Enable using NEON on Clang

Clang doesn't like the way SIMD shifts were implemented, I guess
it fails to inline the constant. Make it a template parameter instead.

4 files changed, 24 insertions, 20 deletions

diff --git a/src/build-data/cc/clang.txt b/src/build-data/cc/clang.txt
index ec4da317d..44dd654f3 100644
--- a/src/build-data/cc/clang.txt
+++ b/src/build-data/cc/clang.txt
@@ -46,6 +46,9 @@ rdrand -> "-mrdrnd"
 rdseed -> "-mrdseed"
 sha    -> "-msha"
 altivec -> "-maltivec"
+
+arm32:neon    -> "-mfpu=neon"
+arm64:neon    -> ""
 </isa_flags>
 
 <mach_opt>
diff --git a/src/lib/block/serpent/serpent_simd/serpent_simd.cpp b/src/lib/block/serpent/serpent_simd/serpent_simd.cpp
index ad4871cf6..8148e321f 100644
--- a/src/lib/block/serpent/serpent_simd/serpent_simd.cpp
+++ b/src/lib/block/serpent/serpent_simd/serpent_simd.cpp
@@ -27,11 +27,11 @@ namespace Botan {
       B0 = B0.rotl<13>();                                          \
       B2 = B2.rotl<3>();                                           \
       B1 ^= B0 ^ B2;                                               \
-      B3 ^= B2 ^ (B0 << 3);                                        \
+      B3 ^= B2 ^ B0.shl<3>();                                      \
       B1 = B1.rotl<1>();                                           \
       B3 = B3.rotl<7>();                                           \
       B0 ^= B1 ^ B3;                                               \
-      B2 ^= B3 ^ (B1 << 7);                                        \
+      B2 ^= B3 ^ B1.shl<7>();                                      \
       B0 = B0.rotl<5>();                                           \
       B2 = B2.rotl<22>();                                          \
    } while(0)
@@ -40,11 +40,11 @@ namespace Botan {
    do {                                                            \
       B2 = B2.rotr<22>();                                          \
       B0 = B0.rotr<5>();                                           \
-      B2 ^= B3 ^ (B1 << 7);                                        \
+      B2 ^= B3 ^ B1.shl<7>();                                      \
       B0 ^= B1 ^ B3;                                               \
       B3 = B3.rotr<7>();                                           \
       B1 = B1.rotr<1>();                                           \
-      B3 ^= B2 ^ (B0 << 3);                                        \
+      B3 ^= B2 ^ B0.shl<3>();                                      \
       B1 ^= B0 ^ B2;                                               \
       B2 = B2.rotr<3>();                                           \
       B0 = B0.rotr<13>();                                          \
diff --git a/src/lib/utils/simd/simd_32.h b/src/lib/utils/simd/simd_32.h
index aca7a8e9a..d9a41c800 100644
--- a/src/lib/utils/simd/simd_32.h
+++ b/src/lib/utils/simd/simd_32.h
@@ -508,37 +508,38 @@ class SIMD_4x32 final
 #endif
          }
 
-      SIMD_4x32 operator<<(size_t shift) const
+
+      template<int SHIFT> SIMD_4x32 shl() const
          {
 #if defined(BOTAN_SIMD_USE_SSE2)
-         return SIMD_4x32(_mm_slli_epi32(m_sse, static_cast<int>(shift)));
+         return SIMD_4x32(_mm_slli_epi32(m_sse, SHIFT));
 
 #elif defined(BOTAN_SIMD_USE_ALTIVEC)
-         const unsigned int s = static_cast<unsigned int>(shift);
+         const unsigned int s = static_cast<unsigned int>(SHIFT);
          return SIMD_4x32(vec_sl(m_vmx, (__vector unsigned int){s, s, s, s}));
 #elif defined(BOTAN_SIMD_USE_NEON)
-         return SIMD_4x32(vshlq_n_u32(m_neon, static_cast<int>(shift)));
+         return SIMD_4x32(vshlq_n_u32(m_neon, SHIFT));
 #else
-         return SIMD_4x32(m_scalar[0] << shift,
-                          m_scalar[1] << shift,
-                          m_scalar[2] << shift,
-                          m_scalar[3] << shift);
+         return SIMD_4x32(m_scalar[0] << SHIFT,
+                          m_scalar[1] << SHIFT,
+                          m_scalar[2] << SHIFT,
+                          m_scalar[3] << SHIFT);
 #endif
          }
 
-      SIMD_4x32 operator>>(size_t shift) const
+      template<int SHIFT> SIMD_4x32 shr() const
          {
 #if defined(BOTAN_SIMD_USE_SSE2)
-         return SIMD_4x32(_mm_srli_epi32(m_sse, static_cast<int>(shift)));
+         return SIMD_4x32(_mm_srli_epi32(m_sse, SHIFT));
 
 #elif defined(BOTAN_SIMD_USE_ALTIVEC)
-         const unsigned int s = static_cast<unsigned int>(shift);
+         const unsigned int s = static_cast<unsigned int>(SHIFT);
          return SIMD_4x32(vec_sr(m_vmx, (__vector unsigned int){s, s, s, s}));
 #elif defined(BOTAN_SIMD_USE_NEON)
-         return SIMD_4x32(vshrq_n_u32(m_neon, static_cast<int>(shift)));
+         return SIMD_4x32(vshrq_n_u32(m_neon, SHIFT));
 #else
-         return SIMD_4x32(m_scalar[0] >> shift, m_scalar[1] >> shift,
-                          m_scalar[2] >> shift, m_scalar[3] >> shift);
+         return SIMD_4x32(m_scalar[0] >> SHIFT, m_scalar[1] >> SHIFT,
+                          m_scalar[2] >> SHIFT, m_scalar[3] >> SHIFT);
 
 #endif
          }
diff --git a/src/tests/test_simd.cpp b/src/tests/test_simd.cpp
index 1f192c6b6..60c46fc67 100644
--- a/src/tests/test_simd.cpp
+++ b/src/tests/test_simd.cpp
@@ -82,10 +82,10 @@ class SIMD_32_Tests final : public Test
          blender = ~blender;
          test_eq(result, "~", blender, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
 
-         blender = blender >> 23;
+         blender = blender.shr<23>();
          test_eq(result, ">>", blender, 0x1FF, 0x1FF, 0x1FF, 0x1FF);
 
-         blender = blender << 27;
+         blender = blender.shl<27>();
          test_eq(result, "<<", blender, 0xF8000000, 0xF8000000, 0xF8000000, 0xF8000000);
 
          blender = ~blender;