arne
/
cryptopp
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Update comments

pull/748/head
Jeffrey Walton 2018-11-19 20:02:51 -05:00
parent f52a141f6e
commit 66b3ba5d66
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
3 changed files with 36 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
TestPrograms/test_ppc_power7.cxx
View File

@ -1,17 +1,20 @@
#include <altivec.h> #include <altivec.h>
// This follows ppc_simd.h. XLC compilers for POWER7 use vec_xlw4 and vec_xstw4. // This follows ppc_simd.h. XLC compilers for POWER7 use vec_xlw4 and
// XLC compilers for POWER8 and above use vec_xl and vec_xst. The way to tell // vec_xstw4. Some XLC compilers for POWER7 and above use vec_xl and
// the difference is, POWER7 XLC compilers are version 12.x and earlier. // vec_xst. The way to tell the difference is, XLC compilers version
// The open question is, how to handle early Clang compilers for POWER7. // 13.0 and earlier use use vec_xlw4 and vec_xstw4 XLC compilers 13.1
// We know the latest Clang compilers support vec_xl and vec_xst. // and later are use vec_xl and vec_xst. The open question is, how to
// handle early Clang compilers for POWER7. We know the latest Clang
// compilers support vec_xl and vec_xst. Also see
// https://www-01.ibm.com/support/docview.wss?uid=swg21683541
#if defined(__xlc__) && (__xlc__ < 0x0d00) #if defined(__xlc__) && (__xlc__ < 0x0d01)
# define __old_xlc__ 1 # define __early_xlc__ 1
#endif #endif
#if defined(__xlC__) && (__xlC__ < 0x0d00) #if defined(__xlC__) && (__xlC__ < 0x0d01)
# define __old_xlC__ 1 # define __early_xlC__ 1
#endif #endif
int main(int argc, char* argv[]) int main(int argc, char* argv[])
@ -19,15 +22,15 @@ int main(int argc, char* argv[])
__vector unsigned char x; __vector unsigned char x;
unsigned char res[16]; unsigned char res[16];
#if defined(__old_xlc__) || defined(__old_xlC__) #if defined(_ARCH_PWR7) && (defined(__early_xlc__) || defined(__early_xlC__))
x=vec_xlw4(0, (unsigned char*)argv[0]); x=vec_xlw4(0, (unsigned char*)argv[0]);
x=vec_add(x,x); x=vec_add(x,x);
vec_xstw4(x, 0, res); vec_xstw4(x, 0, res);
#elif defined(__xlc__) || defined(__xlC__) || defined(__clang__) #elif defined(_ARCH_PWR7) && (defined(__xlc__) || defined(__xlC__) || defined(__clang__))
x=vec_xl(0, (unsigned char*)argv[0]); x=vec_xl(0, (unsigned char*)argv[0]);
x=vec_add(x,x); x=vec_add(x,x);
vec_xst(x, 0, res); vec_xst(x, 0, res);
#elif defined(__GNUC__) #elif defined(_ARCH_PWR7) && defined(__GNUC__)
x=vec_vsx_ld(0, (unsigned char*)argv[0]); x=vec_vsx_ld(0, (unsigned char*)argv[0]);
x=vec_add(x,x); x=vec_add(x,x);
vec_vsx_st(x, 0, res); vec_vsx_st(x, 0, res);

2
ppc_power7.cpp
View File

@ -65,7 +65,7 @@ bool CPU_ProbePower7()
byte b1[19] = {255, 255, 255, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, b2[17]; byte b1[19] = {255, 255, 255, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, b2[17];
// Specifically call the VSX loads and stores // Specifically call the VSX loads and stores
#if defined(__old_xlc__) || defined(__old_xlC__) #if defined(__early_xlc__) || defined(__early_xlC__)
vec_xstw4(vec_xlw4(0, b1+3), 0, b2+1); vec_xstw4(vec_xlw4(0, b1+3), 0, b2+1);
#elif defined(__xlc__) || defined(__xlC__) || defined(__clang__) #elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
vec_xst(vec_xl(0, b1+3), 0, b2+1); vec_xst(vec_xl(0, b1+3), 0, b2+1);

39
ppc_simd.h
View File

@ -39,19 +39,20 @@
# define __CRYPTO__ 1 # define __CRYPTO__ 1
#endif #endif
// Hack to detect early XLC compilers. XLC compilers for POWER7 // Hack to detect early XLC compilers. XLC compilers for POWER7 use
// use vec_xlw4 and vec_xstw4. XLC compilers for POWER8 and above // vec_xlw4 and vec_xstw4. Some XLC compilers for POWER7 and above
// use vec_xl and vec_xst. The way to tell the difference is, // use vec_xl and vec_xst. The way to tell the difference is, XLC
// POWER7 XLC compilers are version 12.x and earlier. The open // compilers version 13.0 and earlier use vec_xlw4 and vec_xstw4.
// question is, how to handle early Clang compilers for POWER7. // XLC compilers 13.1 and later are use vec_xl and vec_xst. The open
// We know the latest Clang compilers support vec_xl and vec_xst. // question is, how to handle early Clang compilers for POWER7. We
// 0x0d00 is hex for version 0x0d (13) and 0x00 (0), or 13.0. // know the latest Clang compilers support vec_xl and vec_xst. Also
// see https://www-01.ibm.com/support/docview.wss?uid=swg21683541.
#if defined(__xlc__) && (__xlc__ < 0x0d00) #if defined(__xlc__) && (__xlc__ < 0x0d01)
# define __old_xlc__ 1 # define __early_xlc__ 1
#endif #endif
#if defined(__xlC__) && (__xlC__ < 0x0d00) #if defined(__xlC__) && (__xlC__ < 0x0d01)
# define __old_xlC__ 1 # define __early_xlC__ 1
#endif #endif
// VecLoad_ALTIVEC and VecStore_ALTIVEC are // VecLoad_ALTIVEC and VecStore_ALTIVEC are
@ -162,7 +163,7 @@ inline uint32x4_p VecLoad_ALTIVEC(int off, const byte src[16])
inline uint32x4_p VecLoad(const byte src[16]) inline uint32x4_p VecLoad(const byte src[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
return (uint32x4_p)vec_xlw4(0, (byte*)src); return (uint32x4_p)vec_xlw4(0, (byte*)src);
# elif defined(__xlc__) || defined(__xlC__) || defined(__clang__) # elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
return (uint32x4_p)vec_xl(0, (byte*)src); return (uint32x4_p)vec_xl(0, (byte*)src);
@ -189,7 +190,7 @@ inline uint32x4_p VecLoad(const byte src[16])
inline uint32x4_p VecLoad(int off, const byte src[16]) inline uint32x4_p VecLoad(int off, const byte src[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
return (uint32x4_p)vec_xlw4(off, (byte*)src); return (uint32x4_p)vec_xlw4(off, (byte*)src);
# elif defined(__xlc__) || defined(__xlC__) || defined(__clang__) # elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
return (uint32x4_p)vec_xl(off, (byte*)src); return (uint32x4_p)vec_xl(off, (byte*)src);
@ -288,7 +289,7 @@ inline uint64x2_p VecLoad(int off, const word64 src[2])
inline uint32x4_p VecLoadBE(const byte src[16]) inline uint32x4_p VecLoadBE(const byte src[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
# if (CRYPTOPP_BIG_ENDIAN) # if (CRYPTOPP_BIG_ENDIAN)
return (uint32x4_p)vec_xlw4(0, (byte*)src); return (uint32x4_p)vec_xlw4(0, (byte*)src);
# else # else
@ -328,7 +329,7 @@ inline uint32x4_p VecLoadBE(const byte src[16])
inline uint32x4_p VecLoadBE(int off, const byte src[16]) inline uint32x4_p VecLoadBE(int off, const byte src[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
# if (CRYPTOPP_BIG_ENDIAN) # if (CRYPTOPP_BIG_ENDIAN)
return (uint32x4_p)vec_xlw4(off, (byte*)src); return (uint32x4_p)vec_xlw4(off, (byte*)src);
# else # else
@ -442,7 +443,7 @@ template<class T>
inline void VecStore(const T data, byte dest[16]) inline void VecStore(const T data, byte dest[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
vec_xstw4((uint8x16_p)data, 0, (byte*)dest); vec_xstw4((uint8x16_p)data, 0, (byte*)dest);
# elif defined(__xlc__) || defined(__xlC__) || defined(__clang__) # elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
vec_xst((uint8x16_p)data, 0, (byte*)dest); vec_xst((uint8x16_p)data, 0, (byte*)dest);
@ -472,7 +473,7 @@ template<class T>
inline void VecStore(const T data, int off, byte dest[16]) inline void VecStore(const T data, int off, byte dest[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
vec_xstw4((uint8x16_p)data, 0, (byte*)dest); vec_xstw4((uint8x16_p)data, 0, (byte*)dest);
# elif defined(__xlc__) || defined(__xlC__) || defined(__clang__) # elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
vec_xst((uint8x16_p)data, off, (byte*)dest); vec_xst((uint8x16_p)data, off, (byte*)dest);
@ -586,7 +587,7 @@ template <class T>
inline void VecStoreBE(const T data, byte dest[16]) inline void VecStoreBE(const T data, byte dest[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
# if (CRYPTOPP_BIG_ENDIAN) # if (CRYPTOPP_BIG_ENDIAN)
vec_xstw4((uint8x16_p)data, 0, (byte*)dest); vec_xstw4((uint8x16_p)data, 0, (byte*)dest);
# else # else
@ -629,7 +630,7 @@ template <class T>
inline void VecStoreBE(const T data, int off, byte dest[16]) inline void VecStoreBE(const T data, int off, byte dest[16])
{ {
#if defined(_ARCH_PWR7) #if defined(_ARCH_PWR7)
# if defined(__old_xlc__) || defined(__old_xlC__) # if defined(__early_xlc__) || defined(__early_xlC__)
# if (CRYPTOPP_BIG_ENDIAN) # if (CRYPTOPP_BIG_ENDIAN)
vec_xstw4((uint8x16_p)data, off, (byte*)dest); vec_xstw4((uint8x16_p)data, off, (byte*)dest);
# else # else