Merge pull request #13451 from vyuuui/numeric_labels

Add numeric label support to assembler

Source/Core/Common/Assembler/GekkoIRGen.cpp

@@ -46,6 +46,7 @@ public:
   void OnLoaddr(std::string_view id) override;
   void OnCloseParen(ParenType type) override;
   void OnLabelDecl(std::string_view name) override;
+  void OnNumericLabelDecl(std::string_view name, u32 num) override;
   void OnVarDecl(std::string_view name) override;
   void PostParseAction() override;
 
@@ -75,6 +76,7 @@ public:
   void AddAbsoluteAddressConv();
   void AddLiteral(u32 lit);
   void AddSymbolResolve(std::string_view sym, bool absolute);
+  void AddNumLabelSymResolve(std::string_view sym, u32 num);
 
   void RunFixups();
 
@@ -97,6 +99,8 @@ private:
   u64* m_active_var;
   size_t m_operand_scan_begin;
 
+  // Ordered top-to-bottom, stores (label number, address)
+  std::vector<std::pair<u32, u32>> m_numlabs;
   std::map<std::string, u32, std::less<>> m_labels;
   std::map<std::string, u64, std::less<>> m_constants;
   std::set<std::string> m_symset;
@@ -362,6 +366,11 @@ void GekkoIRPlugin::OnLabelDecl(std::string_view name)
   m_symset.insert(name_str);
 }
 
+void GekkoIRPlugin::OnNumericLabelDecl(std::string_view, u32 num)
+{
+  m_numlabs.emplace_back(num, m_active_block->BlockEndAddress());
+}
+
 void GekkoIRPlugin::OnVarDecl(std::string_view name)
 {
   const std::string name_str(name);
@@ -555,6 +564,35 @@ void GekkoIRPlugin::AddSymbolResolve(std::string_view sym, bool absolute)
       });
 }
 
+void GekkoIRPlugin::AddNumLabelSymResolve(std::string_view sym, u32 num)
+{
+  const u32 source_address = m_active_block->BlockEndAddress();
+  AssemblerError err_on_fail = AssemblerError{
+      fmt::format("No numeric label '{}' found below here", num),
+      m_owner->lexer.CurrentLine(),
+      m_owner->lexer.LineNumber(),
+      // Lexer should currently point to the label, as it hasn't been eaten yet
+      m_owner->lexer.ColNumber(),
+      sym.size(),
+  };
+
+  // Searching forward only
+  size_t search_start_idx = static_cast<size_t>(m_numlabs.size());
+  m_fixup_stack.emplace(
+      [this, num, source_address, search_start_idx, err_on_fail = std::move(err_on_fail)]() {
+        for (size_t i = search_start_idx; i < m_numlabs.size(); i++)
+        {
+          if (num == m_numlabs[i].first)
+          {
+            return m_numlabs[i].second - source_address;
+          }
+        }
+
+        m_owner->error = std::move(err_on_fail);
+        return u32{0};
+      });
+}
+
 void GekkoIRPlugin::SaveOperandFixup(size_t str_left, size_t str_right)
 {
   m_operand_fixups.emplace_back(std::move(m_fixup_stack.top()));
@@ -715,6 +753,33 @@ void GekkoIRPlugin::EvalTerminalRel(Terminal type, const AssemblerToken& tok)
     break;
   }
 
+  case Terminal::NumLabFwd:
+  {
+    std::optional<u32> val = tok.EvalToken<u32>();
+    ASSERT(val.has_value());
+    AddNumLabelSymResolve(tok.token_val, *val);
+    break;
+  }
+
+  case Terminal::NumLabBwd:
+  {
+    std::optional<u32> mval = tok.EvalToken<u32>();
+    ASSERT(mval.has_value());
+    u32 val = *mval;
+    if (auto label_it = std::find_if(m_numlabs.rbegin(), m_numlabs.rend(),
+                                     [val](std::pair<u32, u32> p) { return p.first == val; });
+        label_it != m_numlabs.rend())
+    {
+      AddLiteral(label_it->second - CurrentAddress());
+    }
+    else
+    {
+      m_owner->EmitErrorHere(fmt::format("No numeric label '{}' found above here", val));
+      return;
+    }
+    break;
+  }
+
   // Parser should disallow this from happening
   default:
     ASSERT(false);
@@ -780,6 +845,31 @@ void GekkoIRPlugin::EvalTerminalAbs(Terminal type, const AssemblerToken& tok)
     break;
   }
 
+  case Terminal::NumLabFwd:
+    m_owner->EmitErrorHere(
+        fmt::format("Forward label references not supported in fully resolved expressons"));
+    break;
+
+  case Terminal::NumLabBwd:
+  {
+    std::optional<u32> mval = tok.EvalToken<u32>();
+    ASSERT(mval.has_value());
+    u32 val = *mval;
+    if (auto label_it = std::find_if(m_numlabs.rbegin(), m_numlabs.rend(),
+                                     [val](std::pair<u32, u32> p) { return p.first == val; });
+        label_it != m_numlabs.rend())
+    {
+      m_eval_stack.push_back(label_it->second);
+    }
+    else
+    {
+      m_owner->EmitErrorHere(fmt::format("No numeric label '{}' found above here", val));
+      return;
+    }
+
+    break;
+  }
+
   // Parser should disallow this from happening
   default:
     ASSERT(false);

Source/Core/Common/Assembler/GekkoLexer.cpp

@@ -201,6 +201,9 @@ std::optional<T> EvalIntegral(TokenType tp, std::string_view val)
     return T{2};
   case TokenType::So:
     return T{3};
+  case TokenType::NumLabFwd:
+  case TokenType::NumLabBwd:
+    return std::accumulate(val.begin(), val.end() - 1, T{0}, dec_step);
   default:
     return std::nullopt;
   }
@@ -723,9 +726,9 @@ AssemblerToken Lexer::LexSingle() const
   }
   else if (h == '0')
   {
-    const char imm_type = Peek();
+    const char nextch = Peek();
 
-    if (imm_type == 'x')
+    if (nextch == 'x')
     {
       token_type = TokenType::HexadecimalLit;
       Step();
@@ -733,21 +736,33 @@ AssemblerToken Lexer::LexSingle() const
       {
       }
     }
-    else if (imm_type == 'b')
+    else if (nextch == 'b')
     {
-      token_type = TokenType::BinaryLit;
       Step();
-      for (char c = Peek(); IsBinary(c); c = Step().Peek())
+      if (!IsBinary(Peek()))
       {
+        token_type = TokenType::NumLabBwd;
+      }
+      else
+      {
+        token_type = TokenType::BinaryLit;
+        for (char c = Peek(); IsBinary(c); c = Step().Peek())
+        {
+        }
       }
     }
-    else if (IsOctal(imm_type))
+    else if (IsOctal(nextch))
     {
       token_type = TokenType::OctalLit;
       for (char c = Peek(); IsOctal(c); c = Step().Peek())
       {
       }
     }
+    else if (nextch == 'f')
+    {
+      Step();
+      token_type = TokenType::NumLabFwd;
+    }
     else
     {
       token_type = TokenType::DecimalLit;
@@ -758,7 +773,22 @@ AssemblerToken Lexer::LexSingle() const
     for (char c = Peek(); std::isdigit(c); c = Step().Peek())
     {
     }
-    token_type = TokenType::DecimalLit;
+    switch (Peek())
+    {
+    case 'f':
+      token_type = TokenType::NumLabFwd;
+      Step();
+      break;
+
+    case 'b':
+      token_type = TokenType::NumLabBwd;
+      Step();
+      break;
+
+    default:
+      token_type = TokenType::DecimalLit;
+      break;
+    }
   }
   else if (h == '<' || h == '>')
   {

Source/Core/Common/Assembler/GekkoLexer.h

@@ -37,6 +37,8 @@ enum class TokenType
   Gt,
   Eq,
   So,
+  NumLabFwd,
+  NumLabBwd,
   // EOL signifies boundaries between instructions, a la ';'
   Eol,
   Eof,

Source/Core/Common/Assembler/GekkoParser.cpp

@@ -84,6 +84,26 @@ void ParseId(ParseState* state)
   }
 }
 
+void ParseNumLocation(ParseState* state)
+{
+  AssemblerToken tok = state->lexer.Lookahead();
+  switch (tok.token_type)
+  {
+  case TokenType::NumLabFwd:
+    state->plugin.OnTerminal(Terminal::NumLabFwd, tok);
+    break;
+
+  case TokenType::NumLabBwd:
+    state->plugin.OnTerminal(Terminal::NumLabBwd, tok);
+    break;
+
+  default:
+    state->EmitErrorHere(fmt::format("Invalid {} with value '{}'", tok.TypeStr(), tok.ValStr()));
+    return;
+  }
+  state->lexer.Eat();
+}
+
 void ParseIdLocation(ParseState* state)
 {
   std::array<AssemblerToken, 3> toks;
@@ -184,6 +204,11 @@ void ParseBaseexpr(ParseState* state)
     ParsePpcBuiltin(state);
     break;
 
+  case TokenType::NumLabFwd:
+  case TokenType::NumLabBwd:
+    ParseNumLocation(state);
+    break;
+
   case TokenType::Dot:
     state->plugin.OnTerminal(Terminal::Dot, state->lexer.Lookahead());
     if (state->error)
@@ -589,6 +614,21 @@ void ParseLabel(ParseState* state)
     }
     state->lexer.EatN<2>();
   }
+
+  if (tokens[0].token_type == TokenType::DecimalLit && tokens[1].token_type == TokenType::Colon)
+  {
+    std::optional<u32> labnum = tokens[0].EvalToken<u32>();
+    if (!labnum)
+    {
+      return;
+    }
+    state->plugin.OnNumericLabelDecl(tokens[0].token_val, *labnum);
+    if (state->error)
+    {
+      return;
+    }
+    state->lexer.EatN<2>();
+  }
 }
 
 void ParseResolvedExpr(ParseState* state)

Source/Core/Common/Assembler/GekkoParser.h

@@ -63,6 +63,8 @@ enum class Terminal
   Eq,
   So,
   Dot,
+  NumLabFwd,
+  NumLabBwd,
 };
 
 enum class ParenType
@@ -111,6 +113,7 @@ public:
   virtual void OnCloseParen(ParenType type) {}
   virtual void OnError() {}
   virtual void OnLabelDecl(std::string_view name) {}
+  virtual void OnNumericLabelDecl(std::string_view name, u32 parse_num) {}
   virtual void OnVarDecl(std::string_view name) {}
 
 protected:

Source/Core/DolphinQt/Debugger/GekkoSyntaxHighlight.cpp

@@ -35,6 +35,8 @@ public:
   {
     switch (type)
     {
+    case Terminal::NumLabFwd:
+    case Terminal::NumLabBwd:
     case Terminal::Id:
       HighlightCurToken(HighlightFormat::Symbol);
       break;
@@ -119,6 +121,13 @@ public:
     m_formatting.emplace_back(len, off, HighlightFormat::Symbol);
   }
 
+  void OnNumericLabelDecl(std::string_view name, u32 parse_num) override
+  {
+    const int len = static_cast<int>(m_owner->lexer.LookaheadRef().token_val.length());
+    const int off = static_cast<int>(m_owner->lexer.ColNumber());
+    m_formatting.emplace_back(len, off, HighlightFormat::Symbol);
+  }
+
   void OnVarDecl(std::string_view name) override { OnLabelDecl(name); }
 
 private:

Source/UnitTests/Common/AssemblerTest.cpp

@@ -2029,7 +2029,7 @@ TEST(Assembler, RangeTest)
   EXPECT_TRUE(!IsFailure(Assemble(uimm_range_3, 0)));
 }
 
-TEST(Assembly, MalformedExpressions)
+TEST(Assembler, MalformedExpressions)
 {
   constexpr char missing_arg[] = "add 0, 1";
   constexpr char missing_paren_0[] = ".4byte (1 + 2), ((3 * 6) + 7";
@@ -2065,7 +2065,7 @@ TEST(Assembly, MalformedExpressions)
 
 // Modified listing of a subroutine, listing generated by IDA
 // Expect bytes are based on disc contents
-TEST(Assembly, RealAssembly)
+TEST(Assembler, RealAssembly)
 {
   constexpr char real_assembly[] = ".locate 0x8046A690\n"
                                    ".defvar back_chain, -0x30\n"
@@ -2223,3 +2223,44 @@ TEST(Assembly, RealAssembly)
     EXPECT_EQ(code_blocks[0].instructions[i], real_expect[i]) << " -> i=" << i;
   }
 }
+
+TEST(Assembler, NumericLabels)
+{
+  constexpr char assembly[] = "0:b 0b\nb 0f\n0:.4byte 0\n0:\n1:.4byte 1b";
+  constexpr u8 expect[] = {0x48, 0x00, 0x00, 0x00, 0x48, 0x00, 0x00, 0x04,
+                           0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0c};
+
+  auto res = Assemble(assembly, 0);
+  ASSERT_TRUE(!IsFailure(res));
+  auto&& code_blocks = GetT(res);
+  ASSERT_EQ(code_blocks.size(), 1);
+
+  ASSERT_EQ(code_blocks[0].instructions.size(), sizeof(expect));
+  for (size_t i = 0; i < code_blocks[0].instructions.size(); i++)
+  {
+    EXPECT_EQ(code_blocks[0].instructions[i], expect[i]) << " -> i=" << i;
+  }
+}
+
+TEST(Assembler, InvalidNumericLabels)
+{
+  constexpr char missing_forward[] = "0:b 0f";
+  constexpr char missing_backward[] = "b 0b\n0:";
+  constexpr char forward_directive[] = ".4byte 0f\n0:";
+  constexpr char missing_backward_directive[] = ".4byte 0b\n0:";
+
+  auto res = Assemble(missing_forward, 0);
+  EXPECT_TRUE(IsFailure(res) && GetFailure(res).message == "No numeric label '0' found below here")
+      << GetFailure(res).message;
+  res = Assemble(missing_backward, 0);
+  EXPECT_TRUE(IsFailure(res) && GetFailure(res).message == "No numeric label '0' found above here")
+      << GetFailure(res).message;
+  res = Assemble(forward_directive, 0);
+  EXPECT_TRUE(IsFailure(res) &&
+              GetFailure(res).message ==
+                  "Forward label references not supported in fully resolved expressons")
+      << GetFailure(res).message;
+  res = Assemble(missing_backward_directive, 0);
+  EXPECT_TRUE(IsFailure(res) && GetFailure(res).message == "No numeric label '0' found above here")
+      << GetFailure(res).message;
+}