Use avx512f across entire checkRangeRead call tree

Keeping the avx fallback

ConflictSet.cpp

@@ -1907,9 +1907,9 @@ bool scan16(const InternalVersionT *vs, const uint8_t *is, int begin, int end,
 //
 // always_inline So that we can optimize when begin or end is a constant.
 template <bool kAVX512>
-inline __attribute((always_inline)) bool scan16(const InternalVersionT *vs,
+inline __attribute__((always_inline)) bool
-                                                int begin, int end,
+scan16(const InternalVersionT *vs, int begin, int end,
-                                                InternalVersionT readVersion) {
+       InternalVersionT readVersion) {
   assert(0 <= begin && begin < 16);
   assert(0 <= end && end <= 16);
   assert(begin <= end);
@@ -2512,6 +2512,33 @@ bool checkRangeReadImpl(Node *n, std::span<const uint8_t> begin,
   return checkRangeLeftSide.ok & checkRangeRightSide.ok;
 }
 
+#ifdef __x86_64__
+// Explicitly instantiate with target avx512f attribute so the compiler can
+// inline compare16_32bit_avx512, and generally use avx512f within more
+// functions
+template __attribute__((target("avx512f"))) bool
+scan16<true>(const InternalVersionT *vs, const uint8_t *is, int begin, int end,
+             InternalVersionT readVersion);
+template __attribute__((always_inline, target("avx512f"))) bool
+scan16<true>(const InternalVersionT *vs, int begin, int end,
+             InternalVersionT readVersion);
+template __attribute__((target("avx512f"))) bool
+checkMaxBetweenExclusive<true>(Node *n, int begin, int end,
+                               InternalVersionT readVersion);
+template __attribute__((target("avx512f"))) bool
+checkRangeStartsWith<true>(Node *n, std::span<const uint8_t> key, int begin,
+                           int end, InternalVersionT readVersion,
+                           ConflictSet::Impl *impl);
+template __attribute__((target("avx512f"))) bool
+CheckRangeLeftSide<true>::step();
+template __attribute__((target("avx512f"))) bool
+CheckRangeRightSide<true>::step();
+template __attribute__((target("avx512f"))) bool
+checkRangeReadImpl<true>(Node *n, std::span<const uint8_t> begin,
+                         std::span<const uint8_t> end,
+                         InternalVersionT readVersion, ConflictSet::Impl *impl);
+#endif
+
 #if defined(__SANITIZE_THREAD__) || !defined(__x86_64__)
 bool checkRangeRead(Node *n, std::span<const uint8_t> begin,
                     std::span<const uint8_t> end, InternalVersionT readVersion,