[libcxx-commits] [libcxx] [libc++] Vectorize std::find (PR #156431)

[Louis Dionne via libcxx-commits]

================
@@ -44,39 +45,100 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 // generic implementation
 template <class _Iter, class _Sent, class _Tp, class _Proj>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 _Iter
-__find(_Iter __first, _Sent __last, const _Tp& __value, _Proj& __proj) {
+__find_loop(_Iter __first, _Sent __last, const _Tp& __value, _Proj& __proj) {
   for (; __first != __last; ++__first)
     if (std::__invoke(__proj, *__first) == __value)
       break;
   return __first;
 }
 
-// trivially equality comparable implementations
-template <class _Tp,
-          class _Up,
-          class _Proj,
-          __enable_if_t<__is_identity<_Proj>::value && __libcpp_is_trivially_equality_comparable<_Tp, _Up>::value &&
-                            sizeof(_Tp) == 1,
-                        int> = 0>
-_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 _Tp* __find(_Tp* __first, _Tp* __last, const _Up& __value, _Proj&) {
-  if (auto __ret = std::__constexpr_memchr(__first, __value, __last - __first))
-    return __ret;
-  return __last;
+template <class _Iter, class _Sent, class _Tp, class _Proj>
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 _Iter
+__find(_Iter __first, _Sent __last, const _Tp& __value, _Proj& __proj) {
+  return std::__find_loop(std::move(__first), std::move(__last), __value, __proj);
 }
 
-#if _LIBCPP_HAS_WIDE_CHARACTERS
-template <class _Tp,
-          class _Up,
-          class _Proj,
-          __enable_if_t<__is_identity<_Proj>::value && __libcpp_is_trivially_equality_comparable<_Tp, _Up>::value &&
-                            sizeof(_Tp) == sizeof(wchar_t) && _LIBCPP_ALIGNOF(_Tp) >= _LIBCPP_ALIGNOF(wchar_t),
-                        int> = 0>
+#if _LIBCPP_VECTORIZE_ALGORITHMS
+template <class _Tp, class _Up>
+[[__nodiscard__]] _LIBCPP_HIDE_FROM_ABI
+_LIBCPP_CONSTEXPR_SINCE_CXX14 _Tp* __find_vectorized(_Tp* __first, _Tp* __last, _Up __value) {
+  if (!__libcpp_is_constant_evaluated()) {
+    constexpr size_t __unroll_count = 4;
+    constexpr size_t __vec_size     = __native_vector_size<_Tp>;
+    using __vec                     = __simd_vector<_Tp, __vec_size>;
+
+    auto __orig_first = __first;
+
+    auto __values = static_cast<__simd_vector<_Up, __vec_size>>(__value);
----------------
ldionne wrote:

```suggestion
    auto __values = static_cast<__simd_vector<_Up, __vec_size>>(__value); // broadcast the value
```

https://github.com/llvm/llvm-project/pull/156431