From 7241d48ec4d0a0a59325d7a9208bdf5e076ba002 Mon Sep 17 00:00:00 2001
From: Roland Reichwein <mail@reichwein.it>
Date: Sun, 2 Jan 2022 16:44:47 +0100
Subject: Use [[likely]] and [[unlikely]]

---
 include/unicode.h           | 24 ++++++++++++++++++++++--
 include/unicode/predicate.h |  4 ++++
 include/unicode/utf.h       | 27 +++++++++++++++++++++++----
 3 files changed, 49 insertions(+), 6 deletions(-)

diff --git a/include/unicode.h b/include/unicode.h
index 2bf17f4..34812df 100644
--- a/include/unicode.h
+++ b/include/unicode.h
@@ -144,7 +144,11 @@ namespace unicode {
    if (((uintptr_t)(void*)addr & arch_optimizer::addr_mask) == 0) {
     while (input_distance_bytes(begin, end) >= accu_size) {
      typename arch_optimizer::accu_type data{*addr};
-     if ((data & arch_optimizer::ascii_mask) == arch_optimizer::ascii_value) {
+     if ((data & arch_optimizer::ascii_mask) == arch_optimizer::ascii_value)
+#if __cplusplus >= 202002L
+     [[likely]]
+#endif
+     {
       arch_optimizer::template append(reinterpret_cast<const typename From::value_type*>(addr), result);
       begin += accu_size / sizeof(typename From::value_type);
       ++addr;
@@ -214,7 +218,11 @@ namespace unicode {
  template<typename From, typename To, bool block_mode = true, typename std::enable_if_t<is_utf_8_v<From>, bool> = true>
  inline static void append_accu(std::basic_string<To>& result, uint64_t& accu, int& bytes_in_accu)
  {
-  if (block_mode && bytes_in_accu == 8 && (accu & 0x8080808080808080) == 0) {
+  if (block_mode && bytes_in_accu == 8 && (accu & 0x8080808080808080) == 0)
+#if __cplusplus >= 202002L
+  [[likely]]
+#endif
+  {
    result.append({
                  static_cast<To>(accu & 0x7F),
                  static_cast<To>((accu >> 8) & 0x7F),
@@ -239,6 +247,9 @@ namespace unicode {
    if (is_valid_unicode<11>(value))
     append_utf<11>(result, value);
    else
+#if __cplusplus >= 202002L
+    [[unlikely]]
+#endif
     throw std::invalid_argument("Invalid Unicode character in 2 byte UTF-8 sequence");
   } else if ((block_mode || bytes_in_accu >= 3) && (accu & 0xC0C0F0) == 0x8080E0) { // 3 byte sequence
    char32_t value {static_cast<char32_t>(((accu & 0x0F) << 12) | ((accu >> 2) & 0x0FC0) | ((accu >> 16) & 0x3f))};
@@ -247,6 +258,9 @@ namespace unicode {
    if (is_valid_unicode<16>(value))
     append_utf<16>(result, value);
    else
+#if __cplusplus >= 202002L
+    [[unlikely]]
+#endif
     throw std::invalid_argument("Invalid Unicode character in 3 byte UTF-8 sequence");
   } else if ((block_mode || bytes_in_accu >= 4) && (accu & 0xC0C0C0F8) == 0x808080F0) { // 4 byte sequence
    char32_t value {static_cast<char32_t>(((accu & 0x07) << 18) | ((accu << 4) & 0x3f000) | ((accu >> 10) & 0xFC0) | ((accu >> 24) & 0x3f))};
@@ -255,8 +269,14 @@ namespace unicode {
    if (is_valid_unicode<21>(value))
     append_utf(result, value);
    else
+#if __cplusplus >= 202002L
+    [[unlikely]]
+#endif
     throw std::invalid_argument("Invalid Unicode character in 4 byte UTF-8 sequence");
   } else
+#if __cplusplus >= 202002L
+   [[unlikely]]
+#endif
    throw std::invalid_argument("Invalid UTF-8 byte sequence");
  }
 
diff --git a/include/unicode/predicate.h b/include/unicode/predicate.h
index 5f8c6a4..e326529 100644
--- a/include/unicode/predicate.h
+++ b/include/unicode/predicate.h
@@ -1,3 +1,7 @@
+//
+// Unicode library - predicates for Unicode characters
+//
+
 #pragma once
 
 namespace unicode {
diff --git a/include/unicode/utf.h b/include/unicode/utf.h
index 046d9c6..6d7c2f7 100644
--- a/include/unicode/utf.h
+++ b/include/unicode/utf.h
@@ -58,9 +58,11 @@ namespace unicode::detail {
     if ((((s[i] & 7) << 2) | ((s[i + 1] >> 4) & 3)) >= 0x11)
      return false; // Unicode too big above 0x10FFFF
     i += 4;
-   } else {
+   } else
+#if __cplusplus >= 202002L
+   [[unlikely]]
+#endif
     return false;
-   }
   }
   return true;
  }
@@ -91,9 +93,11 @@ namespace unicode::detail {
     i++;
    } else if ((i < size - 1) && is_utf16_sequence(s[i], s[i + 1])) {
     i += 2;
-   } else {
+   } else
+#if __cplusplus >= 202002L
+    [[unlikely]]
+#endif
     return false;
-   }
   }
   return true;
  }
@@ -185,6 +189,9 @@ namespace unicode::detail {
 
    if constexpr(sequence_length > 1) {
     if (remaining_code_units() < sequence_length)
+#if __cplusplus >= 202002L
+     [[unlikely]]
+#endif
      throw std::invalid_argument("Bad input: Not enough bytes left for decoding UTF-8 sequence");
    }
 
@@ -192,6 +199,9 @@ namespace unicode::detail {
     std::advance(iterator, sequence_length);
     internal_type result{decode_utf8_sequence(bytes...)};
     if (!unicode::is_valid_unicode<sequence_length * 6>(result))
+#if __cplusplus >= 202002L
+     [[unlikely]]
+#endif
      throw std::invalid_argument("Invalid Unicode character: "s + std::to_string(static_cast<uint32_t>(result)));
     return result;
    } else {
@@ -218,10 +228,16 @@ namespace unicode::detail {
     return unit0;
    } else {
     if (remaining_code_units() < 2)
+#if __cplusplus >= 202002L
+     [[unlikely]]
+#endif
      throw std::invalid_argument("Bad input: Continuation of first UTF-16 unit missing");
 
     char16_t unit1 {static_cast<char16_t>(get_code_unit<1>())};
     if ((unit0 & 0xFC00) != 0xD800 || (unit1 & 0xFC00) != 0xDC00)
+#if __cplusplus >= 202002L
+     [[unlikely]]
+#endif
      throw std::invalid_argument("Bad input: 2 malformed UTF-16 surrogates");
 
     std::advance(iterator, 2);
@@ -235,6 +251,9 @@ namespace unicode::detail {
    internal_type result {static_cast<internal_type>(get_code_unit<0>())};
 
    if (!unicode::is_valid_unicode(result))
+#if __cplusplus >= 202002L
+    [[unlikely]]
+#endif
     throw std::invalid_argument("Invalid Unicode character: "s + std::to_string(static_cast<uint32_t>(result)));
 
    std::advance(iterator, 1);
-- 
cgit v1.2.3