4x interleaved SSE2

1 files changed, 225 insertions, 67 deletions

diff --git a/src/lib/stream/chacha/chacha_sse2/chacha_sse2.cpp b/src/lib/stream/chacha/chacha_sse2/chacha_sse2.cpp
index 34376d84c..8c00ce133 100644
--- a/src/lib/stream/chacha/chacha_sse2/chacha_sse2.cpp
+++ b/src/lib/stream/chacha/chacha_sse2/chacha_sse2.cpp
@@ -16,81 +16,239 @@ void ChaCha::chacha_sse2_x4(byte output[64], u32bit input[16], size_t rounds)
    BOTAN_ASSERT(rounds % 2 == 0, "Valid rounds");
 
    const __m128i* input_mm = reinterpret_cast<const __m128i*>(input);
+   __m128i* output_mm = reinterpret_cast<__m128i*>(output);
 
-   const __m128i input0 = _mm_loadu_si128(input_mm);
-   const __m128i input1 = _mm_loadu_si128(input_mm + 1);
-   const __m128i input2 = _mm_loadu_si128(input_mm + 2);
-   const __m128i input3 = _mm_loadu_si128(input_mm + 3);
-   // TODO: interleave!
+   __m128i input0 = _mm_loadu_si128(input_mm);
+   __m128i input1 = _mm_loadu_si128(input_mm + 1);
+   __m128i input2 = _mm_loadu_si128(input_mm + 2);
+   __m128i input3 = _mm_loadu_si128(input_mm + 3);
+
+   // TODO: try transposing, which would avoid the permutations each round
 
 #define mm_rotl(r, n) \
    _mm_or_si128(_mm_slli_epi32(r, n), _mm_srli_epi32(r, 32-n))
 
-   for(size_t i = 0; i != 4; ++i)
+   __m128i r0_0 = input0;
+   __m128i r0_1 = input1;
+   __m128i r0_2 = input2;
+   __m128i r0_3 = input3;
+
+   __m128i r1_0 = input0;
+   __m128i r1_1 = input1;
+   __m128i r1_2 = input2;
+   __m128i r1_3 = input3;
+   r1_3 = _mm_add_epi64(r0_3, _mm_set_epi64x(0, 1));
+
+   __m128i r2_0 = input0;
+   __m128i r2_1 = input1;
+   __m128i r2_2 = input2;
+   __m128i r2_3 = input3;
+   r2_3 = _mm_add_epi64(r0_3, _mm_set_epi64x(0, 2));
+
+   __m128i r3_0 = input0;
+   __m128i r3_1 = input1;
+   __m128i r3_2 = input2;
+   __m128i r3_3 = input3;
+   r3_3 = _mm_add_epi64(r0_3, _mm_set_epi64x(0, 3));
+
+   for(size_t r = 0; r != rounds / 2; ++r)
       {
-      __m128i r0 = input0;
-      __m128i r1 = input1;
-      __m128i r2 = input2;
-      __m128i r3 = input3;
-
-      r3 = _mm_add_epi64(r3, _mm_set_epi64x(0, i));
-
-      for(size_t r = 0; r != rounds / 2; ++r)
-         {
-         r0 = _mm_add_epi32(r0, r1);
-         r3 = _mm_xor_si128(r3, r0);
-         r3 = mm_rotl(r3, 16);
-
-         r2 = _mm_add_epi32(r2, r3);
-         r1 = _mm_xor_si128(r1, r2);
-         r1 = mm_rotl(r1, 12);
-
-         r0 = _mm_add_epi32(r0, r1);
-         r3 = _mm_xor_si128(r3, r0);
-         r3 = mm_rotl(r3, 8);
-
-         r2 = _mm_add_epi32(r2, r3);
-         r1 = _mm_xor_si128(r1, r2);
-         r1 = mm_rotl(r1, 7);
-
-         r1 = _mm_shuffle_epi32(r1, _MM_SHUFFLE(0, 3, 2, 1));
-         r2 = _mm_shuffle_epi32(r2, _MM_SHUFFLE(1, 0, 3, 2));
-         r3 = _mm_shuffle_epi32(r3, _MM_SHUFFLE(2, 1, 0, 3));
-
-         r0 = _mm_add_epi32(r0, r1);
-         r3 = _mm_xor_si128(r3, r0);
-         r3 = mm_rotl(r3, 16);
-
-         r2 = _mm_add_epi32(r2, r3);
-         r1 = _mm_xor_si128(r1, r2);
-         r1 = mm_rotl(r1, 12);
-
-         r0 = _mm_add_epi32(r0, r1);
-         r3 = _mm_xor_si128(r3, r0);
-         r3 = mm_rotl(r3, 8);
-
-         r2 = _mm_add_epi32(r2, r3);
-         r1 = _mm_xor_si128(r1, r2);
-         r1 = mm_rotl(r1, 7);
-
-         r1 = _mm_shuffle_epi32(r1, _MM_SHUFFLE(2, 1, 0, 3));
-         r2 = _mm_shuffle_epi32(r2, _MM_SHUFFLE(1, 0, 3, 2));
-         r3 = _mm_shuffle_epi32(r3, _MM_SHUFFLE(0, 3, 2, 1));
-         }
-
-      r0 = _mm_add_epi32(r0, input0);
-      r1 = _mm_add_epi32(r1, input1);
-      r2 = _mm_add_epi32(r2, input2);
-      r3 = _mm_add_epi32(r3, input3);
-      r3 = _mm_add_epi64(r3, _mm_set_epi64x(0, i));
-
-      __m128i* output_mm = reinterpret_cast<__m128i*>(output);
-      _mm_storeu_si128(output_mm + 4*i    , r0);
-      _mm_storeu_si128(output_mm + 4*i + 1, r1);
-      _mm_storeu_si128(output_mm + 4*i + 2, r2);
-      _mm_storeu_si128(output_mm + 4*i + 3, r3);
+      r0_0 = _mm_add_epi32(r0_0, r0_1);
+      r1_0 = _mm_add_epi32(r1_0, r1_1);
+      r2_0 = _mm_add_epi32(r2_0, r2_1);
+      r3_0 = _mm_add_epi32(r3_0, r3_1);
+
+      r0_3 = _mm_xor_si128(r0_3, r0_0);
+      r1_3 = _mm_xor_si128(r1_3, r1_0);
+      r2_3 = _mm_xor_si128(r2_3, r2_0);
+      r3_3 = _mm_xor_si128(r3_3, r3_0);
+
+      r0_3 = mm_rotl(r0_3, 16);
+      r1_3 = mm_rotl(r1_3, 16);
+      r2_3 = mm_rotl(r2_3, 16);
+      r3_3 = mm_rotl(r3_3, 16);
+
+      r0_2 = _mm_add_epi32(r0_2, r0_3);
+      r1_2 = _mm_add_epi32(r1_2, r1_3);
+      r2_2 = _mm_add_epi32(r2_2, r2_3);
+      r3_2 = _mm_add_epi32(r3_2, r3_3);
+
+      r0_1 = _mm_xor_si128(r0_1, r0_2);
+      r1_1 = _mm_xor_si128(r1_1, r1_2);
+      r2_1 = _mm_xor_si128(r2_1, r2_2);
+      r3_1 = _mm_xor_si128(r3_1, r3_2);
+
+      r0_1 = mm_rotl(r0_1, 12);
+      r1_1 = mm_rotl(r1_1, 12);
+      r2_1 = mm_rotl(r2_1, 12);
+      r3_1 = mm_rotl(r3_1, 12);
+
+      r0_0 = _mm_add_epi32(r0_0, r0_1);
+      r1_0 = _mm_add_epi32(r1_0, r1_1);
+      r2_0 = _mm_add_epi32(r2_0, r2_1);
+      r3_0 = _mm_add_epi32(r3_0, r3_1);
+
+      r0_3 = _mm_xor_si128(r0_3, r0_0);
+      r1_3 = _mm_xor_si128(r1_3, r1_0);
+      r2_3 = _mm_xor_si128(r2_3, r2_0);
+      r3_3 = _mm_xor_si128(r3_3, r3_0);
+
+      r0_3 = mm_rotl(r0_3, 8);
+      r1_3 = mm_rotl(r1_3, 8);
+      r2_3 = mm_rotl(r2_3, 8);
+      r3_3 = mm_rotl(r3_3, 8);
+
+      r0_2 = _mm_add_epi32(r0_2, r0_3);
+      r1_2 = _mm_add_epi32(r1_2, r1_3);
+      r2_2 = _mm_add_epi32(r2_2, r2_3);
+      r3_2 = _mm_add_epi32(r3_2, r3_3);
+
+      r0_1 = _mm_xor_si128(r0_1, r0_2);
+      r1_1 = _mm_xor_si128(r1_1, r1_2);
+      r2_1 = _mm_xor_si128(r2_1, r2_2);
+      r3_1 = _mm_xor_si128(r3_1, r3_2);
+
+      r0_1 = mm_rotl(r0_1, 7);
+      r1_1 = mm_rotl(r1_1, 7);
+      r2_1 = mm_rotl(r2_1, 7);
+      r3_1 = mm_rotl(r3_1, 7);
+
+      r0_1 = _mm_shuffle_epi32(r0_1, _MM_SHUFFLE(0, 3, 2, 1));
+      r0_2 = _mm_shuffle_epi32(r0_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r0_3 = _mm_shuffle_epi32(r0_3, _MM_SHUFFLE(2, 1, 0, 3));
+
+      r1_1 = _mm_shuffle_epi32(r1_1, _MM_SHUFFLE(0, 3, 2, 1));
+      r1_2 = _mm_shuffle_epi32(r1_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r1_3 = _mm_shuffle_epi32(r1_3, _MM_SHUFFLE(2, 1, 0, 3));
+
+      r2_1 = _mm_shuffle_epi32(r2_1, _MM_SHUFFLE(0, 3, 2, 1));
+      r2_2 = _mm_shuffle_epi32(r2_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r2_3 = _mm_shuffle_epi32(r2_3, _MM_SHUFFLE(2, 1, 0, 3));
+
+      r3_1 = _mm_shuffle_epi32(r3_1, _MM_SHUFFLE(0, 3, 2, 1));
+      r3_2 = _mm_shuffle_epi32(r3_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r3_3 = _mm_shuffle_epi32(r3_3, _MM_SHUFFLE(2, 1, 0, 3));
+
+      r0_0 = _mm_add_epi32(r0_0, r0_1);
+      r1_0 = _mm_add_epi32(r1_0, r1_1);
+      r2_0 = _mm_add_epi32(r2_0, r2_1);
+      r3_0 = _mm_add_epi32(r3_0, r3_1);
+
+      r0_3 = _mm_xor_si128(r0_3, r0_0);
+      r1_3 = _mm_xor_si128(r1_3, r1_0);
+      r2_3 = _mm_xor_si128(r2_3, r2_0);
+      r3_3 = _mm_xor_si128(r3_3, r3_0);
+
+      r0_3 = mm_rotl(r0_3, 16);
+      r1_3 = mm_rotl(r1_3, 16);
+      r2_3 = mm_rotl(r2_3, 16);
+      r3_3 = mm_rotl(r3_3, 16);
+
+      r0_2 = _mm_add_epi32(r0_2, r0_3);
+      r1_2 = _mm_add_epi32(r1_2, r1_3);
+      r2_2 = _mm_add_epi32(r2_2, r2_3);
+      r3_2 = _mm_add_epi32(r3_2, r3_3);
+
+      r0_1 = _mm_xor_si128(r0_1, r0_2);
+      r1_1 = _mm_xor_si128(r1_1, r1_2);
+      r2_1 = _mm_xor_si128(r2_1, r2_2);
+      r3_1 = _mm_xor_si128(r3_1, r3_2);
+
+      r0_1 = mm_rotl(r0_1, 12);
+      r1_1 = mm_rotl(r1_1, 12);
+      r2_1 = mm_rotl(r2_1, 12);
+      r3_1 = mm_rotl(r3_1, 12);
+
+      r0_0 = _mm_add_epi32(r0_0, r0_1);
+      r1_0 = _mm_add_epi32(r1_0, r1_1);
+      r2_0 = _mm_add_epi32(r2_0, r2_1);
+      r3_0 = _mm_add_epi32(r3_0, r3_1);
+
+      r0_3 = _mm_xor_si128(r0_3, r0_0);
+      r1_3 = _mm_xor_si128(r1_3, r1_0);
+      r2_3 = _mm_xor_si128(r2_3, r2_0);
+      r3_3 = _mm_xor_si128(r3_3, r3_0);
+
+      r0_3 = mm_rotl(r0_3, 8);
+      r1_3 = mm_rotl(r1_3, 8);
+      r2_3 = mm_rotl(r2_3, 8);
+      r3_3 = mm_rotl(r3_3, 8);
+
+      r0_2 = _mm_add_epi32(r0_2, r0_3);
+      r1_2 = _mm_add_epi32(r1_2, r1_3);
+      r2_2 = _mm_add_epi32(r2_2, r2_3);
+      r3_2 = _mm_add_epi32(r3_2, r3_3);
+
+      r0_1 = _mm_xor_si128(r0_1, r0_2);
+      r1_1 = _mm_xor_si128(r1_1, r1_2);
+      r2_1 = _mm_xor_si128(r2_1, r2_2);
+      r3_1 = _mm_xor_si128(r3_1, r3_2);
+
+      r0_1 = mm_rotl(r0_1, 7);
+      r1_1 = mm_rotl(r1_1, 7);
+      r2_1 = mm_rotl(r2_1, 7);
+      r3_1 = mm_rotl(r3_1, 7);
+
+      r0_1 = _mm_shuffle_epi32(r0_1, _MM_SHUFFLE(2, 1, 0, 3));
+      r0_2 = _mm_shuffle_epi32(r0_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r0_3 = _mm_shuffle_epi32(r0_3, _MM_SHUFFLE(0, 3, 2, 1));
+
+      r1_1 = _mm_shuffle_epi32(r1_1, _MM_SHUFFLE(2, 1, 0, 3));
+      r1_2 = _mm_shuffle_epi32(r1_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r1_3 = _mm_shuffle_epi32(r1_3, _MM_SHUFFLE(0, 3, 2, 1));
+
+      r2_1 = _mm_shuffle_epi32(r2_1, _MM_SHUFFLE(2, 1, 0, 3));
+      r2_2 = _mm_shuffle_epi32(r2_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r2_3 = _mm_shuffle_epi32(r2_3, _MM_SHUFFLE(0, 3, 2, 1));
+
+      r3_1 = _mm_shuffle_epi32(r3_1, _MM_SHUFFLE(2, 1, 0, 3));
+      r3_2 = _mm_shuffle_epi32(r3_2, _MM_SHUFFLE(1, 0, 3, 2));
+      r3_3 = _mm_shuffle_epi32(r3_3, _MM_SHUFFLE(0, 3, 2, 1));
       }
 
+   r0_0 = _mm_add_epi32(r0_0, input0);
+   r0_1 = _mm_add_epi32(r0_1, input1);
+   r0_2 = _mm_add_epi32(r0_2, input2);
+   r0_3 = _mm_add_epi32(r0_3, input3);
+
+   r1_0 = _mm_add_epi32(r1_0, input0);
+   r1_1 = _mm_add_epi32(r1_1, input1);
+   r1_2 = _mm_add_epi32(r1_2, input2);
+   r1_3 = _mm_add_epi32(r1_3, input3);
+   r1_3 = _mm_add_epi64(r1_3, _mm_set_epi64x(0, 1));
+
+   r2_0 = _mm_add_epi32(r2_0, input0);
+   r2_1 = _mm_add_epi32(r2_1, input1);
+   r2_2 = _mm_add_epi32(r2_2, input2);
+   r2_3 = _mm_add_epi32(r2_3, input3);
+   r2_3 = _mm_add_epi64(r2_3, _mm_set_epi64x(0, 2));
+
+   r3_0 = _mm_add_epi32(r3_0, input0);
+   r3_1 = _mm_add_epi32(r3_1, input1);
+   r3_2 = _mm_add_epi32(r3_2, input2);
+   r3_3 = _mm_add_epi32(r3_3, input3);
+   r3_3 = _mm_add_epi64(r3_3, _mm_set_epi64x(0, 3));
+
+   _mm_storeu_si128(output_mm + 0, r0_0);
+   _mm_storeu_si128(output_mm + 1, r0_1);
+   _mm_storeu_si128(output_mm + 2, r0_2);
+   _mm_storeu_si128(output_mm + 3, r0_3);
+
+   _mm_storeu_si128(output_mm + 4    , r1_0);
+   _mm_storeu_si128(output_mm + 5, r1_1);
+   _mm_storeu_si128(output_mm + 6, r1_2);
+   _mm_storeu_si128(output_mm + 7, r1_3);
+
+   _mm_storeu_si128(output_mm + 8, r2_0);
+   _mm_storeu_si128(output_mm + 9, r2_1);
+   _mm_storeu_si128(output_mm + 10, r2_2);
+   _mm_storeu_si128(output_mm + 11, r2_3);
+
+   _mm_storeu_si128(output_mm + 12    , r3_0);
+   _mm_storeu_si128(output_mm + 13, r3_1);
+   _mm_storeu_si128(output_mm + 14, r3_2);
+   _mm_storeu_si128(output_mm + 15, r3_3);
+
 #undef mm_rotl
 
    }