Improve cpu performance for workload generation in server_bench

ServerBench.cpp

@@ -138,13 +138,17 @@ void tupleAppend(std::string &output, int64_t value) {
 
 void tupleAppend(std::string &output, std::string_view value) {
   output.push_back('\x02');
-  for (auto c : value) {
-    if (c == '\x00') {
-      output.push_back('\x00');
-      output.push_back('\xff');
-    } else {
-      output.push_back(c);
+  if (memchr(value.data(), '\x00', value.size()) != nullptr) {
+    for (auto c : value) {
+      if (c == '\x00') {
+        output.push_back('\x00');
+        output.push_back('\xff');
+      } else {
+        output.push_back(c);
+      }
     }
+  } else {
+    output.insert(output.end(), value.begin(), value.end());
   }
   output.push_back('\x00');
 }
@@ -412,19 +416,18 @@ int main(int argc, char **argv) {
     }
 #endif
 
-    TxQueue<std::unique_ptr<Transaction>> queue{10};
+    TxQueue<Transaction> queue{10};
 
     auto workloadThread = std::thread{[&]() {
       for (int64_t version = kWindowSize;;
            ++version, transactions.fetch_add(1, std::memory_order_relaxed)) {
-        auto tx = std::make_unique<Transaction>(version);
-        queue.push(std::move(tx));
+        queue.push(Transaction(version));
       }
     }};
 
     auto resolverThread = std::thread{[&]() {
       for (;;) {
-        auto tx = queue.pop()->get();
+        auto tx = queue.pop();
         resolver.resolve(tx->reads.data(), tx->reads.size(), tx->writes.data(),
                          tx->writes.size(), tx->version, tx->oldestVersion);
       }