shadps4-emu/ext-cryptopp
1
0
Fork 0
mirror of https://github.com/shadps4-emu/ext-cryptopp.git synced 2024-11-27 03:40:22 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

More Clang and XLC workarounds (PR #904)

Browse Source 
I think this is the final set. Clang 3.8 is OK on Ubuntu Xenial. XLC 11.0 is OK on Power 7.

Clang 3.4 crashes on gcc110.fsffrance.org, but it looks like another compiler bug. I'm not spending time trying to work around it. Clang sucks days of your life at times...
This commit is contained in:
Jeffrey Walton 2019-10-24 16:48:40 -04:00 committed by GitHub
parent 670ceb19a2
commit c363e43968
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 23 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
TestPrograms/test_ppc_power7.cxx
View File

@ -11,10 +11,10 @@ int main(int argc, char* argv[])
__vector unsigned int b = vec_ld(0, (unsigned int*)argv[0]);
__vector unsigned int c = vec_xor(a, b);
# if defined(__VSX__)
__vector unsigned int x = {5,6,7,8};
__vector unsigned int y = vec_xl(0, (unsigned int*)argv[0]);
__vector unsigned int z = vec_xor(x, y);
# if defined(__VSX__)
__vector unsigned long long xx = {1,2};
__vector unsigned long long yy = (__vector unsigned long long)y;
# endif

2
TestPrograms/test_ppc_power8.cxx
View File

@ -8,7 +8,7 @@ int main(int argc, char* argv[])
{
#if defined(_ARCH_PWR8)
__vector unsigned long long r = {1, 2};
__vector unsigned int s = vec_xl(0, (unsigned int*)argv[0]);
__vector unsigned int s = vec_xl(0, (unsigned int*)argv[0]); // Power7
__vector unsigned long long w = (__vector unsigned long long)r;
__vector unsigned long long x = (__vector unsigned long long)s;
__vector unsigned long long y = vec_xor(w, x);

42
ppc_simd.h
View File

@ -45,9 +45,9 @@
/// slightly different implementations.
/// \since Crypto++ 6.0, LLVM Clang compiler support since Crypto++ 8.0
// Use __ALTIVEC__, _ARCH_PWR7 and _ARCH_PWR8 when detecting actual
// availaibility of the feature for the source file being compiled. The
// preprocessor macros depend on compiler options like -maltivec; and
// Use __ALTIVEC__, _ARCH_PWR7, __VSX__, and _ARCH_PWR8 when detecting
// actual availaibility of the feature for the source file being compiled.
// The preprocessor macros depend on compiler options like -maltivec; and
// not compiler versions.
// For GCC see https://gcc.gnu.org/onlinedocs/gcc/Basic-PowerPC-Built-in-Functions.html
@ -262,11 +262,11 @@ inline uint32x4_p VecLoad(const byte src[16])
#if defined(_ARCH_PWR9)
// ISA 3.0 provides vec_xl for short* and char*
return (uint32x4_p)vec_xl(off, CONST_V8_CAST(src));
#elif defined(__VSX__) && defined(__GNUC__) && !defined(__clang__)
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#elif defined(_ARCH_PWR7)
#elif defined(_ARCH_PWR7) && defined(__VSX__)
// ISA 2.06 provides vec_xl, but it lacks short* and char*
return (uint32x4_p)vec_xl(off, CONST_V32_CAST(src));
#elif defined(__VSX__) // will this ever be taken?
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#else
return (uint32x4_p)VecLoad_ALTIVEC(off, src);
#endif
@ -290,11 +290,11 @@ inline uint32x4_p VecLoad(int off, const byte src[16])
#if defined(_ARCH_PWR9)
// ISA 3.0 provides vec_xl for short* and char*
return (uint32x4_p)vec_xl(off, CONST_V8_CAST(src));
#elif defined(__VSX__) && defined(__GNUC__) && !defined(__clang__)
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#elif defined(_ARCH_PWR7)
#elif defined(_ARCH_PWR7) && defined(__VSX__)
// ISA 2.06 provides vec_xl, but it lacks short* and char*
return (uint32x4_p)vec_xl(off, CONST_V32_CAST(src));
#elif defined(__VSX__) // will this ever be taken?
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#else
return (uint32x4_p)VecLoad_ALTIVEC(off, src);
#endif
@ -393,11 +393,11 @@ inline uint32x4_p VecLoadAligned(const byte src[16])
#if defined(_ARCH_PWR9)
// ISA 3.0 provides vec_xl for short* and char*
return (uint32x4_p)vec_xl(off, CONST_V8_CAST(src));
#elif defined(__VSX__) && defined(__GNUC__) && !defined(__clang__)
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#elif defined(_ARCH_PWR7)
#elif defined(_ARCH_PWR7) && defined(__VSX__)
// ISA 2.06 provides vec_xl, but it lacks short* and char*
return (uint32x4_p)vec_xl(off, CONST_V32_CAST(src));
#elif defined(__VSX__) // will this ever be taken?
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#else
return (uint32x4_p)VecLoad_ALTIVEC(off, src);
#endif
@ -420,11 +420,11 @@ inline uint32x4_p VecLoadAligned(int off, const byte src[16])
#if defined(_ARCH_PWR9)
// ISA 3.0 provides vec_xl for short* and char*
return (uint32x4_p)vec_xl(off, CONST_V8_CAST(src));
#elif defined(__VSX__) && defined(__GNUC__) && !defined(__clang__)
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#elif defined(_ARCH_PWR7)
#elif defined(_ARCH_PWR7) && defined(__VSX__)
// ISA 2.06 provides vec_xl, but it lacks short* and char*
return (uint32x4_p)vec_xl(off, CONST_V32_CAST(src));
#elif defined(__VSX__) // will this ever be taken?
return (uint32x4_p)vec_vsx_ld(off, CONST_V32_CAST(src));
#else
return (uint32x4_p)VecLoad_ALTIVEC(off, src);
#endif
@ -582,11 +582,11 @@ inline void VecStore(const T data, byte dest[16])
#if defined(_ARCH_PWR9)
// ISA 3.0 provides vec_xl for short* and char*
vec_xst((uint8x16_p)data, off, NCONST_V8_CAST(dest));
#elif defined(__VSX__) && defined(__GNUC__) && !defined(__clang__)
vec_vsx_st((uint32x4_p)data, off, NCONST_V32_CAST(dest));
#elif defined(_ARCH_PWR7)
#elif defined(_ARCH_PWR7) && defined(__VSX__)
// ISA 2.06 provides vec_xl, but it lacks short* and char*
vec_xst((uint32x4_p)data, off, NCONST_V32_CAST(dest));
#elif defined(__VSX__) // will this ever be taken?
vec_vsx_st((uint32x4_p)data, off, NCONST_V32_CAST(dest));
#else
VecStore_ALTIVEC((uint8x16_p)data, off, NCONST_V8_CAST(dest));
#endif
@ -613,11 +613,11 @@ inline void VecStore(const T data, int off, byte dest[16])
#if defined(_ARCH_PWR9)
// ISA 3.0 provides vec_xl for short* and char*
vec_xst((uint8x16_p)data, off, NCONST_V8_CAST(dest));
#elif defined(__VSX__) && defined(__GNUC__) && !defined(__clang__)
vec_vsx_st((uint32x4_p)data, off, NCONST_V32_CAST(dest));
#elif defined(_ARCH_PWR7)
#elif defined(_ARCH_PWR7) && defined(__VSX__)
// ISA 2.06 provides vec_xl, but it lacks short* and char*
vec_xst((uint32x4_p)data, off, NCONST_V32_CAST(dest));
#elif defined(__VSX__) // will this ever be taken?
vec_vsx_st((uint32x4_p)data, off, NCONST_V32_CAST(dest));
#else
VecStore_ALTIVEC((uint8x16_p)data, off, NCONST_V8_CAST(dest));
#endif