diff --git a/config.compat b/config.compat
index 11bd174c..6392a786 100644
--- a/config.compat
+++ b/config.compat
@@ -502,7 +502,7 @@ NAMESPACE_END
 	#define CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE 0
 #endif
 
-#if !defined(CRYPTOPP_DISABLE_ASM) && !defined(CRYPTOPP_DISABLE_SHA) && !defined(_M_ARM) && ((_MSC_VER >= 1900) || (CRYPTOPP_GCC_VERSION >= 50000) || defined(__SHA__))
+#if !defined(CRYPTOPP_DISABLE_ASM) && !defined(CRYPTOPP_DISABLE_SHA) && !defined(_M_ARM) && ((_MSC_VER >= 1900) || defined(__SHA__))
 	#define CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE 1
 #else
 	#define CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE 0
diff --git a/config.h b/config.h
index dcdf1ac1..f2a007fb 100644
--- a/config.h
+++ b/config.h
@@ -502,7 +502,7 @@ NAMESPACE_END
 	#define CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE 0
 #endif
 
-#if !defined(CRYPTOPP_DISABLE_ASM) && !defined(CRYPTOPP_DISABLE_SHA) && !defined(_M_ARM) && ((_MSC_VER >= 1900) || (CRYPTOPP_GCC_VERSION >= 50000) || defined(__SHA__))
+#if !defined(CRYPTOPP_DISABLE_ASM) && !defined(CRYPTOPP_DISABLE_SHA) && !defined(_M_ARM) && ((_MSC_VER >= 1900) || defined(__SHA__))
 	#define CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE 1
 #else
 	#define CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE 0
diff --git a/sha.cpp b/sha.cpp
index 1b75194a..c8a770ac 100644
--- a/sha.cpp
+++ b/sha.cpp
@@ -750,12 +750,11 @@ size_t SHA224::HashMultipleBlocks(const word32 *input, size_t length)
 #define s0(x) (rotrFixed(x,7)^rotrFixed(x,18)^(x>>3))
 #define s1(x) (rotrFixed(x,17)^rotrFixed(x,19)^(x>>10))
 
-// Smaller but slower
 #if defined(__OPTIMIZE_SIZE__)
+// Smaller but slower
 void SHA256_CXX_Transform(word32 *state, const word32 *data)
 {
-	word32 T[20];
-	word32 W[32];
+	word32 W[32], T[20];
 	unsigned int i = 0, j = 0;
 	word32 *t = T+8;
 
@@ -824,15 +823,10 @@ void SHA256_CXX_Transform(word32 *state, const word32 *data)
     state[7] += t[7];
 }
 #else
+// Bigger but faster
 void SHA256_CXX_Transform(word32 *state, const word32 *data)
 {
-	word32 W[16];
-#if (defined(CRYPTOPP_X86_ASM_AVAILABLE) || defined(CRYPTOPP_X32_ASM_AVAILABLE) || defined(CRYPTOPP_X64_MASM_AVAILABLE)) && !defined(CRYPTOPP_DISABLE_SHA_ASM)
-	// this byte reverse is a waste of time, but this function is only called by MDC
-	ByteReverse(W, data, SHA256::BLOCKSIZE);
-	X86_SHA256_HashBlocks(state, W, SHA256::BLOCKSIZE - !HasSSE2());
-#else
-	word32 T[8];
+	word32 W[16], T[8];
     /* Copy context->state[] to working vars */
 	memcpy(T, state, sizeof(T));
     /* 64 operations, partially loop unrolled */
@@ -852,9 +846,8 @@ void SHA256_CXX_Transform(word32 *state, const word32 *data)
     state[5] += f(0);
     state[6] += g(0);
     state[7] += h(0);
-#endif
 }
-#endif
+#endif  // __OPTIMIZE_SIZE__
 
 #undef S0
 #undef S1
@@ -862,16 +855,28 @@ void SHA256_CXX_Transform(word32 *state, const word32 *data)
 #undef s1
 #undef R
 
+#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE
+static void SHA256_SSE2_Transform(word32 *state, const word32 *data)
+{
+	// this byte reverse is a waste of time, but this function is only called by MDC
+	word32 W[16];
+	ByteReverse(W, data, SHA256::BLOCKSIZE);
+	X86_SHA256_HashBlocks(state, W, SHA256::BLOCKSIZE - !HasSSE2());
+}
+#endif  // CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE
+
 #if CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE
 static void SHA256_SSE_SHA_Transform(word32 *state, const word32 *data)
 {
     return SHA256_SSE_SHA_HashBlocks(state, data, SHA256::BLOCKSIZE);
 }
+#endif  // CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE
 
 ///////////////////////////////////
 // start of Walton/Gulley's code //
 ///////////////////////////////////
 
+#if CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE
 // Based on http://software.intel.com/en-us/articles/intel-sha-extensions and code by Sean Gulley.
 static void SHA256_SSE_SHA_HashBlocks(word32 *state, const word32 *data, size_t length)
 {
@@ -1081,6 +1086,11 @@ pfnSHATransform InitializeSHA256Transform()
         return &SHA256_SSE_SHA_Transform;
     else
 #endif
+#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE
+    if (HasSSE2())
+        return &SHA256_SSE2_Transform;
+    else
+#endif
 
     return &SHA256_CXX_Transform;
 }
diff --git a/sha.h b/sha.h
index 97853b3a..b80817aa 100644
--- a/sha.h
+++ b/sha.h
@@ -2,6 +2,7 @@
 
 //! \file sha.h
 //! \brief Classes for SHA-1 and SHA-2 family of message digests
+//! \since SHA1 since Crypto++ 1.0, SHA2 since Crypto++ 4.0, Intel SHA extensions since Crypto++ 5.7
 
 #ifndef CRYPTOPP_SHA_H
 #define CRYPTOPP_SHA_H
@@ -20,7 +21,7 @@ NAMESPACE_BEGIN(CryptoPP)
 //! \class SHA1
 //! \brief SHA-1 message digest
 //! \sa <a href="http://www.weidai.com/scan-mirror/md.html#SHA-1">SHA-1</a>
-//! \since Crypto++ 1.0
+//! \since Crypto++ 1.0, Intel SHA extensions since Crypto++ 5.7
 class CRYPTOPP_DLL SHA1 : public IteratedHashWithStaticTransform<word32, BigEndian, 64, 20, SHA1>
 {
 public:
@@ -34,7 +35,7 @@ typedef SHA1 SHA;	// for backwards compatibility
 //! \class SHA256
 //! \brief SHA-256 message digest
 //! \sa <a href="http://www.weidai.com/scan-mirror/md.html#SHA-256">SHA-256</a>
-//! \since Crypto++ 4.0
+//! \since Crypto++ 4.0, Intel SHA extensions since Crypto++ 5.7
 class CRYPTOPP_DLL SHA256 : public IteratedHashWithStaticTransform<word32, BigEndian, 64, 32, SHA256, 32, true>
 {
 public:
@@ -49,7 +50,7 @@ public:
 //! \class SHA224
 //! \brief SHA-224 message digest
 //! \sa <a href="http://www.weidai.com/scan-mirror/md.html#SHA-224">SHA-224</a>
-//! \since Crypto++ 4.0
+//! \since Crypto++ 4.0, Intel SHA extensions since Crypto++ 5.7
 class CRYPTOPP_DLL SHA224 : public IteratedHashWithStaticTransform<word32, BigEndian, 64, 32, SHA224, 28, true>
 {
 public: