eax], edx
pop eax
MAYBE_3DLATER:
EXIT:
popfd
pop ecx
pop edx
popreg
pop ebp
ret

View file

@ -322,10 +322,9 @@ ThreadPool::ThreadPool( const char* name, U32 numThreads )
// Platform::SystemInfo will not yet have been initialized.
U32 numLogical = 0;
U32 numPhysical = 0;
U32 numCores = 0;
CPUInfo::CPUCount( numLogical, numCores, numPhysical );
CPUInfo::CPUCount( numLogical, numCores );
const U32 baseCount = getMax( numLogical, numCores );
mNumThreads = (baseCount > 0) ? baseCount : 2;