added a subset of Crypto++ 5.6.0 with 48% faster ASM SHA-256, combined speedup 2.5x faster vs 0.3.3, thanks BlackEye for figuring out the alignment problem

cryptopp/iterhash.h

@@ -0,0 +1,29 @@
+#ifndef CRYPTOPP_ITERHASH_H
+#define CRYPTOPP_ITERHASH_H
+
+#include "secblock.h"
+
+NAMESPACE_BEGIN(CryptoPP)
+
+// *** trimmed down dependency from iterhash.h ***
+template <class T_HashWordType, class T_Endianness, unsigned int T_BlockSize, unsigned int T_StateSize, class T_Transform, unsigned int T_DigestSize = 0, bool T_StateAligned = false>
+class CRYPTOPP_NO_VTABLE IteratedHashWithStaticTransform
+{
+public:
+	CRYPTOPP_CONSTANT(DIGESTSIZE = T_DigestSize ? T_DigestSize : T_StateSize)
+	unsigned int DigestSize() const {return DIGESTSIZE;};
+    typedef T_HashWordType HashWordType;
+    CRYPTOPP_CONSTANT(BLOCKSIZE = T_BlockSize)
+
+protected:
+	IteratedHashWithStaticTransform() {this->Init();}
+	void HashEndianCorrectedBlock(const T_HashWordType *data) {T_Transform::Transform(this->m_state, data);}
+	void Init() {T_Transform::InitState(this->m_state);}
+
+	T_HashWordType* StateBuf() {return this->m_state;}
+	FixedSizeAlignedSecBlock<T_HashWordType, T_BlockSize/sizeof(T_HashWordType), T_StateAligned> m_state;
+};
+
+NAMESPACE_END
+
+#endif