Add SSE2 Serpent decryption

2 files changed, 307 insertions, 1 deletions

diff --git a/src/block/serpent_sse2/serp_sse2.cpp b/src/block/serpent_sse2/serp_sse2.cpp
index 5ce7d8f47..a3319afec 100644
--- a/src/block/serpent_sse2/serp_sse2.cpp
+++ b/src/block/serpent_sse2/serp_sse2.cpp
@@ -24,11 +24,14 @@ namespace {
    } while(0);
 
 /*
-* Serpent's linear transformation
+* Serpent's linear transformations
 */
 #define rotate_left_m128(vec, rot)              \
    _mm_or_si128(_mm_slli_epi32(vec, rot), _mm_srli_epi32(vec, 32-rot))
 
+#define rotate_right_m128(vec, rot)              \
+   _mm_or_si128(_mm_srli_epi32(vec, rot), _mm_slli_epi32(vec, 32-rot))
+
 #define transform(B0, B1, B2, B3)                                       \
    do {                                                                 \
       B0 = rotate_left_m128(B0, 13);                                    \
@@ -43,6 +46,20 @@ namespace {
       B2 = rotate_left_m128(B2, 22);                                    \
    } while(0);
 
+#define i_transform(B0, B1, B2, B3)                                     \
+   do {                                                                 \
+      B2 = rotate_right_m128(B2, 22);                                   \
+      B0 = rotate_right_m128(B0, 5);                                    \
+      B2 = _mm_xor_si128(B2, _mm_xor_si128(B3, _mm_slli_epi32(B1, 7))); \
+      B0 = _mm_xor_si128(B0, _mm_xor_si128(B1, B3));                    \
+      B3 = rotate_right_m128(B3, 7);                                    \
+      B1 = rotate_right_m128(B1, 1);                                    \
+      B3 = _mm_xor_si128(B3, _mm_xor_si128(B2, _mm_slli_epi32(B0, 3))); \
+      B1 = _mm_xor_si128(B1, _mm_xor_si128(B0, B2));                    \
+      B2 = rotate_right_m128(B2, 3);                                    \
+      B0 = rotate_right_m128(B0, 13);                                   \
+   } while(0);
+
 /*
 * 4x4 SSE2 integer matrix transpose
 */
@@ -122,6 +139,70 @@ void serpent_encrypt_4(const byte in[64],
    _mm_storeu_si128(out_mm + 3, B3);
    }
 
+/*
+* SSE2 Serpent Decryption of 4 blocks in parallel
+*/
+void serpent_decrypt_4(const byte in[64],
+                       byte out[64],
+                       const u32bit keys_32[132])
+   {
+   const __m128i all_ones = _mm_set1_epi8(0xFF);
+
+   const __m128i* keys = (const __m128i*)(keys_32);
+   __m128i* out_mm = (__m128i*)(out);
+   __m128i* in_mm = (__m128i*)(in);
+
+   __m128i B0 = _mm_loadu_si128(in_mm);
+   __m128i B1 = _mm_loadu_si128(in_mm + 1);
+   __m128i B2 = _mm_loadu_si128(in_mm + 2);
+   __m128i B3 = _mm_loadu_si128(in_mm + 3);
+
+   transpose(B0, B1, B2, B3);
+
+   key_xor(32,B0,B1,B2,B3);  SBoxD8(B0,B1,B2,B3); key_xor(31,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD7(B0,B1,B2,B3); key_xor(30,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD6(B0,B1,B2,B3); key_xor(29,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD5(B0,B1,B2,B3); key_xor(28,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD4(B0,B1,B2,B3); key_xor(27,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD3(B0,B1,B2,B3); key_xor(26,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD2(B0,B1,B2,B3); key_xor(25,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD1(B0,B1,B2,B3); key_xor(24,B0,B1,B2,B3);
+
+   i_transform(B0,B1,B2,B3); SBoxD8(B0,B1,B2,B3); key_xor(23,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD7(B0,B1,B2,B3); key_xor(22,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD6(B0,B1,B2,B3); key_xor(21,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD5(B0,B1,B2,B3); key_xor(20,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD4(B0,B1,B2,B3); key_xor(19,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD3(B0,B1,B2,B3); key_xor(18,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD2(B0,B1,B2,B3); key_xor(17,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD1(B0,B1,B2,B3); key_xor(16,B0,B1,B2,B3);
+
+   i_transform(B0,B1,B2,B3); SBoxD8(B0,B1,B2,B3); key_xor(15,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD7(B0,B1,B2,B3); key_xor(14,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD6(B0,B1,B2,B3); key_xor(13,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD5(B0,B1,B2,B3); key_xor(12,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD4(B0,B1,B2,B3); key_xor(11,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD3(B0,B1,B2,B3); key_xor(10,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD2(B0,B1,B2,B3); key_xor( 9,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD1(B0,B1,B2,B3); key_xor( 8,B0,B1,B2,B3);
+
+   i_transform(B0,B1,B2,B3); SBoxD8(B0,B1,B2,B3); key_xor( 7,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD7(B0,B1,B2,B3); key_xor( 6,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD6(B0,B1,B2,B3); key_xor( 5,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD5(B0,B1,B2,B3); key_xor( 4,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD4(B0,B1,B2,B3); key_xor( 3,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD3(B0,B1,B2,B3); key_xor( 2,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD2(B0,B1,B2,B3); key_xor( 1,B0,B1,B2,B3);
+   i_transform(B0,B1,B2,B3); SBoxD1(B0,B1,B2,B3); key_xor( 0,B0,B1,B2,B3);
+
+   transpose(B0, B1, B2, B3);
+
+   _mm_storeu_si128(out_mm    , B0);
+   _mm_storeu_si128(out_mm + 1, B1);
+   _mm_storeu_si128(out_mm + 2, B2);
+   _mm_storeu_si128(out_mm + 3, B3);
+   }
+
 }
 
 /*
@@ -150,6 +231,14 @@ void Serpent_SSE2::encrypt_n(const byte in[], byte out[], u32bit blocks) const
 */
 void Serpent_SSE2::decrypt_n(const byte in[], byte out[], u32bit blocks) const
    {
+   while(blocks >= 4)
+      {
+      serpent_decrypt_4(in, out, this->round_key);
+      in += 4 * BLOCK_SIZE;
+      out += 4 * BLOCK_SIZE;
+      blocks -= 4;
+      }
+
    for(u32bit i = 0; i != blocks; ++i)
       {
       Serpent::decrypt_n(in, out, 1);
diff --git a/src/block/serpent_sse2/serp_sse2_sbox.h b/src/block/serpent_sse2/serp_sse2_sbox.h
index 1660643ad..40c552e87 100644
--- a/src/block/serpent_sse2/serp_sse2_sbox.h
+++ b/src/block/serpent_sse2/serp_sse2_sbox.h
@@ -214,4 +214,221 @@
       B0 = B4;                                    \
    } while(0);
 
+#define SBoxD1(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      B2 = _mm_xor_si128(B2, all_ones);           \
+      __m128i B4 = B1;                            \
+      B1 = _mm_or_si128(B1, B0);                  \
+      B4 = _mm_xor_si128(B4, all_ones);           \
+      B1 = _mm_xor_si128(B1, B2);                 \
+      B2 = _mm_or_si128(B2, B4);                  \
+      B1 = _mm_xor_si128(B1, B3);                 \
+      B0 = _mm_xor_si128(B0, B4);                 \
+      B2 = _mm_xor_si128(B2, B0);                 \
+      B0 = _mm_and_si128(B0, B3);                 \
+      B4 = _mm_xor_si128(B4, B0);                 \
+      B0 = _mm_or_si128(B0, B1);                  \
+      B0 = _mm_xor_si128(B0, B2);                 \
+      B3 = _mm_xor_si128(B3, B4);                 \
+      B2 = _mm_xor_si128(B2, B1);                 \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B2 = _mm_and_si128(B2, B3);                 \
+      B4 = _mm_xor_si128(B4, B2);                 \
+      B2 = B1;                                    \
+      B1 = B4;                                    \
+      } while(0);
+
+#define SBoxD2(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      __m128i B4 = B1;                            \
+      B1 = _mm_xor_si128(B1, B3);                 \
+      B3 = _mm_and_si128(B3, B1);                 \
+      B4 = _mm_xor_si128(B4, B2);                 \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      B0 = _mm_or_si128(B0, B1);                  \
+      B2 = _mm_xor_si128(B2, B3);                 \
+      B0 = _mm_xor_si128(B0, B4);                 \
+      B0 = _mm_or_si128(B0, B2);                  \
+      B1 = _mm_xor_si128(B1, B3);                 \
+      B0 = _mm_xor_si128(B0, B1);                 \
+      B1 = _mm_or_si128(B1, B3);                  \
+      B1 = _mm_xor_si128(B1, B0);                 \
+      B4 = _mm_xor_si128(B4, all_ones);           \
+      B4 = _mm_xor_si128(B4, B1);                 \
+      B1 = _mm_or_si128(B1, B0);                  \
+      B1 = _mm_xor_si128(B1, B0);                 \
+      B1 = _mm_or_si128(B1, B4);                  \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B1 = B0;                                    \
+      B0 = B4;                                    \
+      B4 = B2;                                    \
+      B2 = B3;                                    \
+      B3 = B4;                                    \
+      } while(0);
+
+#define SBoxD3(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      B2 = _mm_xor_si128(B2, B3);                 \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      __m128i B4 = B3;                            \
+      B3 = _mm_and_si128(B3, B2);                 \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B1 = _mm_or_si128(B1, B2);                  \
+      B1 = _mm_xor_si128(B1, B4);                 \
+      B4 = _mm_and_si128(B4, B3);                 \
+      B2 = _mm_xor_si128(B2, B3);                 \
+      B4 = _mm_and_si128(B4, B0);                 \
+      B4 = _mm_xor_si128(B4, B2);                 \
+      B2 = _mm_and_si128(B2, B1);                 \
+      B2 = _mm_or_si128(B2, B0);                  \
+      B3 = _mm_xor_si128(B3, all_ones);           \
+      B2 = _mm_xor_si128(B2, B3);                 \
+      B0 = _mm_xor_si128(B0, B3);                 \
+      B0 = _mm_and_si128(B0, B1);                 \
+      B3 = _mm_xor_si128(B3, B4);                 \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      B0 = B1;                                    \
+      B1 = B4;                                    \
+      } while(0);
+
+#define SBoxD4(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      __m128i B4 = B2;                            \
+      B2 = _mm_xor_si128(B2, B1);                 \
+      B0 = _mm_xor_si128(B0, B2);                 \
+      B4 = _mm_and_si128(B4, B2);                 \
+      B4 = _mm_xor_si128(B4, B0);                 \
+      B0 = _mm_and_si128(B0, B1);                 \
+      B1 = _mm_xor_si128(B1, B3);                 \
+      B3 = _mm_or_si128(B3, B4);                  \
+      B2 = _mm_xor_si128(B2, B3);                 \
+      B0 = _mm_xor_si128(B0, B3);                 \
+      B1 = _mm_xor_si128(B1, B4);                 \
+      B3 = _mm_and_si128(B3, B2);                 \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B1 = _mm_xor_si128(B1, B0);                 \
+      B1 = _mm_or_si128(B1, B2);                  \
+      B0 = _mm_xor_si128(B0, B3);                 \
+      B1 = _mm_xor_si128(B1, B4);                 \
+      B0 = _mm_xor_si128(B0, B1);                 \
+      B4 = B0;                                    \
+      B0 = B2;                                    \
+      B2 = B3;                                    \
+      B3 = B4;                                    \
+      } while(0);
+
+#define SBoxD5(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      __m128i B4 = B2;                            \
+      B2 = _mm_and_si128(B2, B3);                 \
+      B2 = _mm_xor_si128(B2, B1);                 \
+      B1 = _mm_or_si128(B1, B3);                  \
+      B1 = _mm_and_si128(B1, B0);                 \
+      B4 = _mm_xor_si128(B4, B2);                 \
+      B4 = _mm_xor_si128(B4, B1);                 \
+      B1 = _mm_and_si128(B1, B2);                 \
+      B0 = _mm_xor_si128(B0, all_ones);           \
+      B3 = _mm_xor_si128(B3, B4);                 \
+      B1 = _mm_xor_si128(B1, B3);                 \
+      B3 = _mm_and_si128(B3, B0);                 \
+      B3 = _mm_xor_si128(B3, B2);                 \
+      B0 = _mm_xor_si128(B0, B1);                 \
+      B2 = _mm_and_si128(B2, B0);                 \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      B2 = _mm_xor_si128(B2, B4);                 \
+      B2 = _mm_or_si128(B2, B3);                  \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      B2 = _mm_xor_si128(B2, B1);                 \
+      B1 = B3;                                    \
+      B3 = B4;                                    \
+      } while(0);
+
+#define SBoxD6(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      B1 = _mm_xor_si128(B1, all_ones);           \
+      __m128i B4 = B3;                            \
+      B2 = _mm_xor_si128(B2, B1);                 \
+      B3 = _mm_or_si128(B3, B0);                  \
+      B3 = _mm_xor_si128(B3, B2);                 \
+      B2 = _mm_or_si128(B2, B1);                  \
+      B2 = _mm_and_si128(B2, B0);                 \
+      B4 = _mm_xor_si128(B4, B3);                 \
+      B2 = _mm_xor_si128(B2, B4);                 \
+      B4 = _mm_or_si128(B4, B0);                  \
+      B4 = _mm_xor_si128(B4, B1);                 \
+      B1 = _mm_and_si128(B1, B2);                 \
+      B1 = _mm_xor_si128(B1, B3);                 \
+      B4 = _mm_xor_si128(B4, B2);                 \
+      B3 = _mm_and_si128(B3, B4);                 \
+      B4 = _mm_xor_si128(B4, B1);                 \
+      B3 = _mm_xor_si128(B3, B4);                 \
+      B4 = _mm_xor_si128(B4, all_ones);           \
+      B3 = _mm_xor_si128(B3, B0);                 \
+      B0 = B1;                                    \
+      B1 = B4;                                    \
+      B4 = B3;                                    \
+      B3 = B2;                                    \
+      B2 = B4;                                    \
+      } while(0);
+
+#define SBoxD7(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      B0 = _mm_xor_si128(B0, B2);                 \
+      __m128i B4 = B2;                            \
+      B2 = _mm_and_si128(B2, B0);                 \
+      B4 = _mm_xor_si128(B4, B3);                 \
+      B2 = _mm_xor_si128(B2, all_ones);           \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B2 = _mm_xor_si128(B2, B3);                 \
+      B4 = _mm_or_si128(B4, B0);                  \
+      B0 = _mm_xor_si128(B0, B2);                 \
+      B3 = _mm_xor_si128(B3, B4);                 \
+      B4 = _mm_xor_si128(B4, B1);                 \
+      B1 = _mm_and_si128(B1, B3);                 \
+      B1 = _mm_xor_si128(B1, B0);                 \
+      B0 = _mm_xor_si128(B0, B3);                 \
+      B0 = _mm_or_si128(B0, B2);                  \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B4 = _mm_xor_si128(B4, B0);                 \
+      B0 = B1;                                    \
+      B1 = B2;                                    \
+      B2 = B4;                                    \
+      } while(0);
+
+#define SBoxD8(B0, B1, B2, B3)                    \
+   do                                             \
+      {                                           \
+      __m128i B4 = B2;                            \
+      B2 = _mm_xor_si128(B2, B0);                 \
+      B0 = _mm_and_si128(B0, B3);                 \
+      B4 = _mm_or_si128(B4, B3);                  \
+      B2 = _mm_xor_si128(B2, all_ones);           \
+      B3 = _mm_xor_si128(B3, B1);                 \
+      B1 = _mm_or_si128(B1, B0);                  \
+      B0 = _mm_xor_si128(B0, B2);                 \
+      B2 = _mm_and_si128(B2, B4);                 \
+      B3 = _mm_and_si128(B3, B4);                 \
+      B1 = _mm_xor_si128(B1, B2);                 \
+      B2 = _mm_xor_si128(B2, B0);                 \
+      B0 = _mm_or_si128(B0, B2);                  \
+      B4 = _mm_xor_si128(B4, B1);                 \
+      B0 = _mm_xor_si128(B0, B3);                 \
+      B3 = _mm_xor_si128(B3, B4);                 \
+      B4 = _mm_or_si128(B4, B0);                  \
+      B3 = _mm_xor_si128(B3, B2);                 \
+      B4 = _mm_xor_si128(B4, B2);                 \
+      B2 = B1;                                    \
+      B1 = B0;                                    \
+      B0 = B3;                                    \
+      B3 = B4;                                    \
+      } while(0);
+
 #endif