diff --git a/iterhash.cpp b/iterhash.cpp
index 722d4f77..ce2e5cc3 100644
--- a/iterhash.cpp
+++ b/iterhash.cpp
@@ -82,11 +82,14 @@ template <class T, class BASE> byte * IteratedHashBase<T, BASE>::CreateUpdateSpa
 
 template <class T, class BASE> size_t IteratedHashBase<T, BASE>::HashMultipleBlocks(const T *input, size_t length)
 {
-	// Hardware based SHA1 and SHA256 correct blocks themselves due to hardware requirements.
-	//  For Intel, SHA1 will effectively call ByteReverse(). SHA256 formats data to Intel
-	//  requirements, which means eight words ABCD EFGH are transformed to ABEF CDGH.
+#if CRYPTOPP_BOOL_SSE_SHA_INTRINSICS_AVAILABLE
+	// SHA-1 and SHA-256 only
+	static const bool noReverse = HasSHA() && this->BlockSize() <= 64;
+#else
+	const bool noReverse = NativeByteOrderIs(this->GetByteOrder());
+#endif
+
 	unsigned int blockSize = this->BlockSize();
-	bool noReverse = NativeByteOrderIs(this->GetByteOrder());
 	T* dataBuf = this->DataBuf();
 	do
 	{
diff --git a/sha.cpp b/sha.cpp
index f64ff33d..2c7f32a6 100644
--- a/sha.cpp
+++ b/sha.cpp
@@ -112,7 +112,7 @@ static void SHA1_SSE_SHA_Transform(word32 *state, const word32 *data)
     ABCD = _mm_loadu_si128((__m128i*) state);
     E0 = _mm_set_epi32(state[4], 0, 0, 0);
     ABCD = _mm_shuffle_epi32(ABCD, 0x1B);
-    MASK = _mm_set_epi8(3,2,1,0, 7,6,5,4, 11,10,9,8, 15,14,13,12);
+    MASK = _mm_set_epi8(0,1,2,3, 4,5,6,7, 8,9,10,11, 12,13,14,15);
 
     // Save current hash
     ABCD_SAVE = ABCD;