[Mlir-commits] [mlir] [llvm] [clang] [llvm] Improve implementation of StringRef::find_last_of and cie (PR #71865)

[Mehdi Amini]

================
@@ -274,6 +278,23 @@ StringRef::size_type StringRef::find_first_not_of(StringRef Chars,
 /// Note: O(size() + Chars.size())
 StringRef::size_type StringRef::find_last_of(StringRef Chars,
                                              size_t From) const {
+#ifdef __SSE2__
+  if (Chars.size() == 2) {
+    __m128i Needle0 = _mm_set1_epi8(Chars[0]);
+    __m128i Needle1 = _mm_set1_epi8(Chars[1]);
+    size_type Sz = std::min(From, Length);
+    do {
+      Sz = Sz < 16 ? 0 : Sz - 16;
+      __m128i Buffer = _mm_loadu_si128((const __m128i *)(Data + Sz));
+      unsigned Mask = _mm_movemask_epi8(_mm_or_si128(
+          _mm_cmpeq_epi8(Buffer, Needle0), _mm_cmpeq_epi8(Buffer, Needle1)));
+      if (Mask != 0) {
+        return Sz + sizeof(Mask) * CHAR_BIT - llvm::countl_zero(Mask);
+      }
+    } while (Sz);
+    return npos;
+  }
+#endif
----------------
joker-eph wrote:

Can this be abstracted or made out-of-line?
I'm wondering about the scalability of HW-specific intrinsics in-line (anticipating for the incoming `#elif defined(ARM64)`...)

https://github.com/llvm/llvm-project/pull/71865