From 5584ea843d93749c6af15fa3278fd42d477a703a Mon Sep 17 00:00:00 2001
From: inga-lovinde <52715130+inga-lovinde@users.noreply.github.com>
Date: Wed, 5 Apr 2017 15:25:15 +0300
Subject: [PATCH] Performance fixes

---
 .../WhiteRabbit.UnmanagedBridge.vcxproj       |  1 +
 dotnet/WhiteRabbit.UnmanagedBridge/md5.cpp    | 43 +++++++++++--------
 dotnet/WhiteRabbit/MD5Digest.cs               |  9 ++--
 3 files changed, 30 insertions(+), 23 deletions(-)

diff --git a/dotnet/WhiteRabbit.UnmanagedBridge/WhiteRabbit.UnmanagedBridge.vcxproj b/dotnet/WhiteRabbit.UnmanagedBridge/WhiteRabbit.UnmanagedBridge.vcxproj
index a24e9df..c870702 100644
--- a/dotnet/WhiteRabbit.UnmanagedBridge/WhiteRabbit.UnmanagedBridge.vcxproj
+++ b/dotnet/WhiteRabbit.UnmanagedBridge/WhiteRabbit.UnmanagedBridge.vcxproj
@@ -126,6 +126,7 @@
       <IntrinsicFunctions>true</IntrinsicFunctions>
       <FavorSizeOrSpeed>Speed</FavorSizeOrSpeed>
       <AssemblerOutput>AssemblyAndSourceCode</AssemblerOutput>
+      <EnableEnhancedInstructionSet>StreamingSIMDExtensions2</EnableEnhancedInstructionSet>
     </ClCompile>
     <Link>
       <AdditionalDependencies />
diff --git a/dotnet/WhiteRabbit.UnmanagedBridge/md5.cpp b/dotnet/WhiteRabbit.UnmanagedBridge/md5.cpp
index 2fe4c9b..c33b6bd 100644
--- a/dotnet/WhiteRabbit.UnmanagedBridge/md5.cpp
+++ b/dotnet/WhiteRabbit.UnmanagedBridge/md5.cpp
@@ -18,17 +18,21 @@ typedef __m256i MD5Vector;
 #define OP_BLEND(a, b, x) OP_OR(OP_AND(x, b), OP_ANDNOT(x, a))
 
 #define CREATE_VECTOR(a) _mm256_set1_epi32(a)
-#define CREATE_VECTOR_FROM_INPUT(input, offset) _mm256_set1_epi32(input[offset])
+#define CREATE_VECTOR_FROM_INPUT(input, offset) _mm256_set_epi32( \
+    input[offset + 0 * 8], \
+    input[offset + 1 * 8], \
+    input[offset + 2 * 8], \
+    input[offset + 3 * 8], \
+    input[offset + 4 * 8], \
+    input[offset + 5 * 8], \
+    input[offset + 6 * 8], \
+    input[offset + 7 * 8])
 
 #define WRITE_TO_OUTPUT(a, output) \
-    output[0] = a.m256i_u32[0]; \
-    output[1] = a.m256i_u32[1]; \
-    output[2] = a.m256i_u32[2]; \
-    output[3] = a.m256i_u32[3]; \
-    output[4] = a.m256i_u32[4]; \
-    output[5] = a.m256i_u32[5]; \
-    output[6] = a.m256i_u32[6]; \
-    output[7] = a.m256i_u32[7];
+    ((unsigned __int64*)output)[0] = a.m256i_u64[0]; \
+    ((unsigned __int64*)output)[1] = a.m256i_u64[1]; \
+    ((unsigned __int64*)output)[2] = a.m256i_u64[2]; \
+    ((unsigned __int64*)output)[3] = a.m256i_u64[3];
 
 #elif SIMD
 
@@ -41,6 +45,7 @@ typedef __m128i MD5Vector;
 #define OP_ADD(a, b) _mm_add_epi32(a, b)
 #define OP_ROT(a, r) OP_OR(_mm_slli_epi32(a, r), _mm_srli_epi32(a, 32 - (r)))
 #define OP_BLEND(a, b, x) OP_OR(OP_AND(x, b), OP_ANDNOT(x, a))
+//#define OP_BLEND(a, b, x) OP_XOR(a, OP_AND(x, OP_XOR(b, a)))
 
 #define CREATE_VECTOR(a) _mm_set1_epi32(a)
 #define CREATE_VECTOR_FROM_INPUT(input, offset) _mm_set_epi32( \
@@ -158,19 +163,21 @@ static const MD5Parameters Parameters = {
 #define Xor(a, b, c) OP_XOR(a, OP_XOR(b, c))
 #define I(a, b, c) OP_XOR(a, OP_OR(b, OP_NEG(c)))
 
-#define LeftRotate(r, x) OP_ROT(x, r)
+#define StepOuter(r, a, b, x) \
+    a = x; \
+    a = OP_ADD(b, OP_ROT(a, r));
 
-#define Step1(r, a, b, c, d, k, w) OP_ADD(b, LeftRotate(r, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w)))))
-#define Step1E(r, a, b, c, d, k)   OP_ADD(b, LeftRotate(r, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), a))))
+#define Step1(r, a, b, c, d, k, w) StepOuter(r, a, b, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w))))
+#define Step1E(r, a, b, c, d, k)   StepOuter(r, a, b, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), a)))
 
-#define Step2(r, a, b, c, d, k, w) OP_ADD(c, LeftRotate(r, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w)))))
-#define Step2E(r, a, b, c, d, k)   OP_ADD(c, LeftRotate(r, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), a))))
+#define Step2(r, a, b, c, d, k, w) StepOuter(r, a, c, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w))))
+#define Step2E(r, a, b, c, d, k)   StepOuter(r, a, c, OP_ADD(Blend(d, c, b), OP_ADD(CREATE_VECTOR(k), a)))
 
-#define Step3(r, a, b, c, d, k, w) OP_ADD(b, LeftRotate(r, OP_ADD(Xor(b, c, d), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w)))))
-#define Step3E(r, a, b, c, d, k)   OP_ADD(b, LeftRotate(r, OP_ADD(Xor(b, c, d), OP_ADD(CREATE_VECTOR(k), a))))
+#define Step3(r, a, b, c, d, k, w) StepOuter(r, a, b, OP_ADD(Xor(b, c, d), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w))))
+#define Step3E(r, a, b, c, d, k)   StepOuter(r, a, b, OP_ADD(Xor(b, c, d), OP_ADD(CREATE_VECTOR(k), a)))
 
-#define Step4(r, a, b, c, d, k, w) OP_ADD(b, LeftRotate(r, OP_ADD(I(c, b, d), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w)))))
-#define Step4E(r, a, b, c, d, k)   OP_ADD(b, LeftRotate(r, OP_ADD(I(c, b, d), OP_ADD(CREATE_VECTOR(k), a))))
+#define Step4(r, a, b, c, d, k, w) StepOuter(r, a, b, OP_ADD(I(c, b, d), OP_ADD(CREATE_VECTOR(k), OP_ADD(a, w))))
+#define Step4E(r, a, b, c, d, k)   StepOuter(r, a, b, OP_ADD(I(c, b, d), OP_ADD(CREATE_VECTOR(k), a)))
 
 void md5(unsigned __int32 * input, unsigned __int32 * output)
 {
diff --git a/dotnet/WhiteRabbit/MD5Digest.cs b/dotnet/WhiteRabbit/MD5Digest.cs
index 2ff6e02..154210d 100644
--- a/dotnet/WhiteRabbit/MD5Digest.cs
+++ b/dotnet/WhiteRabbit/MD5Digest.cs
@@ -1,9 +1,8 @@
-using System.Numerics;
-using System.Runtime.CompilerServices;
-using WhiteRabbitUnmanagedBridge;
-
-namespace WhiteRabbit
+namespace WhiteRabbit
 {
+    using System.Runtime.CompilerServices;
+    using WhiteRabbitUnmanagedBridge;
+
     internal static class MD5Digest
     {
         // It only returns first component of MD5 hash