Pre-splat SIMON and SPECK keys when appropriate for Altivec (PR #910)

SIMON and SPECK keys can be pre-splatted in the forward direction when Altivec instructions will be used. Pre-splatting does not work for the reverse transformation. It breaks modes like CBC, so the speed-up is only applied to the forward transformation.

1 files changed, 15 insertions, 8 deletions

diff --git a/simon128_simd.cpp b/simon128_simd.cpp
index a91da0f5..b780cf7e 100644
--- a/simon128_simd.cpp
+++ b/simon128_simd.cpp
@@ -66,7 +66,7 @@ using CryptoPP::vec_swap;  // SunCC
 #if defined(_MSC_VER) && !defined(_M_ARM64)

 inline uint64x2_t vld1q_dup_u64(const uint64_t* ptr)

 {

-	return vmovq_n_u64(*ptr);

+    return vmovq_n_u64(*ptr);

 }

 #endif

 

@@ -536,6 +536,7 @@ using CryptoPP::uint64x2_p;
 

 using CryptoPP::VecAnd;

 using CryptoPP::VecXor;

+using CryptoPP::VecLoad;

 using CryptoPP::VecPermute;

 

 // Rotate left by bit count

@@ -576,16 +577,18 @@ inline void SIMON128_Enc_Block(uint32x4_p &block, const word64 *subkeys, unsigne
 

     for (int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)

     {

-        const uint64x2_p rk1 = vec_splats((unsigned long long)subkeys[i]);

-        y1 = VecXor(VecXor(y1, SIMON128_f(x1)), rk1);

+        // Round keys are pre-splated in forward direction

+        const uint64x2_p rk1 = VecLoad(subkeys+i*2);

+        const uint64x2_p rk2 = VecLoad(subkeys+i*2+2);

 

-        const uint64x2_p rk2 = vec_splats((unsigned long long)subkeys[i+1]);

+        y1 = VecXor(VecXor(y1, SIMON128_f(x1)), rk1);

         x1 = VecXor(VecXor(x1, SIMON128_f(y1)), rk2);

     }

 

     if (rounds & 1)

     {

-        const uint64x2_p rk = vec_splats((unsigned long long)subkeys[rounds-1]);

+        // Round keys are pre-splated in forward direction

+        const uint64x2_p rk = VecLoad(subkeys+rounds*2-2);

         y1 = VecXor(VecXor(y1, SIMON128_f(x1)), rk);

         std::swap(x1, y1);

     }

@@ -667,12 +670,14 @@ inline void SIMON128_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 

     for (int i = 0; i < static_cast<int>(rounds & ~1)-1; i += 2)

     {

-        const uint64x2_p rk1 = vec_splats((unsigned long long)subkeys[i]);

+        // Round keys are pre-splated in forward direction

+        const uint64x2_p rk1 = VecLoad(subkeys+i*2);

+        const uint64x2_p rk2 = VecLoad(subkeys+i*2+2);

+

         y1 = VecXor(VecXor(y1, SIMON128_f(x1)), rk1);

         y2 = VecXor(VecXor(y2, SIMON128_f(x2)), rk1);

         y3 = VecXor(VecXor(y3, SIMON128_f(x3)), rk1);

 

-        const uint64x2_p rk2 = vec_splats((unsigned long long)subkeys[i+1]);

         x1 = VecXor(VecXor(x1, SIMON128_f(y1)), rk2);

         x2 = VecXor(VecXor(x2, SIMON128_f(y2)), rk2);

         x3 = VecXor(VecXor(x3, SIMON128_f(y3)), rk2);

@@ -680,7 +685,9 @@ inline void SIMON128_Enc_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
 

     if (rounds & 1)

     {

-        const uint64x2_p rk = vec_splats((unsigned long long)subkeys[rounds-1]);

+        // Round keys are pre-splated in forward direction

+        const uint64x2_p rk = VecLoad(subkeys+rounds*2-2);

+

         y1 = VecXor(VecXor(y1, SIMON128_f(x1)), rk);

         y2 = VecXor(VecXor(y2, SIMON128_f(x2)), rk);

         y3 = VecXor(VecXor(y3, SIMON128_f(x3)), rk);