diff --git a/GNUmakefile b/GNUmakefile
index 49ad24ef..ae19f145 100755
--- a/GNUmakefile
+++ b/GNUmakefile
@@ -882,6 +882,8 @@ gcm-simd.o : gcm-simd.cpp
 	$(CXX) $(strip $(CXXFLAGS) $(GCM_FLAG) -c) $<
 
 # AESNI or ARMv7a/ARMv8a available
+rijndael.o : rijndael.cpp
+	$(CXX) $(strip $(CXXFLAGS) $(AES_FLAG) -c) $<
 rijndael-simd.o : rijndael-simd.cpp
 	$(CXX) $(strip $(CXXFLAGS) $(AES_FLAG) -c) $<
 
diff --git a/config.h b/config.h
index 9b69e904..c22be086 100644
--- a/config.h
+++ b/config.h
@@ -521,9 +521,6 @@ NAMESPACE_END
 	#define CRYPTOPP_AESNI_AVAILABLE 1
 #endif
 
-// TODO:
-#undef CRYPTOPP_AESNI_AVAILABLE
-
 #if !defined(CRYPTOPP_DISABLE_ASM) && !defined(CRYPTOPP_DISABLE_SHA) && \
 	(defined(__SHA__) || (CRYPTOPP_MSC_VERSION >= 1900) || \
 	(CRYPTOPP_GCC_VERSION >= 40900) || (__INTEL_COMPILER >= 1300) || \
diff --git a/gcm.cpp b/gcm.cpp
index a3228103..404eaea7 100644
--- a/gcm.cpp
+++ b/gcm.cpp
@@ -29,8 +29,7 @@
 NAMESPACE_BEGIN(CryptoPP)
 
 #if (CRYPTOPP_CLMUL_AVAILABLE)
-# include "wmmintrin.h"
-# include "tmmintrin.h"
+# include "emmintrin.h"
 #endif
 
 #if (CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 || CRYPTOPP_BOOL_X64)
@@ -94,17 +93,8 @@ extern void GCM_ReverseHashBufferIfNeeded_SSSE3(byte *hashBuffer);
 #endif
 
 #if CRYPTOPP_CLMUL_AVAILABLE
-extern __m128i GCM_Multiply_CLMUL(const __m128i &x, const __m128i &h, const __m128i &r);
 extern void GCM_SetKeyWithoutResync_CLMUL(const byte *hashKey, byte *mulTable, unsigned int tableSize);
 extern size_t GCM_AuthenticateBlocks_CLMUL(const byte *data, size_t len, const byte *mtable, byte *hbuffer);
-
-CRYPTOPP_ALIGN_DATA(16)
-const word64 s_clmulConstants64[] = {
-    W64LIT(0xe100000000000000), W64LIT(0xc200000000000000),
-    W64LIT(0x08090a0b0c0d0e0f), W64LIT(0x0001020304050607),
-    W64LIT(0x0001020304050607), W64LIT(0x08090a0b0c0d0e0f)};
-
-const __m128i *s_clmulConstants = (const __m128i *)(const void *)s_clmulConstants64;
 const unsigned int s_cltableSizeInBlocks = 8;
 #endif  // CRYPTOPP_CLMUL_AVAILABLE
 
@@ -113,18 +103,8 @@ extern void GCM_ReverseHashBufferIfNeeded_NEON(byte *hashBuffer);
 #endif
 
 #if CRYPTOPP_ARM_PMULL_AVAILABLE
-extern uint64x2_t GCM_Multiply_PMULL(const uint64x2_t &x, const uint64x2_t &h, const uint64x2_t &r);
 extern void GCM_SetKeyWithoutResync_PMULL(const byte *hashKey, byte *mulTable, unsigned int tableSize);
 extern size_t GCM_AuthenticateBlocks_PMULL(const byte *data, size_t len, const byte *mtable, byte *hbuffer);
-
-CRYPTOPP_ALIGN_DATA(16)
-const word64 s_clmulConstants64[] = {
-    W64LIT(0xe100000000000000), W64LIT(0xc200000000000000),  // Used for ARM and x86; polynomial coefficients
-    W64LIT(0x08090a0b0c0d0e0f), W64LIT(0x0001020304050607),  // Unused for ARM; used for x86 _mm_shuffle_epi8
-    W64LIT(0x0001020304050607), W64LIT(0x08090a0b0c0d0e0f)   // Unused for ARM; used for x86 _mm_shuffle_epi8
-};
-
-const uint64x2_t *s_clmulConstants = (const uint64x2_t *)s_clmulConstants64;
 const unsigned int s_cltableSizeInBlocks = 8;
 #endif  // CRYPTOPP_ARM_PMULL_AVAILABLE
 
diff --git a/rijndael-simd.cpp b/rijndael-simd.cpp
new file mode 100644
index 00000000..a6575c34
--- /dev/null
+++ b/rijndael-simd.cpp
@@ -0,0 +1,195 @@
+// rijndael-simd.cpp - written and placed in the public domain by
+//                     Jeffrey Walton, Uri Blumenthal and Marcel Raad.
+//
+//    This source file uses intrinsics to gain access to AES-NI and
+//    ARMv8a AES instructions. A separate source file is needed
+//    because additional CXXFLAGS are required to enable the
+//    appropriate instructions sets in some build configurations.
+
+#include "pch.h"
+#include "config.h"
+#include "misc.h"
+
+// Clang and GCC hoops...
+#if !(defined(__ARM_FEATURE_CRYPTO) || defined(_MSC_VER))
+# undef CRYPTOPP_ARM_AES_AVAILABLE
+#endif
+
+#if (CRYPTOPP_SSE42_AVAILABLE)
+# include "nmmintrin.h"
+#endif
+
+#if (CRYPTOPP_AESNI_AVAILABLE)
+# include "wmmintrin.h"
+#endif
+
+#if (CRYPTOPP_ARM_AES_AVAILABLE)
+# include "arm_neon.h"
+# include "arm_acle.h"
+#endif
+
+#ifdef CRYPTOPP_GNU_STYLE_INLINE_ASSEMBLY
+# include <signal.h>
+# include <setjmp.h>
+#endif
+
+#ifndef EXCEPTION_EXECUTE_HANDLER
+# define EXCEPTION_EXECUTE_HANDLER 1
+#endif
+
+NAMESPACE_BEGIN(CryptoPP)
+
+#ifdef CRYPTOPP_GNU_STYLE_INLINE_ASSEMBLY
+extern "C" {
+    typedef void (*SigHandler)(int);
+
+	static jmp_buf s_jmpSIGILL;
+	static void SigIllHandler(int)
+	{
+		longjmp(s_jmpSIGILL, 1);
+	}
+};
+#endif  // Not CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY
+
+#if (CRYPTOPP_BOOL_ARM32 || CRYPTOPP_BOOL_ARM64)
+bool CPU_TryAES_ARMV8()
+{
+#if (CRYPTOPP_ARM_AES_AVAILABLE)
+# if defined(CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY)
+	volatile bool result = true;
+	__try
+	{
+		// AES encrypt and decrypt
+		uint8x16_t data = vdupq_n_u8(0), key = vdupq_n_u8(0);
+		uint8x16_t r1 = vaeseq_u8(data, key);
+		uint8x16_t r2 = vaesdq_u8(data, key);
+
+		result = !!(vgetq_lane_u8(r1,0) | vgetq_lane_u8(r2,7));
+	}
+	__except (EXCEPTION_EXECUTE_HANDLER)
+	{
+		return false;
+	}
+	return result;
+# else
+	// longjmp and clobber warnings. Volatile is required.
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
+	volatile bool result = true;
+
+	volatile SigHandler oldHandler = signal(SIGILL, SigIllHandlerAES);
+	if (oldHandler == SIG_ERR)
+		return false;
+
+	volatile sigset_t oldMask;
+	if (sigprocmask(0, NULLPTR, (sigset_t*)&oldMask))
+		return false;
+
+	if (setjmp(s_jmpNoAES))
+		result = false;
+	else
+	{
+		uint8x16_t data = vdupq_n_u8(0), key = vdupq_n_u8(0);
+		uint8x16_t r1 = vaeseq_u8(data, key);
+		uint8x16_t r2 = vaesdq_u8(data, key);
+
+		// Hack... GCC optimizes away the code and returns true
+		result = !!(vgetq_lane_u8(r1,0) | vgetq_lane_u8(r2,7));
+	}
+
+	sigprocmask(SIG_SETMASK, (sigset_t*)&oldMask, NULLPTR);
+	signal(SIGILL, oldHandler);
+	return result;
+# endif
+#else
+	return false;
+#endif  // CRYPTOPP_ARM_AES_AVAILABLE
+}
+#endif  // ARM32 or ARM64
+
+#if (CRYPTOPP_ARM_AES_AVAILABLE)
+#endif  // CRYPTOPP_ARM_AES_AVAILABLE
+
+#if (CRYPTOPP_AESNI_AVAILABLE)
+void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32 *rk)
+{
+	const unsigned rounds = keyLen/4 + 6;
+	static const word32 rcLE[] = {
+		0x01, 0x02, 0x04, 0x08,
+		0x10, 0x20, 0x40, 0x80,
+		0x1B, 0x36, /* for 128-bit blocks, Rijndael never uses more than 10 rcon values */
+	};
+
+	const word32 *ro = rcLE, *rc = rcLE;
+	CRYPTOPP_UNUSED(ro);
+
+	__m128i temp = _mm_loadu_si128((__m128i *)(void *)(userKey+keyLen-16));
+	std::memcpy(rk, userKey, keyLen);
+
+	// keySize: m_key allocates 4*(rounds+1 word32's.
+	const size_t keySize = 4*(rounds+1);
+	const word32* end = rk + keySize;
+	while (true)
+	{
+		CRYPTOPP_ASSERT(rc < ro + COUNTOF(rcLE));
+		rk[keyLen/4] = rk[0] ^ _mm_extract_epi32(_mm_aeskeygenassist_si128(temp, 0), 3) ^ *(rc++);
+		rk[keyLen/4+1] = rk[1] ^ rk[keyLen/4];
+		rk[keyLen/4+2] = rk[2] ^ rk[keyLen/4+1];
+		rk[keyLen/4+3] = rk[3] ^ rk[keyLen/4+2];
+
+		if (rk + keyLen/4 + 4 == end)
+			break;
+
+		if (keyLen == 24)
+		{
+			rk[10] = rk[ 4] ^ rk[ 9];
+			rk[11] = rk[ 5] ^ rk[10];
+
+			CRYPTOPP_ASSERT(keySize >= 12);
+			temp = _mm_insert_epi32(temp, rk[11], 3);
+		}
+		else if (keyLen == 32)
+		{
+			CRYPTOPP_ASSERT(keySize >= 12);
+			temp = _mm_insert_epi32(temp, rk[11], 3);
+			rk[12] = rk[ 4] ^ _mm_extract_epi32(_mm_aeskeygenassist_si128(temp, 0), 2);
+			rk[13] = rk[ 5] ^ rk[12];
+			rk[14] = rk[ 6] ^ rk[13];
+			rk[15] = rk[ 7] ^ rk[14];
+
+			CRYPTOPP_ASSERT(keySize >= 16);
+			temp = _mm_insert_epi32(temp, rk[15], 3);
+		}
+		else
+		{
+			CRYPTOPP_ASSERT(keySize >= 8);
+			temp = _mm_insert_epi32(temp, rk[7], 3);
+		}
+
+		rk += keyLen/4;
+	}
+}
+
+void Rijndael_UncheckedSetKeyRev_SSE4_AESNI(word32 *key, unsigned int rounds)
+{
+	unsigned int i, j;
+	__m128i temp;
+
+#if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120)
+	// __m128i is an unsigned long long[2], and support for swapping it was not added until C++11.
+	// SunCC 12.1 - 12.3 fail to consume the swap; while SunCC 12.4 consumes it without -std=c++11.
+	vec_swap(*(__m128i *)(key), *(__m128i *)(key+4*rounds));
+#else
+	std::swap(*(__m128i *)(void *)(key), *(__m128i *)(void *)(key+4*rounds));
+#endif
+	for (i = 4, j = 4*rounds-4; i < j; i += 4, j -= 4)
+	{
+		temp = _mm_aesimc_si128(*(__m128i *)(void *)(key+i));
+		*(__m128i *)(void *)(key+i) = _mm_aesimc_si128(*(__m128i *)(void *)(key+j));
+		*(__m128i *)(void *)(key+j) = temp;
+	}
+
+	*(__m128i *)(void *)(key+i) = _mm_aesimc_si128(*(__m128i *)(void *)(key+i));	
+}
+#endif  // CRYPTOPP_AESNI_AVAILABLE
+
+NAMESPACE_END
diff --git a/rijndael.cpp b/rijndael.cpp
index b3045b30..f54fec60 100644
--- a/rijndael.cpp
+++ b/rijndael.cpp
@@ -74,6 +74,17 @@ 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"
+# include "arm_acle.h"
+#endif
+
 NAMESPACE_BEGIN(CryptoPP)
 
 // Hack for http://github.com/weidai11/cryptopp/issues/42 and http://github.com/weidai11/cryptopp/issues/132
@@ -214,11 +225,16 @@ void Rijndael::Base::FillDecTable()
 	s_TdFilled = true;
 }
 
-void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keylen, const NameValuePairs &)
-{
-	AssertValidKeyLength(keylen);
+#if (CRYPTOPP_AESNI_AVAILABLE)
+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);
+#endif
 
-	m_rounds = keylen/4 + 6;
+void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keyLen, const NameValuePairs &)
+{
+	AssertValidKeyLength(keyLen);
+
+	m_rounds = keyLen/4 + 6;
 	m_key.New(4*(m_rounds+1));
 
 	word32 *rk = m_key;
@@ -227,110 +243,36 @@ void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keylen, c
 	// MSVC 2008 SP1 generates bad code for _mm_extract_epi32() when compiling for X64
 	if (HasAESNI() && HasSSE4())
 	{
-		static const word32 rcLE[] = {
-			0x01, 0x02, 0x04, 0x08,
-			0x10, 0x20, 0x40, 0x80,
-			0x1B, 0x36, /* for 128-bit blocks, Rijndael never uses more than 10 rcon values */
-		};
-
-		// Coverity finding, appears to be false positive. Assert the condition.
-		const word32 *ro = rcLE, *rc = rcLE;
-		CRYPTOPP_UNUSED(ro);
-
-		__m128i temp = _mm_loadu_si128((__m128i *)(void *)(userKey+keylen-16));
-		memcpy(rk, userKey, keylen);
-
-		while (true)
-		{
-			// Coverity finding, appears to be false positive. Assert the condition.
-			CRYPTOPP_ASSERT(rc < ro + COUNTOF(rcLE));
-			rk[keylen/4] = rk[0] ^ _mm_extract_epi32(_mm_aeskeygenassist_si128(temp, 0), 3) ^ *(rc++);
-			rk[keylen/4+1] = rk[1] ^ rk[keylen/4];
-			rk[keylen/4+2] = rk[2] ^ rk[keylen/4+1];
-			rk[keylen/4+3] = rk[3] ^ rk[keylen/4+2];
-
-			if (rk + keylen/4 + 4 == m_key.end())
-				break;
-
-			if (keylen == 24)
-			{
-				rk[10] = rk[ 4] ^ rk[ 9];
-				rk[11] = rk[ 5] ^ rk[10];
-				// Coverity finding, appears to be false positive. Assert the condition.
-				CRYPTOPP_ASSERT(m_key.size() >= 12);
-				temp = _mm_insert_epi32(temp, rk[11], 3);
-			}
-			else if (keylen == 32)
-			{
-				// Coverity finding, appears to be false positive. Assert the condition.
-				CRYPTOPP_ASSERT(m_key.size() >= 12);
-				temp = _mm_insert_epi32(temp, rk[11], 3);
-    			rk[12] = rk[ 4] ^ _mm_extract_epi32(_mm_aeskeygenassist_si128(temp, 0), 2);
-    			rk[13] = rk[ 5] ^ rk[12];
-    			rk[14] = rk[ 6] ^ rk[13];
-    			rk[15] = rk[ 7] ^ rk[14];
-				// Coverity finding, appears to be false positive. Assert the condition.
-				CRYPTOPP_ASSERT(m_key.size() >= 16);
-				temp = _mm_insert_epi32(temp, rk[15], 3);
-			}
-			else
-			{
-				// Coverity finding, appears to be false positive. Assert the condition.
-				CRYPTOPP_ASSERT(m_key.size() >= 8);
-				temp = _mm_insert_epi32(temp, rk[7], 3);
-			}
-
-			rk += keylen/4;
-		}
-
+		Rijndael_UncheckedSetKey_SSE4_AESNI(userKey, keyLen, rk);
 		if (!IsForwardTransformation())
-		{
-			rk = m_key;
-			unsigned int i, j;
-
-#if defined(__SUNPRO_CC) && (__SUNPRO_CC <= 0x5120)
-			// __m128i is an unsigned long long[2], and support for swapping it was not added until C++11.
-			// SunCC 12.1 - 12.3 fail to consume the swap; while SunCC 12.4 consumes it without -std=c++11.
-			vec_swap(*(__m128i *)(rk), *(__m128i *)(rk+4*m_rounds));
-#else
-			std::swap(*(__m128i *)(void *)(rk), *(__m128i *)(void *)(rk+4*m_rounds));
-#endif
-			for (i = 4, j = 4*m_rounds-4; i < j; i += 4, j -= 4)
-			{
-				temp = _mm_aesimc_si128(*(__m128i *)(void *)(rk+i));
-				*(__m128i *)(void *)(rk+i) = _mm_aesimc_si128(*(__m128i *)(void *)(rk+j));
-				*(__m128i *)(void *)(rk+j) = temp;
-			}
-
-			*(__m128i *)(void *)(rk+i) = _mm_aesimc_si128(*(__m128i *)(void *)(rk+i));
-		}
+			Rijndael_UncheckedSetKeyRev_SSE4_AESNI(m_key, m_rounds);
 
 		return;
 	}
 #endif
 
-	GetUserKey(BIG_ENDIAN_ORDER, rk, keylen/4, userKey, keylen);
+	GetUserKey(BIG_ENDIAN_ORDER, rk, keyLen/4, userKey, keyLen);
 	const word32 *rc = rcon;
 	word32 temp;
 
 	while (true)
 	{
-		temp  = rk[keylen/4-1];
+		temp  = rk[keyLen/4-1];
 		word32 x = (word32(Se[GETBYTE(temp, 2)]) << 24) ^ (word32(Se[GETBYTE(temp, 1)]) << 16) ^ (word32(Se[GETBYTE(temp, 0)]) << 8) ^ Se[GETBYTE(temp, 3)];
-		rk[keylen/4] = rk[0] ^ x ^ *(rc++);
-		rk[keylen/4+1] = rk[1] ^ rk[keylen/4];
-		rk[keylen/4+2] = rk[2] ^ rk[keylen/4+1];
-		rk[keylen/4+3] = rk[3] ^ rk[keylen/4+2];
+		rk[keyLen/4] = rk[0] ^ x ^ *(rc++);
+		rk[keyLen/4+1] = rk[1] ^ rk[keyLen/4];
+		rk[keyLen/4+2] = rk[2] ^ rk[keyLen/4+1];
+		rk[keyLen/4+3] = rk[3] ^ rk[keyLen/4+2];
 
-		if (rk + keylen/4 + 4 == m_key.end())
+		if (rk + keyLen/4 + 4 == m_key.end())
 			break;
 
-		if (keylen == 24)
+		if (keyLen == 24)
 		{
 			rk[10] = rk[ 4] ^ rk[ 9];
 			rk[11] = rk[ 5] ^ rk[10];
 		}
-		else if (keylen == 32)
+		else if (keyLen == 32)
 		{
     		temp = rk[11];
     		rk[12] = rk[ 4] ^ (word32(Se[GETBYTE(temp, 3)]) << 24) ^ (word32(Se[GETBYTE(temp, 2)]) << 16) ^ (word32(Se[GETBYTE(temp, 1)]) << 8) ^ Se[GETBYTE(temp, 0)];
@@ -338,7 +280,7 @@ void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keylen, c
     		rk[14] = rk[ 6] ^ rk[13];
     		rk[15] = rk[ 7] ^ rk[14];
 		}
-		rk += keylen/4;
+		rk += keyLen/4;
 	}
 
 	rk = m_key;