diff --git a/gcm.cpp b/gcm.cpp
index 404eaea7..e75c43a4 100644
--- a/gcm.cpp
+++ b/gcm.cpp
@@ -269,7 +269,7 @@ void GCM_Base::SetKeyWithoutResync(const byte *userKey, size_t keylength, const
 
 inline void GCM_Base::ReverseHashBufferIfNeeded()
 {
-#if CRYPTOPP_CLMUL_AVAILABLE
+#if CRYPTOPP_SSSE3_AVAILABLE
     if (HasCLMUL())
     {
         GCM_ReverseHashBufferIfNeeded_SSSE3(HashBuffer());
diff --git a/rijndael-simd.cpp b/rijndael-simd.cpp
index 39a4dd96..b2ff8edb 100644
--- a/rijndael-simd.cpp
+++ b/rijndael-simd.cpp
@@ -16,8 +16,16 @@
 #endif
 
 #if (CRYPTOPP_SSE41_AVAILABLE)
+// Hack... Apple conflates SSE4.1 and SSE4.2. Without __SSE4_2__,
+//   Apple fails the compile with "SSE4.2 instruction set not enabled"
+//   when "nmmintrin.h" is included. Its non-trivial for us to
+//   automatically add -msse4.2 for Apple Clang. We also want to
+//   avoid problems on low-end Atoms which have AES but lack SSE4.2.
+# if (CRYPTOPP_APPLE_CLANG_VERSION)
+#  define __SSE4_2__ 1
+# endif
 # include "nmmintrin.h"
-#endif
+#endif  // CRYPTOPP_SSE41_AVAILABLE
 
 #if (CRYPTOPP_AESNI_AVAILABLE)
 # include "wmmintrin.h"
@@ -129,7 +137,8 @@ void AESNI_Enc_Block(__m128i &block, MAYBE_CONST __m128i *subkeys, unsigned int
 	block = _mm_aesenclast_si128(block, subkeys[rounds]);
 }
 
-inline void AESNI_Enc_4_Blocks(__m128i &block0, __m128i &block1, __m128i &block2, __m128i &block3, MAYBE_CONST __m128i *subkeys, unsigned int rounds)
+inline void AESNI_Enc_4_Blocks(__m128i &block0, __m128i &block1, __m128i &block2, __m128i &block3,
+                               MAYBE_CONST __m128i *subkeys, unsigned int rounds)
 {
 	__m128i rk = subkeys[0];
 	block0 = _mm_xor_si128(block0, rk);
@@ -163,7 +172,8 @@ void AESNI_Dec_Block(__m128i &block, MAYBE_CONST __m128i *subkeys, unsigned int
 	block = _mm_aesdeclast_si128(block, subkeys[rounds]);
 }
 
-void AESNI_Dec_4_Blocks(__m128i &block0, __m128i &block1, __m128i &block2, __m128i &block3, MAYBE_CONST __m128i *subkeys, unsigned int rounds)
+void AESNI_Dec_4_Blocks(__m128i &block0, __m128i &block1, __m128i &block2, __m128i &block3,
+                        MAYBE_CONST __m128i *subkeys, unsigned int rounds)
 {
 	__m128i rk = subkeys[0];
 	block0 = _mm_xor_si128(block0, rk);
@@ -298,16 +308,18 @@ inline size_t Rijndael_AdvancedProcessBlocks_AESNI(F1 func1, F4 func4, MAYBE_CON
 	return length;
 }
 
-size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(MAYBE_CONST __m128i *subkeys, unsigned int rounds, const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
+size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(MAYBE_CONST word32 *subkeys, unsigned int rounds, const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
 {
+	MAYBE_CONST __m128i* keys = reinterpret_cast<MAYBE_CONST __m128i*>(subkeys);
 	return Rijndael_AdvancedProcessBlocks_AESNI(AESNI_Enc_Block, AESNI_Enc_4_Blocks,
-		subkeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
+		keys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
 }
 
-size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(MAYBE_CONST __m128i *subkeys, unsigned int rounds, const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
+size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(MAYBE_CONST word32 *subkeys, unsigned int rounds, const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
 {
+	MAYBE_CONST __m128i* keys = reinterpret_cast<MAYBE_CONST __m128i*>(subkeys);
 	return Rijndael_AdvancedProcessBlocks_AESNI(AESNI_Dec_Block, AESNI_Dec_4_Blocks,
-		subkeys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
+		keys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
 }
 
 void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32 *rk)
diff --git a/rijndael.cpp b/rijndael.cpp
index bea3b45e..4d803001 100644
--- a/rijndael.cpp
+++ b/rijndael.cpp
@@ -74,11 +74,6 @@ being unloaded from L1 cache, until that round is finished.
 #include "misc.h"
 #include "cpu.h"
 
-// TODO: remove...
-#if (CRYPTOPP_AESNI_AVAILABLE)
-# include "wmmintrin.h"
-#endif
-
 // TODO: remove...
 #if (CRYPTOPP_ARM_AES_AVAILABLE)
 # include "arm_neon.h"
@@ -229,9 +224,9 @@ void Rijndael::Base::FillDecTable()
 extern void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32* rk);
 extern void Rijndael_UncheckedSetKeyRev_SSE4_AESNI(word32 *key, unsigned int rounds);
 
-extern size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(MAYBE_CONST __m128i *subkeys, unsigned int rounds,
+extern size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(const word32 *subkeys, unsigned int rounds,
 		const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags);
-extern size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(MAYBE_CONST __m128i *subkeys, unsigned int rounds,
+extern size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(const word32 *subkeys, unsigned int rounds,
 		const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags);
 #endif
 
@@ -1053,8 +1048,7 @@ size_t Rijndael::Enc::AdvancedProcessBlocks(const byte *inBlocks, const byte *xo
 {
 #if CRYPTOPP_AESNI_AVAILABLE
 	if (HasAESNI())
-		return Rijndael_AdvancedProcessBlocks_Enc_AESNI((MAYBE_CONST __m128i *)(const void *)m_key.begin(),
-			m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);
+		return Rijndael_AdvancedProcessBlocks_Enc_AESNI(m_key.begin(), m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);
 
 #endif
 
@@ -1117,8 +1111,7 @@ size_t Rijndael::Dec::AdvancedProcessBlocks(const byte *inBlocks, const byte *xo
 {
 #if CRYPTOPP_AESNI_AVAILABLE
 	if (HasAESNI())
-		return Rijndael_AdvancedProcessBlocks_Dec_AESNI((MAYBE_CONST __m128i *)(const void *)m_key.begin(),
-			m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);
+		return Rijndael_AdvancedProcessBlocks_Dec_AESNI(m_key.begin(), m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);
 #endif
 
 	return BlockTransformation::AdvancedProcessBlocks(inBlocks, xorBlocks, outBlocks, length, flags);