Add explicit AVX2 function annotations

Needed for single amalagamation file with AVX2 enabled.

3 files changed, 30 insertions, 2 deletions

diff --git a/src/lib/block/serpent/serpent.cpp b/src/lib/block/serpent/serpent.cpp
index 7bdda0f9f..77cf9f32b 100644
--- a/src/lib/block/serpent/serpent.cpp
+++ b/src/lib/block/serpent/serpent.cpp
@@ -293,4 +293,6 @@ std::string Serpent::provider() const
    return "base";
    }
 
+#undef key_xor
+
 }
diff --git a/src/lib/block/serpent/serpent_avx2/serpent_avx2.cpp b/src/lib/block/serpent/serpent_avx2/serpent_avx2.cpp
index 92ea909a8..3438440da 100644
--- a/src/lib/block/serpent/serpent_avx2/serpent_avx2.cpp
+++ b/src/lib/block/serpent/serpent_avx2/serpent_avx2.cpp
@@ -50,6 +50,7 @@ namespace Botan {
       B0 = B0.rotr<13>();                                          \
    } while(0)
 
+BOTAN_FUNC_ISA("avx2")
 void Serpent::avx2_encrypt_8(const uint8_t in[128], uint8_t out[128]) const
    {
    SIMD_8x32 B0 = SIMD_8x32::load_le(in);
@@ -99,6 +100,7 @@ void Serpent::avx2_encrypt_8(const uint8_t in[128], uint8_t out[128]) const
    B3.store_le(out + 96);
    }
 
+BOTAN_FUNC_ISA("avx2")
 void Serpent::avx2_decrypt_8(const uint8_t in[128], uint8_t out[128]) const
    {
    SIMD_8x32 B0 = SIMD_8x32::load_le(in);
@@ -152,4 +154,8 @@ void Serpent::avx2_decrypt_8(const uint8_t in[128], uint8_t out[128]) const
    B3.store_le(out + 96);
    }
 
+#undef key_xor
+#undef transform
+#undef i_transform
+
 }
diff --git a/src/lib/utils/simd/simd_avx2/simd_avx2.h b/src/lib/utils/simd/simd_avx2/simd_avx2.h
index 3161af962..6f4b37b24 100644
--- a/src/lib/utils/simd/simd_avx2/simd_avx2.h
+++ b/src/lib/utils/simd/simd_avx2/simd_avx2.h
@@ -22,42 +22,50 @@ class SIMD_8x32 final
       SIMD_8x32& operator=(SIMD_8x32&& other) = default;
       SIMD_8x32(SIMD_8x32&& other) = default;
 
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32()
          {
          m_avx2 = _mm256_setzero_si256();
          }
 
+      BOTAN_FUNC_ISA("avx2")
       explicit SIMD_8x32(const uint32_t B[8])
          {
          m_avx2 = _mm256_loadu_si256(reinterpret_cast<const __m256i*>(B));
          }
 
+      BOTAN_FUNC_ISA("avx2")
       static SIMD_8x32 splat(uint32_t B)
          {
          return SIMD_8x32(_mm256_set1_epi32(B));
          }
 
+      BOTAN_FUNC_ISA("avx2")
       static SIMD_8x32 load_le(const uint8_t* in)
          {
          return SIMD_8x32(_mm256_loadu_si256(reinterpret_cast<const __m256i*>(in)));
          }
 
+      BOTAN_FUNC_ISA("avx2")
       static SIMD_8x32 load_be(const uint8_t* in)
          {
          return load_le(in).bswap();
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void store_le(uint8_t out[]) const
          {
          _mm256_storeu_si256(reinterpret_cast<__m256i*>(out), m_avx2);
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void store_be(uint8_t out[]) const
          {
          bswap().store_le(out);
          }
 
       template<size_t ROT>
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32 rotl() const
          {
          static_assert(ROT > 0 && ROT < 32, "Invalid rotation constant");
@@ -67,6 +75,7 @@ class SIMD_8x32 final
          }
 
       template<size_t ROT>
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32 rotr() const
          {
          return this->rotl<32-ROT>();
@@ -107,52 +116,60 @@ class SIMD_8x32 final
          return retval;
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void operator+=(const SIMD_8x32& other)
          {
          m_avx2 = _mm256_add_epi32(m_avx2, other.m_avx2);
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void operator-=(const SIMD_8x32& other)
          {
          m_avx2 = _mm256_sub_epi32(m_avx2, other.m_avx2);
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void operator^=(const SIMD_8x32& other)
          {
          m_avx2 = _mm256_xor_si256(m_avx2, other.m_avx2);
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void operator|=(const SIMD_8x32& other)
          {
          m_avx2 = _mm256_or_si256(m_avx2, other.m_avx2);
          }
 
+      BOTAN_FUNC_ISA("avx2")
       void operator&=(const SIMD_8x32& other)
          {
          m_avx2 = _mm256_and_si256(m_avx2, other.m_avx2);
          }
 
-      template<int SHIFT> SIMD_8x32 shl() const
+      template<int SHIFT> BOTAN_FUNC_ISA("avx2") SIMD_8x32 shl() const
          {
          return SIMD_8x32(_mm256_slli_epi32(m_avx2, SHIFT));
          }
 
-      template<int SHIFT> SIMD_8x32 shr() const
+      template<int SHIFT> BOTAN_FUNC_ISA("avx2")SIMD_8x32 shr() const
          {
          return SIMD_8x32(_mm256_srli_epi32(m_avx2, SHIFT));
          }
 
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32 operator~() const
          {
          return SIMD_8x32(_mm256_xor_si256(m_avx2, _mm256_set1_epi32(0xFFFFFFFF)));
          }
 
       // (~reg) & other
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32 andc(const SIMD_8x32& other) const
          {
          return SIMD_8x32(_mm256_andnot_si256(m_avx2, other.m_avx2));
          }
 
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32 bswap() const
          {
          const uint8_t BSWAP_MASK[32] = { 3, 2, 1, 0,
@@ -171,6 +188,7 @@ class SIMD_8x32 final
          return SIMD_8x32(output);
          }
 
+      BOTAN_FUNC_ISA("avx2")
       static void transpose(SIMD_8x32& B0, SIMD_8x32& B1,
                             SIMD_8x32& B2, SIMD_8x32& B3)
          {
@@ -186,6 +204,8 @@ class SIMD_8x32 final
          }
 
    private:
+
+      BOTAN_FUNC_ISA("avx2")
       SIMD_8x32(__m256i x) : m_avx2(x) {}
 
       __m256i m_avx2;