• Main Page
  • Namespaces
  • Classes
  • Files
  • File List
  • File Members

cpu.h

00001 #ifndef CRYPTOPP_CPU_H
00002 #define CRYPTOPP_CPU_H
00003 
00004 #ifdef CRYPTOPP_GENERATE_X64_MASM
00005 
00006 #define CRYPTOPP_X86_ASM_AVAILABLE
00007 #define CRYPTOPP_BOOL_X64 1
00008 #define CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE 1
00009 #define NAMESPACE_END
00010 
00011 #else
00012 
00013 #include "config.h"
00014 
00015 #ifdef CRYPTOPP_MSVC6PP_OR_LATER
00016     #include <emmintrin.h>
00017 #endif
00018 
00019 NAMESPACE_BEGIN(CryptoPP)
00020 
00021 #if defined(CRYPTOPP_X86_ASM_AVAILABLE) || (_MSC_VER >= 1400 && CRYPTOPP_BOOL_X64)
00022 
00023 #define CRYPTOPP_CPUID_AVAILABLE
00024 
00025 // these should not be used directly
00026 extern CRYPTOPP_DLL bool g_x86DetectionDone;
00027 extern CRYPTOPP_DLL bool g_hasSSE2;
00028 extern CRYPTOPP_DLL bool g_hasISSE;
00029 extern CRYPTOPP_DLL bool g_hasMMX;
00030 extern CRYPTOPP_DLL bool g_hasSSSE3;
00031 extern CRYPTOPP_DLL bool g_isP4;
00032 extern CRYPTOPP_DLL word32 g_cacheLineSize;
00033 CRYPTOPP_DLL void CRYPTOPP_API DetectX86Features();
00034 
00035 CRYPTOPP_DLL bool CRYPTOPP_API CpuId(word32 input, word32 *output);
00036 
00037 #if CRYPTOPP_BOOL_X64
00038 inline bool HasSSE2()   {return true;}
00039 inline bool HasISSE()   {return true;}
00040 inline bool HasMMX()    {return true;}
00041 #else
00042 
00043 inline bool HasSSE2()
00044 {
00045     if (!g_x86DetectionDone)
00046         DetectX86Features();
00047     return g_hasSSE2;
00048 }
00049 
00050 inline bool HasISSE()
00051 {
00052     if (!g_x86DetectionDone)
00053         DetectX86Features();
00054     return g_hasISSE;
00055 }
00056 
00057 inline bool HasMMX()
00058 {
00059     if (!g_x86DetectionDone)
00060         DetectX86Features();
00061     return g_hasMMX;
00062 }
00063 
00064 #endif
00065 
00066 inline bool HasSSSE3()
00067 {
00068     if (!g_x86DetectionDone)
00069         DetectX86Features();
00070     return g_hasSSSE3;
00071 }
00072 
00073 inline bool IsP4()
00074 {
00075     if (!g_x86DetectionDone)
00076         DetectX86Features();
00077     return g_isP4;
00078 }
00079 
00080 inline int GetCacheLineSize()
00081 {
00082     if (!g_x86DetectionDone)
00083         DetectX86Features();
00084     return g_cacheLineSize;
00085 }
00086 
00087 #else
00088 
00089 inline int GetCacheLineSize()
00090 {
00091     return CRYPTOPP_L1_CACHE_LINE_SIZE;
00092 }
00093 
00094 inline bool HasSSSE3()  {return false;}
00095 inline bool IsP4()      {return false;}
00096 
00097 // assume MMX and SSE2 if intrinsics are enabled
00098 #if CRYPTOPP_BOOL_SSE2_INTRINSICS_AVAILABLE || CRYPTOPP_BOOL_X64
00099 inline bool HasSSE2()   {return true;}
00100 inline bool HasISSE()   {return true;}
00101 inline bool HasMMX()    {return true;}
00102 #else
00103 inline bool HasSSE2()   {return false;}
00104 inline bool HasISSE()   {return false;}
00105 inline bool HasMMX()    {return false;}
00106 #endif
00107 
00108 #endif      // #ifdef CRYPTOPP_X86_ASM_AVAILABLE || _MSC_VER >= 1400
00109 
00110 #endif
00111 
00112 #ifdef CRYPTOPP_GENERATE_X64_MASM
00113     #define AS1(x) x*newline*
00114     #define AS2(x, y) x, y*newline*
00115     #define AS3(x, y, z) x, y, z*newline*
00116     #define ASS(x, y, a, b, c, d) x, y, a*64+b*16+c*4+d*newline*
00117     #define ASL(x) label##x:*newline*
00118     #define ASJ(x, y, z) x label##y*newline*
00119     #define ASC(x, y) x label##y*newline*
00120     #define AS_HEX(y) 0##y##h
00121 #elif defined(__GNUC__)
00122     // define these in two steps to allow arguments to be expanded
00123     #define GNU_AS1(x) #x ";"
00124     #define GNU_AS2(x, y) #x ", " #y ";"
00125     #define GNU_AS3(x, y, z) #x ", " #y ", " #z ";"
00126     #define GNU_ASL(x) "\n" #x ":"
00127     #define GNU_ASJ(x, y, z) #x " " #y #z ";"
00128     #define AS1(x) GNU_AS1(x)
00129     #define AS2(x, y) GNU_AS2(x, y)
00130     #define AS3(x, y, z) GNU_AS3(x, y, z)
00131     #define ASS(x, y, a, b, c, d) #x ", " #y ", " #a "*64+" #b "*16+" #c "*4+" #d ";"
00132     #define ASL(x) GNU_ASL(x)
00133     #define ASJ(x, y, z) GNU_ASJ(x, y, z)
00134     #define ASC(x, y) #x " " #y ";"
00135     #define CRYPTOPP_NAKED
00136     #define AS_HEX(y) 0x##y
00137 #else
00138     #define AS1(x) __asm {x}
00139     #define AS2(x, y) __asm {x, y}
00140     #define AS3(x, y, z) __asm {x, y, z}
00141     #define ASS(x, y, a, b, c, d) __asm {x, y, _MM_SHUFFLE(a, b, c, d)}
00142     #define ASL(x) __asm {label##x:}
00143     #define ASJ(x, y, z) __asm {x label##y}
00144     #define ASC(x, y) __asm {x label##y}
00145     #define CRYPTOPP_NAKED __declspec(naked)
00146     #define AS_HEX(y) 0x##y
00147 #endif
00148 
00149 #define IF0(y)
00150 #define IF1(y) y
00151 
00152 #ifdef CRYPTOPP_GENERATE_X64_MASM
00153 #define ASM_MOD(x, y) ((x) MOD (y))
00154 #define XMMWORD_PTR XMMWORD PTR
00155 #else
00156 // GNU assembler doesn't seem to have mod operator
00157 #define ASM_MOD(x, y) ((x)-((x)/(y))*(y))
00158 // GAS 2.15 doesn't support XMMWORD PTR. it seems necessary only for MASM
00159 #define XMMWORD_PTR
00160 #endif
00161 
00162 #if CRYPTOPP_BOOL_X86
00163     #define AS_REG_1 ecx
00164     #define AS_REG_2 edx
00165     #define AS_REG_3 esi
00166     #define AS_REG_4 edi
00167     #define AS_REG_5 eax
00168     #define AS_REG_6 ebx
00169     #define AS_REG_7 ebp
00170     #define AS_REG_1d ecx
00171     #define AS_REG_2d edx
00172     #define AS_REG_3d esi
00173     #define AS_REG_4d edi
00174     #define AS_REG_5d eax
00175     #define AS_REG_6d ebx
00176     #define AS_REG_7d ebp
00177     #define WORD_SZ 4
00178     #define WORD_REG(x) e##x
00179     #define WORD_PTR DWORD PTR
00180     #define AS_PUSH_IF86(x) AS1(push e##x)
00181     #define AS_POP_IF86(x) AS1(pop e##x)
00182     #define AS_JCXZ jecxz
00183 #elif CRYPTOPP_BOOL_X64
00184     #ifdef CRYPTOPP_GENERATE_X64_MASM
00185         #define AS_REG_1 rcx
00186         #define AS_REG_2 rdx
00187         #define AS_REG_3 r8
00188         #define AS_REG_4 r9
00189         #define AS_REG_5 rax
00190         #define AS_REG_6 r10
00191         #define AS_REG_7 r11
00192         #define AS_REG_1d ecx
00193         #define AS_REG_2d edx
00194         #define AS_REG_3d r8d
00195         #define AS_REG_4d r9d
00196         #define AS_REG_5d eax
00197         #define AS_REG_6d r10d
00198         #define AS_REG_7d r11d
00199     #else
00200         #define AS_REG_1 rdi
00201         #define AS_REG_2 rsi
00202         #define AS_REG_3 rdx
00203         #define AS_REG_4 rcx
00204         #define AS_REG_5 r8
00205         #define AS_REG_6 r9
00206         #define AS_REG_7 r10
00207         #define AS_REG_1d edi
00208         #define AS_REG_2d esi
00209         #define AS_REG_3d edx
00210         #define AS_REG_4d ecx
00211         #define AS_REG_5d r8d
00212         #define AS_REG_6d r9d
00213         #define AS_REG_7d r10d
00214     #endif
00215     #define WORD_SZ 8
00216     #define WORD_REG(x) r##x
00217     #define WORD_PTR QWORD PTR
00218     #define AS_PUSH_IF86(x)
00219     #define AS_POP_IF86(x)
00220     #define AS_JCXZ jrcxz
00221 #endif
00222 
00223 // helper macro for stream cipher output
00224 #define AS_XMM_OUTPUT4(labelPrefix, inputPtr, outputPtr, x0, x1, x2, x3, t, p0, p1, p2, p3, increment)\
00225     AS2(    test    inputPtr, inputPtr)\
00226     ASC(    jz,     labelPrefix##3)\
00227     AS2(    test    inputPtr, 15)\
00228     ASC(    jnz,    labelPrefix##7)\
00229     AS2(    pxor    xmm##x0, [inputPtr+p0*16])\
00230     AS2(    pxor    xmm##x1, [inputPtr+p1*16])\
00231     AS2(    pxor    xmm##x2, [inputPtr+p2*16])\
00232     AS2(    pxor    xmm##x3, [inputPtr+p3*16])\
00233     AS2(    add     inputPtr, increment*16)\
00234     ASC(    jmp,    labelPrefix##3)\
00235     ASL(labelPrefix##7)\
00236     AS2(    movdqu  xmm##t, [inputPtr+p0*16])\
00237     AS2(    pxor    xmm##x0, xmm##t)\
00238     AS2(    movdqu  xmm##t, [inputPtr+p1*16])\
00239     AS2(    pxor    xmm##x1, xmm##t)\
00240     AS2(    movdqu  xmm##t, [inputPtr+p2*16])\
00241     AS2(    pxor    xmm##x2, xmm##t)\
00242     AS2(    movdqu  xmm##t, [inputPtr+p3*16])\
00243     AS2(    pxor    xmm##x3, xmm##t)\
00244     AS2(    add     inputPtr, increment*16)\
00245     ASL(labelPrefix##3)\
00246     AS2(    test    outputPtr, 15)\
00247     ASC(    jnz,    labelPrefix##8)\
00248     AS2(    movdqa  [outputPtr+p0*16], xmm##x0)\
00249     AS2(    movdqa  [outputPtr+p1*16], xmm##x1)\
00250     AS2(    movdqa  [outputPtr+p2*16], xmm##x2)\
00251     AS2(    movdqa  [outputPtr+p3*16], xmm##x3)\
00252     ASC(    jmp,    labelPrefix##9)\
00253     ASL(labelPrefix##8)\
00254     AS2(    movdqu  [outputPtr+p0*16], xmm##x0)\
00255     AS2(    movdqu  [outputPtr+p1*16], xmm##x1)\
00256     AS2(    movdqu  [outputPtr+p2*16], xmm##x2)\
00257     AS2(    movdqu  [outputPtr+p3*16], xmm##x3)\
00258     ASL(labelPrefix##9)\
00259     AS2(    add     outputPtr, increment*16)
00260 
00261 NAMESPACE_END
00262 
00263 #endif

Generated on Sun Oct 17 2010 02:13:04 for Crypto++ by  doxygen 1.7.1