Use PowerPC unaligned loads and stores with Power8 (GH #825, PR #826)

Use PowerPC unaligned loads and stores with Power8. Formerly we were using Power7 as the floor because the IBM POWER Architecture manuals said unaligned loads and stores were available. However, some compilers generate bad code for unaligned loads and stores using `-march=power7`, so bump to a known good.

GNUmakefile

@@ -10,8 +10,6 @@ SHELL = /bin/sh
 
 # If needed
 TMPDIR ?= /tmp
-# Used for ARMv7 and NEON.
-FP_ABI ?= hard
 # Used for feature tests
 TOUT ?= a.out
 TOUT := $(strip $(TOUT))
@@ -443,26 +441,26 @@ ifeq ($(DETECT_FEATURES),1)
 ifneq ($(IS_ARM32),0)
 
   TPROG = TestPrograms/test_arm_neon.cxx
-  TOPT = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+  TOPT = -march=armv7-a -mfpu=neon
   HAVE_OPT = $(shell $(CXX) $(TCXXFLAGS) $(ZOPT) $(TOPT) $(TPROG) -o $(TOUT) 2>&1 | tr ' ' '\n' | wc -l)
   ifeq ($(strip $(HAVE_OPT)),0)
-    NEON_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    NEON_FLAG = -march=armv7-a -mfpu=neon
-    ARIA_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    ARIA_FLAG = -march=armv7-a -mfpu=neon
-    AES_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    AES_FLAG = -march=armv7-a -mfpu=neon
-    CRC_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    CRC_FLAG = -march=armv7-a -mfpu=neon
-    GCM_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    GCM_FLAG = -march=armv7-a -mfpu=neon
-    BLAKE2B_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    BLAKE2B_FLAG = -march=armv7-a -mfpu=neon
-    BLAKE2S_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    BLAKE2S_FLAG = -march=armv7-a -mfpu=neon
-    CHACHA_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    CHACHA_FLAG = -march=armv7-a -mfpu=neon
-    CHAM_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    CHAM_FLAG = -march=armv7-a -mfpu=neon
-    LEA_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    LEA_FLAG = -march=armv7-a -mfpu=neon
-    SHA_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SHA_FLAG = -march=armv7-a -mfpu=neon
-    SIMECK_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SIMECK_FLAG = -march=armv7-a -mfpu=neon
-    SIMON64_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SIMON64_FLAG = -march=armv7-a -mfpu=neon
-    SIMON128_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SIMON128_FLAG = -march=armv7-a -mfpu=neon
-    SPECK64_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SPECK64_FLAG = -march=armv7-a -mfpu=neon
-    SPECK128_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SPECK128_FLAG = -march=armv7-a -mfpu=neon
-    SM4_FLAG = -march=armv7-a -mfloat-abi=$(FP_ABI) -mfpu=neon
+    SM4_FLAG = -march=armv7-a -mfpu=neon
   else
     CXXFLAGS += -DCRYPTOPP_DISABLE_ASM
   endif
@@ -620,13 +618,21 @@ ifeq ($(DETECT_FEATURES),1)
   TOPT = $(POWER8_FLAG)
   HAVE_OPT = $(shell $(CXX) $(TCXXFLAGS) $(ZOPT) $(TOPT) $(TPROG) -o $(TOUT) 2>&1 | tr ' ' '\n' | wc -l)
   ifeq ($(strip $(HAVE_OPT)),0)
+    AES_FLAG = $(POWER8_FLAG)
+    ARIA_FLAG = $(POWER8_FLAG)
     BLAKE2B_FLAG = $(POWER8_FLAG)
+    BLAKE2S_FLAG = $(POWER8_FLAG)
+    CHACHA_FLAG = $(POWER8_FLAG)
+    CHAM_FLAG = $(POWER8_FLAG)
     CRC_FLAG = $(POWER8_FLAG)
     GCM_FLAG = $(POWER8_FLAG)
     GF2N_FLAG = $(POWER8_FLAG)
-    AES_FLAG = $(POWER8_FLAG)
+    LEA_FLAG = $(POWER8_FLAG)
     SHA_FLAG = $(POWER8_FLAG)
     SHACAL2_FLAG = $(POWER8_FLAG)
+    SIMECK_FLAG = $(POWER8_FLAG)
+    SIMON64_FLAG = $(POWER8_FLAG)
+    SPECK64_FLAG = $(POWER8_FLAG)
     SIMON128_FLAG = $(POWER8_FLAG)
     SPECK128_FLAG = $(POWER8_FLAG)
   else
@@ -639,16 +645,7 @@ ifeq ($(DETECT_FEATURES),1)
   TPROG = TestPrograms/test_ppc_power7.cxx
   TOPT = $(POWER7_FLAG)
   HAVE_OPT = $(shell $(CXX) $(TCXXFLAGS) $(ZOPT) $(TOPT) $(TPROG) -o $(TOUT) 2>&1 | tr ' ' '\n' | wc -l)
-  ifeq ($(strip $(HAVE_OPT)),0)
+  ifneq ($(strip $(HAVE_OPT)),0)
-    ARIA_FLAG = $(POWER7_FLAG)
-    BLAKE2S_FLAG = $(POWER7_FLAG)
-    CHACHA_FLAG = $(POWER7_FLAG)
-    CHAM_FLAG = $(POWER7_FLAG)
-    LEA_FLAG = $(POWER7_FLAG)
-    SIMECK_FLAG = $(POWER7_FLAG)
-    SIMON64_FLAG = $(POWER7_FLAG)
-    SPECK64_FLAG = $(POWER7_FLAG)
-  else
     POWER7_FLAG =
   endif
 
@@ -691,18 +688,12 @@ ifeq ($(DETECT_FEATURES),1)
   #####################################################################
   # Fixups for algorithms that can drop to a lower ISA, if needed
 
-  # Drop to Power7 if Power8 is not available.
+  # Drop to Power4 if Power8 not available
   ifeq ($(POWER8_FLAG),)
-    ifneq ($(POWER7_FLAG),)
-      GCM_FLAG = $(POWER7_FLAG)
-    endif
-  endif
-
-  # Drop to Power4 if Power7 not available
-  ifeq ($(POWER7_FLAG),)
     ifneq ($(ALTIVEC_FLAG),)
       BLAKE2S_FLAG = $(ALTIVEC_FLAG)
       CHACHA_FLAG = $(ALTIVEC_FLAG)
+      GCM_FLAG = $(ALTIVEC_FLAG)
       SIMON64_FLAG = $(ALTIVEC_FLAG)
       SPECK64_FLAG = $(ALTIVEC_FLAG)
     endif
@@ -1428,7 +1419,7 @@ endif # Dependencies
 
 # Cryptogams ARM asm implementation.
 aes_armv4.o : aes_armv4.S
-	$(CC) $(strip $(CXXFLAGS) $(CRYPTOGAMS_AES_FLAG) -mfloat-abi=$(FP_ABI) -c) $<
+	$(CC) $(strip $(CXXFLAGS) $(CRYPTOGAMS_AES_FLAG) -c) $<
 
 # SSSE3 or NEON available
 aria_simd.o : aria_simd.cpp

blake2.cpp

@@ -38,8 +38,8 @@
 // https://github.com/weidai11/cryptopp/issues/743
 #if defined(__xlC__) && (__xlC__ < 0x0d01)
 # define CRYPTOPP_DISABLE_ALTIVEC 1
-# define CRYPTOPP_POWER7_ALTIVEC 1
 # undef CRYPTOPP_POWER7_AVAILABLE
+# undef CRYPTOPP_POWER8_AVAILABLE
 # undef CRYPTOPP_ALTIVEC_AVAILABLE
 #endif
 
@@ -171,8 +171,8 @@ extern void BLAKE2_Compress32_NEON(const byte* input, BLAKE2s_State& state);
 extern void BLAKE2_Compress64_NEON(const byte* input, BLAKE2b_State& state);
 #endif
 
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
-extern void BLAKE2_Compress32_POWER7(const byte* input, BLAKE2s_State& state);
+extern void BLAKE2_Compress32_POWER8(const byte* input, BLAKE2s_State& state);
 #elif CRYPTOPP_ALTIVEC_AVAILABLE
 extern void BLAKE2_Compress32_ALTIVEC(const byte* input, BLAKE2s_State& state);
 #endif
@@ -233,8 +233,8 @@ unsigned int BLAKE2s::OptimalDataAlignment() const
         return 4;
     else
 #endif
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
-    if (HasPower7())
+    if (HasPower8())
         return 16;
     else
 #elif (CRYPTOPP_ALTIVEC_AVAILABLE)
@@ -257,9 +257,9 @@ std::string BLAKE2s::AlgorithmProvider() const
         return "NEON";
     else
 #endif
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
-    if (HasPower7())
+    if (HasPower8())
-        return "Power7";
+        return "Power8";
     else
 #elif (CRYPTOPP_ALTIVEC_AVAILABLE)
     if (HasAltivec())
@@ -690,10 +690,10 @@ void BLAKE2s::Compress(const byte *input)
         return BLAKE2_Compress32_NEON(input, m_state);
     }
 #endif
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
-    if(HasPower7())
+    if(HasPower8())
     {
-        return BLAKE2_Compress32_POWER7(input, m_state);
+        return BLAKE2_Compress32_POWER8(input, m_state);
     }
 #elif CRYPTOPP_ALTIVEC_AVAILABLE
     if(HasAltivec())

blake2b_simd.cpp

@@ -29,6 +29,7 @@
 #if defined(__xlC__) && (__xlC__ < 0x0d01)
 # define CRYPTOPP_DISABLE_ALTIVEC 1
 # undef CRYPTOPP_POWER7_AVAILABLE
+# undef CRYPTOPP_POWER8_AVAILABLE
 # undef CRYPTOPP_ALTIVEC_AVAILABLE
 #endif
 

blake2s_simd.cpp

@@ -38,8 +38,8 @@
 // https://github.com/weidai11/cryptopp/issues/743
 #if defined(__xlC__) && (__xlC__ < 0x0d01)
 # define CRYPTOPP_DISABLE_ALTIVEC 1
-# define CRYPTOPP_POWER7_ALTIVEC 1
 # undef CRYPTOPP_POWER7_AVAILABLE
+# undef CRYPTOPP_POWER8_AVAILABLE
 # undef CRYPTOPP_ALTIVEC_AVAILABLE
 #endif
 
@@ -692,7 +692,7 @@ void BLAKE2_Compress32_NEON(const byte* input, BLAKE2s_State& state)
 }
 #endif  // CRYPTOPP_ARM_NEON_AVAILABLE
 
-#if (CRYPTOPP_POWER7_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE)
 
 inline uint32x4_p VecLoad32(const void* p)
 {
@@ -838,7 +838,7 @@ inline uint32x4_p VectorSet32(const uint32x4_p a, const uint32x4_p b,
     const uint32x4_p t0 = VectorSet32<W,X>(a, b);
     const uint32x4_p t1 = VectorSet32<Y,Z>(c, d);
 
-    // Power7 follows SSE2's implementation, and this is _mm_set_epi32.
+    // PowerPC follows SSE2's implementation, and this is _mm_set_epi32.
     const uint8x16_p mask = {20,21,22,23, 16,17,18,19, 4,5,6,7, 0,1,2,3};
     return VecPermute(t0, t1, mask);
 }
@@ -1015,11 +1015,11 @@ void BLAKE2_Compress32_CORE(const byte* input, BLAKE2s_State& state)
     VecStore32LE(state.h()+0, VecXor(ff0, VecXor(row1, row3)));
     VecStore32LE(state.h()+4, VecXor(ff1, VecXor(row2, row4)));
 }
-#endif  // CRYPTOPP_POWER7_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE
+#endif  // CRYPTOPP_POWER8_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE
 
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
 
-void BLAKE2_Compress32_POWER7(const byte* input, BLAKE2s_State& state)
+void BLAKE2_Compress32_POWER8(const byte* input, BLAKE2s_State& state)
 {
     BLAKE2_Compress32_CORE(input, state);
 }

chacha.cpp

@@ -28,8 +28,8 @@ extern void ChaCha_OperateKeystream_AVX2(const word32 *state, const byte* input,
 extern void ChaCha_OperateKeystream_SSE2(const word32 *state, const byte* input, byte *output, unsigned int rounds);
 #endif
 
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
-extern void ChaCha_OperateKeystream_POWER7(const word32 *state, const byte* input, byte *output, unsigned int rounds);
+extern void ChaCha_OperateKeystream_POWER8(const word32 *state, const byte* input, byte *output, unsigned int rounds);
 #elif (CRYPTOPP_ALTIVEC_AVAILABLE)
 extern void ChaCha_OperateKeystream_ALTIVEC(const word32 *state, const byte* input, byte *output, unsigned int rounds);
 #endif
@@ -153,13 +153,13 @@ void ChaCha_OperateKeystream(KeystreamOperation operation,
         }
 #endif
 
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
-        if (HasPower7())
+        if (HasPower8())
         {
             while (iterationCount >= 4 && MultiBlockSafe(state[12], 4))
             {
                 const bool xorInput = (operation & INPUT_NULL) != INPUT_NULL;
-                ChaCha_OperateKeystream_POWER7(state, xorInput ? input : NULLPTR, output, rounds);
+                ChaCha_OperateKeystream_POWER8(state, xorInput ? input : NULLPTR, output, rounds);
 
                 // MultiBlockSafe avoids overflow on the counter words
                 state[12] += 4;
@@ -267,9 +267,9 @@ std::string ChaCha_AlgorithmProvider()
         return "NEON";
     else
 #endif
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
-    if (HasPower7())
+    if (HasPower8())
-        return "Power7";
+        return "Power8";
     else
 #elif (CRYPTOPP_ALTIVEC_AVAILABLE)
     if (HasAltivec())

chacha_simd.cpp

@@ -209,7 +209,7 @@ inline __m128i RotateLeft<16>(const __m128i val)
 
 #if (CRYPTOPP_ALTIVEC_AVAILABLE)
 
-// ChaCha_OperateKeystream_POWER7 is optimized for POWER7. However, Altivec
+// ChaCha_OperateKeystream_POWER8 is optimized for POWER7. However, Altivec
 // is supported by using vec_ld and vec_st, and using a composite VecAdd
 // that supports 64-bit element adds. vec_ld and vec_st add significant
 // overhead when memory is not aligned. Despite the drawbacks Altivec
@@ -827,7 +827,7 @@ void ChaCha_OperateKeystream_SSE2(const word32 *state, const byte* input, byte *
 
 #endif  // CRYPTOPP_SSE2_INTRIN_AVAILABLE
 
-#if (CRYPTOPP_POWER7_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE)
 
 // ChaCha_OperateKeystream_CORE will use either POWER7 or ALTIVEC,
 // depending on the flags used to compile this source file. The
@@ -1096,11 +1096,11 @@ inline void ChaCha_OperateKeystream_CORE(const word32 *state, const byte* input,
     VecStore32LE(output + 15*16, r3_3);
 }
 
-#endif  // CRYPTOPP_POWER7_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE
+#endif  // CRYPTOPP_POWER8_AVAILABLE || CRYPTOPP_ALTIVEC_AVAILABLE
 
-#if (CRYPTOPP_POWER7_AVAILABLE)
+#if (CRYPTOPP_POWER8_AVAILABLE)
 
-void ChaCha_OperateKeystream_POWER7(const word32 *state, const byte* input, byte *output, unsigned int rounds)
+void ChaCha_OperateKeystream_POWER8(const word32 *state, const byte* input, byte *output, unsigned int rounds)
 {
     ChaCha_OperateKeystream_CORE(state, input, output, rounds);
 }

config.h

@@ -74,24 +74,6 @@
 // Also see https://bugs.llvm.org/show_bug.cgi?id=39895 .
 // #define CRYPTOPP_DISABLE_MIXED_ASM 1
 
-// Several compilers discard SIMD code that loads unaligned data. The symptom
-// is often self test failures and UBsan findings for unaligned loads. For
-// example, Power7 can load unaligned data using vec_vsx_ld but some versions
-// of GCC and Clang require 16-byte aligned data when using the builtin.
-// It is not limited to SSE and PowerPC code. Define this to disable
-// Crypto++ code that uses potentially problematic builtins or intrinsics.
-// Also see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88234 and
-// https://bugs.llvm.org/show_bug.cgi?id=39704
-// #define CRYPTOPP_BUGGY_SIMD_LOAD_AND_STORE 1
-
-// This list will probably grow over time as more compilers are identified.
-#if defined(CRYPTOPP_BUGGY_SIMD_LOAD_AND_STORE)
-# define CRYPTOPP_DISABLE_LEA_SIMD 1
-# define CRYPTOPP_DISABLE_SIMON_SIMD 1
-# define CRYPTOPP_DISABLE_SPECK_SIMD 1
-# define CRYPTOPP_DISABLE_SM4_SIMD 1
-#endif
-
 // Define CRYPTOPP_NO_CXX11 to avoid C++11 related features shown at the
 // end of this file. Some compilers and standard C++ headers advertise C++11
 // but they are really just C++03 with some additional C++11 headers and

gcm.cpp

@@ -75,8 +75,8 @@ extern void GCM_Xor16_SSE2(byte *a, const byte *b, const byte *c);
 extern void GCM_Xor16_NEON(byte *a, const byte *b, const byte *c);
 #endif
 
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
-extern void GCM_Xor16_POWER7(byte *a, const byte *b, const byte *c);
+extern void GCM_Xor16_POWER8(byte *a, const byte *b, const byte *c);
 #endif
 
 #if CRYPTOPP_CLMUL_AVAILABLE
@@ -213,11 +213,11 @@ void GCM_Base::SetKeyWithoutResync(const byte *userKey, size_t keylength, const
                     for (k=1; k<j; k++)
                         GCM_Xor16_NEON(mulTable+i*256*16+(j+k)*16, mulTable+i*256*16+j*16, mulTable+i*256*16+k*16);
             else
-#elif CRYPTOPP_POWER7_AVAILABLE
+#elif CRYPTOPP_POWER8_AVAILABLE
-            if (HasPower7())
+            if (HasPower8())
                 for (j=2; j<=0x80; j*=2)
                     for (k=1; k<j; k++)
-                        GCM_Xor16_POWER7(mulTable+i*256*16+(j+k)*16, mulTable+i*256*16+j*16, mulTable+i*256*16+k*16);
+                        GCM_Xor16_POWER8(mulTable+i*256*16+(j+k)*16, mulTable+i*256*16+j*16, mulTable+i*256*16+k*16);
             else
 #endif
                 for (j=2; j<=0x80; j*=2)
@@ -277,13 +277,13 @@ void GCM_Base::SetKeyWithoutResync(const byte *userKey, size_t keylength, const
                         GCM_Xor16_NEON(mulTable+1024+i*256+(j+k)*16, mulTable+1024+i*256+j*16, mulTable+1024+i*256+k*16);
                     }
             else
-#elif CRYPTOPP_POWER7_AVAILABLE
+#elif CRYPTOPP_POWER8_AVAILABLE
-            if (HasPower7())
+            if (HasPower8())
                 for (j=2; j<=8; j*=2)
                     for (k=1; k<j; k++)
                     {
-                        GCM_Xor16_POWER7(mulTable+i*256+(j+k)*16, mulTable+i*256+j*16, mulTable+i*256+k*16);
+                        GCM_Xor16_POWER8(mulTable+i*256+(j+k)*16, mulTable+i*256+j*16, mulTable+i*256+k*16);
-                        GCM_Xor16_POWER7(mulTable+1024+i*256+(j+k)*16, mulTable+1024+i*256+j*16, mulTable+1024+i*256+k*16);
+                        GCM_Xor16_POWER8(mulTable+1024+i*256+(j+k)*16, mulTable+1024+i*256+j*16, mulTable+1024+i*256+k*16);
                     }
             else
 #endif
@@ -369,8 +369,8 @@ unsigned int GCM_Base::OptimalDataAlignment() const
         HasSSE2() ? 16 :
 #elif CRYPTOPP_ARM_NEON_AVAILABLE
         HasNEON() ? 4 :
-#elif CRYPTOPP_POWER7_AVAILABLE
+#elif CRYPTOPP_POWER8_AVAILABLE
-        HasPower7() ? 16 :
+        HasPower8() ? 16 :
 #endif
         GetBlockCipher().OptimalDataAlignment();
 }

gcm_simd.cpp

@@ -569,12 +569,12 @@ void GCM_ReverseHashBufferIfNeeded_CLMUL(byte *hashBuffer)
 
 // ***************************** POWER8 ***************************** //
 
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
-void GCM_Xor16_POWER7(byte *a, const byte *b, const byte *c)
+void GCM_Xor16_POWER8(byte *a, const byte *b, const byte *c)
 {
     VecStore(VecXor(VecLoad(b), VecLoad(c)), a);
 }
-#endif  // CRYPTOPP_POWER7_AVAILABLE
+#endif  // CRYPTOPP_POWER8_AVAILABLE
 
 #if CRYPTOPP_POWER8_VMULL_AVAILABLE
 

ppc_simd.h

@@ -57,7 +57,7 @@
 //
 //   inline uint32x4_p VecLoad(const byte src[16])
 //   {
-//   #if defined(_ARCH_PWR7)
+//   #if defined(_ARCH_PWR8)
 //       return (uint32x4_p) *(uint8x16_p*)((byte*)src);
 //   #else
 //       return VecLoad_ALTIVEC(src);
@@ -128,7 +128,7 @@ typedef __vector unsigned short  uint16x8_p;
 /// \since Crypto++ 6.0
 typedef __vector unsigned int    uint32x4_p;
 
-#if defined(_ARCH_PWR7) || defined(CRYPTOPP_DOXYGEN_PROCESSING)
+#if defined(_ARCH_PWR8) || defined(CRYPTOPP_DOXYGEN_PROCESSING)
 /// \brief Vector of 64-bit elements
 /// \details uint64x2_p is available on POWER7 and above. Some supporting
 ///   functions, like 64-bit <tt>vec_add</tt> (<tt>vaddudm</tt>), did not
@@ -137,7 +137,7 @@ typedef __vector unsigned int    uint32x4_p;
 ///   __vector unsigned long long
 /// \since Crypto++ 6.0
 typedef __vector unsigned long long uint64x2_p;
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 
 /// \brief The 0 vector
 /// \returns a 32-bit vector of 0's
@@ -252,7 +252,7 @@ inline uint32x4_p VecLoad_ALTIVEC(int off, const byte src[16])
 /// \since Crypto++ 6.0
 inline uint32x4_p VecLoad(const byte src[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
     return (uint32x4_p)vec_xlw4(0, (byte*)src);
 #  elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
@@ -280,7 +280,7 @@ inline uint32x4_p VecLoad(const byte src[16])
 /// \since Crypto++ 6.0
 inline uint32x4_p VecLoad(int off, const byte src[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
     return (uint32x4_p)vec_xlw4(off, (byte*)src);
 #  elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
@@ -328,7 +328,7 @@ inline uint32x4_p VecLoad(int off, const word32 src[4])
     return VecLoad(off, (const byte*)src);
 }
 
-#if defined(_ARCH_PWR7) || defined(CRYPTOPP_DOXYGEN_PROCESSING)
+#if defined(_ARCH_PWR8) || defined(CRYPTOPP_DOXYGEN_PROCESSING)
 
 /// \brief Loads a vector from a word array
 /// \param src the word array
@@ -367,7 +367,7 @@ inline uint64x2_p VecLoad(int off, const word64 src[2])
     return (uint64x2_p)VecLoad(off, (const byte*)src);
 }
 
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 
 /// \brief Loads a vector from an aligned byte array
 /// \param src the byte array
@@ -382,7 +382,7 @@ inline uint64x2_p VecLoad(int off, const word64 src[2])
 /// \since Crypto++ 8.0
 inline uint32x4_p VecLoadAligned(const byte src[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
     return (uint32x4_p)vec_xlw4(0, (byte*)src);
 #  elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
@@ -390,10 +390,10 @@ inline uint32x4_p VecLoadAligned(const byte src[16])
 #  else
     return (uint32x4_p)vec_vsx_ld(0, (byte*)src);
 #  endif
-#else  // _ARCH_PWR7
+#else  // _ARCH_PWR8
     CRYPTOPP_ASSERT(((uintptr_t)src) % 16 == 0);
     return (uint32x4_p)vec_ld(0, (byte*)src);
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 }
 
 /// \brief Loads a vector from an aligned byte array
@@ -410,7 +410,7 @@ inline uint32x4_p VecLoadAligned(const byte src[16])
 /// \since Crypto++ 8.0
 inline uint32x4_p VecLoadAligned(int off, const byte src[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
     return (uint32x4_p)vec_xlw4(off, (byte*)src);
 #  elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
@@ -418,10 +418,10 @@ inline uint32x4_p VecLoadAligned(int off, const byte src[16])
 #  else
     return (uint32x4_p)vec_vsx_ld(off, (byte*)src);
 #  endif
-#else  // _ARCH_PWR7
+#else  // _ARCH_PWR8
     CRYPTOPP_ASSERT((((uintptr_t)src)+off) % 16 == 0);
     return (uint32x4_p)vec_ld(off, (byte*)src);
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 }
 
 /// \brief Loads a vector from a byte array
@@ -439,7 +439,7 @@ inline uint32x4_p VecLoadAligned(int off, const byte src[16])
 /// \since Crypto++ 6.0
 inline uint32x4_p VecLoadBE(const byte src[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
 #    if (CRYPTOPP_BIG_ENDIAN)
        return (uint32x4_p)vec_xlw4(0, (byte*)src);
@@ -455,13 +455,13 @@ inline uint32x4_p VecLoadBE(const byte src[16])
        return (uint32x4_p)VecReverse(vec_vsx_ld(0, (byte*)src));
 #    endif
 #  endif
-#else  // _ARCH_PWR7
+#else  // _ARCH_PWR8
 #  if (CRYPTOPP_BIG_ENDIAN)
      return (uint32x4_p)VecLoad((const byte*)src);
 #  else
      return (uint32x4_p)VecReverse(VecLoad((const byte*)src));
 #  endif
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 }
 
 /// \brief Loads a vector from a byte array
@@ -480,7 +480,7 @@ inline uint32x4_p VecLoadBE(const byte src[16])
 /// \since Crypto++ 6.0
 inline uint32x4_p VecLoadBE(int off, const byte src[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
 #    if (CRYPTOPP_BIG_ENDIAN)
        return (uint32x4_p)vec_xlw4(off, (byte*)src);
@@ -496,13 +496,13 @@ inline uint32x4_p VecLoadBE(int off, const byte src[16])
        return (uint32x4_p)VecReverse(vec_vsx_ld(off, (byte*)src));
 #    endif
 #  endif
-#else  // _ARCH_PWR7
+#else  // _ARCH_PWR8
 #  if (CRYPTOPP_BIG_ENDIAN)
      return (uint32x4_p)VecLoad(off, (const byte*)src);
 #  else
      return (uint32x4_p)VecReverse(VecLoad(off, (const byte*)src));
 #  endif
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 }
 
 //@}
@@ -604,7 +604,7 @@ inline void VecStore_ALTIVEC(const T data, int off, byte dest[16])
 template<class T>
 inline void VecStore(const T data, byte dest[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
     vec_xstw4((uint8x16_p)data, 0, (byte*)dest);
 #  elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
@@ -635,7 +635,7 @@ inline void VecStore(const T data, byte dest[16])
 template<class T>
 inline void VecStore(const T data, int off, byte dest[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
     vec_xstw4((uint8x16_p)data, off, (byte*)dest);
 #  elif defined(__xlc__) || defined(__xlC__) || defined(__clang__)
@@ -750,7 +750,7 @@ inline void VecStore(const T data, int off, word64 dest[2])
 template <class T>
 inline void VecStoreBE(const T data, byte dest[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
 #    if (CRYPTOPP_BIG_ENDIAN)
        vec_xstw4((uint8x16_p)data, 0, (byte*)dest);
@@ -766,13 +766,13 @@ inline void VecStoreBE(const T data, byte dest[16])
        vec_vsx_st((uint8x16_p)VecReverse(data), 0, (byte*)dest);
 #    endif
 #  endif
-#else  // _ARCH_PWR7
+#else  // _ARCH_PWR8
 #  if (CRYPTOPP_BIG_ENDIAN)
      VecStore_ALTIVEC((uint8x16_p)data, 0, (byte*)dest);
 #  else
      VecStore_ALTIVEC((uint8x16_p)VecReverse(data), 0, (byte*)dest);
 #  endif
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 }
 
 /// \brief Stores a vector to a byte array
@@ -794,7 +794,7 @@ inline void VecStoreBE(const T data, byte dest[16])
 template <class T>
 inline void VecStoreBE(const T data, int off, byte dest[16])
 {
-#if defined(_ARCH_PWR7)
+#if defined(_ARCH_PWR8)
 #  if defined(__early_xlc__) || defined(__early_xlC__)
 #    if (CRYPTOPP_BIG_ENDIAN)
        vec_xstw4((uint8x16_p)data, off, (byte*)dest);
@@ -810,13 +810,13 @@ inline void VecStoreBE(const T data, int off, byte dest[16])
        vec_vsx_st((uint8x16_p)VecReverse(data), off, (byte*)dest);
 #    endif
 #  endif
-#else  // _ARCH_PWR7
+#else  // _ARCH_PWR8
 #  if (CRYPTOPP_BIG_ENDIAN)
      VecStore_ALTIVEC((uint8x16_p)data, off, (byte*)dest);
 #  else
      VecStore_ALTIVEC((uint8x16_p)VecReverse(data), off, (byte*)dest);
 #  endif
-#endif  // _ARCH_PWR7
+#endif  // _ARCH_PWR8
 }
 
 /// \brief Stores a vector to a word array

simon.cpp

@@ -255,9 +255,9 @@ std::string SIMON64::Base::AlgorithmProvider() const
     if (HasNEON())
         return "NEON";
 # endif
-# if (CRYPTOPP_POWER7_AVAILABLE)
+# if (CRYPTOPP_POWER8_AVAILABLE)
-    if (HasPower7())
+    if (HasPower8())
-        return "Power7";
+        return "Power8";
 # endif
 # if (CRYPTOPP_ALTIVEC_AVAILABLE)
     if (HasAltivec())

simon128_simd.cpp

@@ -44,7 +44,7 @@
 # include <arm_acle.h>
 #endif
 
-#if defined(CRYPTOPP_POWER7_AVAILABLE)
+#if defined(CRYPTOPP_POWER8_AVAILABLE)
 # include "adv_simd.h"
 # include "ppc_simd.h"
 #endif

simon64_simd.cpp

@@ -576,7 +576,7 @@ inline void SIMON64_Enc_Block(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk1 = vec_splats(subkeys[i]);
         const uint32x4_p rk2 = vec_splats(subkeys[i+1]);
 #else
@@ -592,7 +592,7 @@ inline void SIMON64_Enc_Block(uint32x4_p &block0, uint32x4_p &block1,
 
     if (rounds & 1)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[rounds-1]);
 #else
         const uint8x16_p m = {0,1,2,3, 0,1,2,3, 0,1,2,3, 0,1,2,3};
@@ -634,7 +634,7 @@ inline void SIMON64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1,
     if (rounds & 1)
     {
         std::swap(x1, y1);
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[rounds-1]);
 #else
         const uint8x16_p m = {0,1,2,3, 0,1,2,3, 0,1,2,3, 0,1,2,3};
@@ -647,7 +647,7 @@ inline void SIMON64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk1 = vec_splats(subkeys[i+1]);
         const uint32x4_p rk2 = vec_splats(subkeys[i]);
 #else
@@ -696,7 +696,7 @@ inline void SIMON64_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk1 = vec_splats(subkeys[i]);
         const uint32x4_p rk2 = vec_splats(subkeys[i+1]);
 #else
@@ -717,7 +717,7 @@ inline void SIMON64_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 
     if (rounds & 1)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[rounds-1]);
 #else
         const uint8x16_p m = {0,1,2,3, 0,1,2,3, 0,1,2,3, 0,1,2,3};
@@ -771,7 +771,7 @@ inline void SIMON64_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
     {
         std::swap(x1, y1); std::swap(x2, y2); std::swap(x3, y3);
 
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[rounds-1]);
 #else
         const uint8x16_p m = {0,1,2,3, 0,1,2,3, 0,1,2,3, 0,1,2,3};
@@ -786,7 +786,7 @@ inline void SIMON64_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i = static_cast<int>(rounds-2); i >= 0; i -= 2)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk1 = vec_splats(subkeys[i+1]);
         const uint32x4_p rk2 = vec_splats(subkeys[i]);
 #else

speck.cpp

@@ -235,9 +235,9 @@ std::string SPECK64::Base::AlgorithmProvider() const
     if (HasNEON())
         return "NEON";
 # endif
-# if (CRYPTOPP_POWER7_AVAILABLE)
+# if (CRYPTOPP_POWER8_AVAILABLE)
-    if (HasPower7())
+    if (HasPower8())
-        return "Power7";
+        return "Power8";
 # endif
 # if (CRYPTOPP_ALTIVEC_AVAILABLE)
     if (HasAltivec())

speck64_simd.cpp

@@ -517,7 +517,7 @@ void SPECK64_Enc_Block(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i=0; i < static_cast<int>(rounds); ++i)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[i]);
 #else
         // subkeys has extra elements so memory backs the last subkey
@@ -564,7 +564,7 @@ void SPECK64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i = static_cast<int>(rounds-1); i >= 0; --i)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[i]);
 #else
         // subkeys has extra elements so memory backs the last subkey
@@ -616,7 +616,7 @@ void SPECK64_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i=0; i < static_cast<int>(rounds); ++i)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[i]);
 #else
         // subkeys has extra elements so memory backs the last subkey
@@ -685,7 +685,7 @@ void SPECK64_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 
     for (int i = static_cast<int>(rounds-1); i >= 0; --i)
     {
-#if CRYPTOPP_POWER7_AVAILABLE
+#if CRYPTOPP_POWER8_AVAILABLE
         const uint32x4_p rk = vec_splats(subkeys[i]);
 #else
         // subkeys has extra elements so memory backs the last subkey