From f1d3a0eb43d99462618fcc5c022f922631d7a587 Mon Sep 17 00:00:00 2001
From: Nikita Skovoroda <chalkerx@gmail.com>
Date: Sat, 20 Dec 2025 03:27:28 +0400
Subject: [PATCH] src: improve StringBytes::Encode perf on UTF8

---
 src/encoding_binding.cc |  2 ++
 src/string_bytes.cc     | 10 ++++++++++
 2 files changed, 12 insertions(+)

diff --git a/src/encoding_binding.cc b/src/encoding_binding.cc
index 1506442f704ee1..7683d205aa6a3e 100644
--- a/src/encoding_binding.cc
+++ b/src/encoding_binding.cc
@@ -379,6 +379,8 @@ void BindingData::DecodeUTF8(const FunctionCallbackInfo<Value>& args) {
       return node::THROW_ERR_ENCODING_INVALID_ENCODED_DATA(
           env->isolate(), "The encoded data was not valid for encoding utf-8");
     }
+
+    // TODO(chalker): save on utf8 validity recheck in StringBytes::Encode()
   }
 
   if (length == 0) return args.GetReturnValue().SetEmptyString();
diff --git a/src/string_bytes.cc b/src/string_bytes.cc
index 8f6bedd63e25b3..bbf0595a2d848c 100644
--- a/src/string_bytes.cc
+++ b/src/string_bytes.cc
@@ -539,6 +539,16 @@ MaybeLocal<Value> StringBytes::Encode(Isolate* isolate,
         return ExternOneByteString::NewFromCopy(isolate, buf, buflen);
       }
 
+      if (simdutf::validate_utf8(buf, buflen)) {
+        // We know that we are non-ASCII (and are unlikely Latin1), use 2-byte
+        // In the most likely case of valid UTF-8, we can use this fast impl
+        size_t u16size = simdutf::utf16_length_from_utf8(buf, buflen);
+        uint16_t* dst = node::UncheckedMalloc<uint16_t>(u16size);
+        size_t utf16len = simdutf::convert_valid_utf8_to_utf16(
+            buf, buflen, reinterpret_cast<char16_t*>(dst));
+        return ExternTwoByteString::New(isolate, dst, utf16len);
+      }
+
       val =
           String::NewFromUtf8(isolate, buf, v8::NewStringType::kNormal, buflen);
       Local<String> str;