From ced7cff64f87ae4fec781e00e4e111f6c188dea2 Mon Sep 17 00:00:00 2001
From: Jeffrey Walton <noloader@gmail.com>
Date: Fri, 22 Sep 2017 09:39:36 -0400
Subject: [PATCH] Add Power8 SHA256 and SHA512 support (GH #513)

---
 ppc-simd.cpp | 41 +++++++++++++++++++++++++++--------------
 sha-simd.cpp |  2 +-
 sha.cpp      | 14 ++++++++++++++
 3 files changed, 42 insertions(+), 15 deletions(-)

diff --git a/ppc-simd.cpp b/ppc-simd.cpp
index cabe8e7c..652c18cb 100644
--- a/ppc-simd.cpp
+++ b/ppc-simd.cpp
@@ -75,22 +75,15 @@ bool CPU_ProbeAltivec()
 		result = false;
 	else
 	{
-		CRYPTOPP_ALIGN_DATA(16)
 		const byte b1[16] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
-		CRYPTOPP_ALIGN_DATA(16)
 		const byte b2[16] = {1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
-		CRYPTOPP_ALIGN_DATA(16) byte b3[16];
-#if defined(CRYPTOPP_XLC_VERSION)
-		const uint8x16_p8 v1 = VectorLoad(0, b1);
-		const uint8x16_p8 v2 = VectorLoad(0, b2);
-		const uint8x16_p8 v3 = VectorXor(v1, v2);
-		vec_st(v3, 0, (byte*)b3);
-#elif defined(CRYPTOPP_GCC_VERSION)
-		const uint64x2_p8 v1 = (uint64x2_p8)VectorLoad(0, b1);
-		const uint64x2_p8 v2 = (uint64x2_p8)VectorLoad(0, b2);
-		const uint64x2_p8 v3 = (uint64x2_p8)VectorXor(v1, v2);
-		vec_st((uint8x16_p8)v3, 0, (byte*)b3);
-#endif
+		byte b3[16];
+
+		const uint8x16_p8 v1 = (uint8x16_p8)VectorLoad(0, b1);
+		const uint8x16_p8 v2 = (uint8x16_p8)VectorLoad(0, b2);
+		const uint8x16_p8 v3 = (uint8x16_p8)VectorXor(v1, v2);
+		VectorStore(v3, b3);
+
 		result = (0 == std::memcmp(b2, b3, 16));
 	}
 
@@ -210,6 +203,8 @@ bool CPU_ProbeAES()
 		uint8x16_p8 s = (uint8x16_p8)VectorLoad(0, state);
 		s = VectorEncrypt(s, k);
 		s = VectorEncryptLast(s, k);
+		s = VectorDecrypt(s, k);
+		s = VectorDecryptLast(s, k);
 		VectorStore(s, r);
 
 		result = (0 != std::memcmp(r, z, 16));
@@ -247,7 +242,16 @@ bool CPU_ProbeSHA256()
 		result = false;
 	else
 	{
+		byte r[16], z[16] = {0};
+		uint8x16_p8 x = ((uint8x16_p8){0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0});
 
+		x = VectorSHA256<0,0>(x);
+		x = VectorSHA256<0,1>(x);
+		x = VectorSHA256<1,0>(x);
+		x = VectorSHA256<1,1>(x);
+		VectorStore(x, r);
+
+		result = (0 != std::memcmp(r, z, 16));
 	}
 
 	sigprocmask(SIG_SETMASK, (sigset_t*)&oldMask, NULLPTR);
@@ -282,7 +286,16 @@ bool CPU_ProbeSHA512()
 		result = false;
 	else
 	{
+		byte r[16], z[16] = {0};
+		uint8x16_p8 x = ((uint8x16_p8){0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0});
 
+		x = VectorSHA512<0,0>(x);
+		x = VectorSHA512<0,1>(x);
+		x = VectorSHA512<1,0>(x);
+		x = VectorSHA512<1,1>(x);
+		VectorStore(x, r);
+
+		result = (0 != std::memcmp(r, z, 16));
 	}
 
 	sigprocmask(SIG_SETMASK, (sigset_t*)&oldMask, NULLPTR);
diff --git a/sha-simd.cpp b/sha-simd.cpp
index 9b16fb1c..4250c41b 100644
--- a/sha-simd.cpp
+++ b/sha-simd.cpp
@@ -991,7 +991,7 @@ void SHA512_HashMultipleBlocks_POWER8(word64 *state, const word64 *data, size_t
     CRYPTOPP_ASSERT(state);
     CRYPTOPP_ASSERT(data);
     CRYPTOPP_ASSERT(length >= SHA512::BLOCKSIZE);
-	
+
 	CRYPTOPP_ASSERT(0);
 }
 
diff --git a/sha.cpp b/sha.cpp
index 04dfb4d9..d1c8d454 100644
--- a/sha.cpp
+++ b/sha.cpp
@@ -783,6 +783,13 @@ size_t SHA224::HashMultipleBlocks(const word32 *input, size_t length)
         return length & (SHA256::BLOCKSIZE - 1);
     }
 #endif
+#if CRYPTOPP_POWER8_SHA_AVAILABLE
+    if (HasSHA256())
+    {
+        SHA256_HashMultipleBlocks_POWER8(m_state, input, length, BIG_ENDIAN_ORDER);
+        return length & (SHA256::BLOCKSIZE - 1);
+    }
+#endif
 
     const bool noReverse = NativeByteOrderIs(this->GetByteOrder());
     word32 *dataBuf = this->DataBuf();
@@ -1164,6 +1171,13 @@ void SHA512::Transform(word64 *state, const word64 *data)
         return;
     }
 #endif
+#if CRYPTOPP_POWER8_SHA_AVAILABLE
+    if (HasSHA512())
+    {
+        SHA512_HashMultipleBlocks_POWER8(state, data, SHA512::BLOCKSIZE, BIG_ENDIAN_ORDER);
+        return;
+    }
+#endif
 
     SHA512_HashBlock_CXX(state, data);
 }