SIMD implementation of scan16 for x86

Closes #29

ConflictSet.cpp

@@ -1713,9 +1713,10 @@ downLeftSpine:
 bool scan16(const int64_t *vs, const uint8_t *is, int begin, int end,
             int64_t readVersion) {
 
+  assert(end - begin < 256);
+
 #ifdef HAS_ARM_NEON
 
-  assert(end - begin < 256);
   uint8x16_t indices;
   memcpy(&indices, is, 16);
   // 0xff for each in bounds
@@ -1727,7 +1728,21 @@ bool scan16(const int64_t *vs, const uint8_t *is, int begin, int end,
 
   uint64_t compared = 0;
   for (int i = 0; i < 16; ++i) {
-    compared |= (uint64_t(vs[i] > readVersion) << (i << 2));
+    compared |= uint64_t(vs[i] > readVersion) << (i << 2);
+  }
+  return !(compared & mask);
+
+#elif defined(HAS_AVX)
+
+  __m128i indices;
+  memcpy(&indices, is, 16);
+  indices = _mm_sub_epi8(indices, _mm_set1_epi8(begin));
+  uint32_t mask = ~_mm_movemask_epi8(_mm_cmpeq_epi8(
+      indices, _mm_max_epu8(indices, _mm_set1_epi8(end - begin))));
+
+  uint32_t compared = 0;
+  for (int i = 0; i < 16; ++i) {
+    compared |= (vs[i] > readVersion) << i;
   }
   return !(compared & mask);
 
@@ -1739,7 +1754,7 @@ bool scan16(const int64_t *vs, const uint8_t *is, int begin, int end,
 
   uint32_t compared = 0;
   for (int i = 0; i < 16; ++i) {
-    compared |= ((vs[i] > readVersion) << i);
+    compared |= (vs[i] > readVersion) << i;
   }
   uint32_t mask = 0;
   for (int i = 0; i < 16; ++i) {