First program: Can add 2 integers and return result via exit code

5 files changed, 108 insertions, 14 deletions

diff --git a/asm/encode.cpp b/asm/encode.cpp
index ea50cb7..8bf33c0 100644
--- a/asm/encode.cpp
+++ b/asm/encode.cpp
@@ -1,6 +1,68 @@
+// Intel specific conversion: Abstract Graph -> Machine specific segment
 #include "encode.h"
 
+#include "asm/assembler.h"
+#include "minicc.h"
+
+#include <boost/endian/conversion.hpp>
+
+#include <exception>
+
 void Asm::toMachineCode(const FlowGraph::Graph& graph, Segment& segment)
 {
+ segment.clear();
+
+ for (const std::shared_ptr<FlowGraph::Node>& node: graph) {
+  try {
+   FlowGraph::BinaryOperation& op {dynamic_cast<FlowGraph::BinaryOperation&>(*node)};
+
+   auto operands {op.operands()};
+   // TODO: ignore destination (0) for now
+
+   if (operands[1].type() != FlowGraph::DataType::Int) {
+    std::runtime_error("Bad type for operand 1: "s + std::to_string(int(operands[1].type())));
+   }
+
+   if (operands[2].type() != FlowGraph::DataType::Int) {
+    std::runtime_error("Bad type for operand 2: "s + std::to_string(int(operands[2].type())));
+   }
+
+   if (!operands[1].storage())
+    throw std::runtime_error("ICE: Operand 1 storage is 0");
+   if (!operands[2].storage())
+    throw std::runtime_error("ICE: Operand 2 storage is 0");
+
+   uint32_t immediate1{};
+   try {
+    FlowGraph::Constant& value1 {dynamic_cast<FlowGraph::Constant&>(*operands[1].storage())};
+    if (value1.value().size() < sizeof(uint32_t))
+     throw std::runtime_error("ICE: Int data from operand 1 needs at least 4 bytes, got "s + std::to_string(value1.value().size()));
+
+    immediate1 = boost::endian::little_to_native(*(reinterpret_cast<const uint32_t*>(value1.value().data())));
+   } catch (const std::bad_cast& ex) {
+    std::runtime_error("Bad value for operand 1: Constant expected");
+   }
+   
+   uint32_t immediate2{};
+   try {
+    FlowGraph::Constant& value2 {dynamic_cast<FlowGraph::Constant&>(*operands[2].storage())};
+    if (value2.value().size() < sizeof(uint32_t))
+     throw std::runtime_error("ICE: Int data from operand 2 needs at least 4 bytes, got "s + std::to_string(value2.value().size()));
+
+    immediate2 = boost::endian::little_to_native(*(reinterpret_cast<const uint32_t*>(value2.value().data())));
+   } catch (const std::bad_cast& ex) {
+    std::runtime_error("Bad value for operand 2: Constant expected");
+   }
+
+   Asm::Args args1{{Asm::Args::Register32("edi"), Asm::Args::Immediate32(immediate1)}};
+   segment.push_back(makeOp("mov", args1));
+   
+   Asm::Args args2{{Asm::Args::Register32("edi"), Asm::Args::Immediate32(immediate2)}};
+   segment.push_back(makeOp("add", args2));
+
+  } catch (const std::bad_cast& ex) {
+   std::runtime_error("ICE: Encoding: Unsupported node: "s + ex.what());
+  }
+ }
 }
 
diff --git a/asm/intel64/add.cpp b/asm/intel64/add.cpp
index 106ffec..4438895 100644
--- a/asm/intel64/add.cpp
+++ b/asm/intel64/add.cpp
@@ -12,8 +12,12 @@ Op_add::Op_add(Asm::Args& args)
  if (args[0].type() == typeid(Asm::Args::Register32) &&
      std::any_cast<Asm::Args::Register32>(args[0]).name() == "eax" &&
      args[1].type() == typeid(Asm::Args::Immediate32))
- { // add eax, imm32
+ { // add eax, imm32 (before "add reg32, imm32"! It's shorter.)
   machine_code = std::vector<uint8_t>{ 0x05 } + std::any_cast<Asm::Args::Immediate32>(args[1]).getCode();
+ } else if (args[0].type() == typeid(Asm::Args::Register32) &&
+     args[1].type() == typeid(Asm::Args::Immediate32))
+ { // add reg32, imm32
+  machine_code = std::vector<uint8_t>{ 0x81 } + ModRM("/0", std::any_cast<Asm::Args::Register32>(args[0]).name()) + std::any_cast<Asm::Args::Immediate32>(args[1]).getCode();
  } else if (args[0].type() == typeid(Asm::Args::Register64) &&
             std::any_cast<Asm::Args::Register64>(args[0]).name() == "rax"  &&
             args[1].type() == typeid(Asm::Args::Immediate32))
diff --git a/asm/intel64/codes.cpp b/asm/intel64/codes.cpp
index 66a08dd..5d93a57 100644
--- a/asm/intel64/codes.cpp
+++ b/asm/intel64/codes.cpp
@@ -29,14 +29,14 @@ namespace {
   {"dl", 2}, {"dh", 6},
 
   {"ax", 0}, {"sp", 4},
-  {"bx", 3}, {"bp", 7},
-  {"cx", 1}, {"si", 5},
-  {"dx", 2}, {"di", 6},
+  {"bx", 3}, {"bp", 5},
+  {"cx", 1}, {"si", 6},
+  {"dx", 2}, {"di", 7},
 
   {"eax", 0}, {"esp", 4},
-  {"ebx", 3}, {"ebp", 7},
-  {"ecx", 1}, {"esi", 5},
-  {"edx", 2}, {"edi", 6},
+  {"ebx", 3}, {"ebp", 5},
+  {"ecx", 1}, {"esi", 6},
+  {"edx", 2}, {"edi", 7},
  };
 
 }
@@ -44,15 +44,26 @@ namespace {
 // Manual, page 530
 // Reg + Reg/Memory
 uint8_t ModRM(const std::string& reg, const std::string& rm) {
- // TODO: extend
  uint8_t result{0b11000000};
 
- auto index1{ IndexOfRegister.find(reg) };
- if (index1 == IndexOfRegister.end())
-  throw std::runtime_error("Unknown register for arg1: "s + reg);
-
- result |= (index1->second << 3);
-
+ size_t val_reg{};
+ // reg
+ if (reg.size() > 0 && reg[0] == '/') { // "/digit"
+  try {
+   val_reg = stoull(reg.substr(1));
+  } catch (const std::exception& ex) {
+   throw std::runtime_error("ModRM: Bad digit in arg1: "s + reg);
+  }
+ } else { // reg
+  auto index1{ IndexOfRegister.find(reg) };
+  if (index1 == IndexOfRegister.end())
+   throw std::runtime_error("ModRM: Unknown register for arg1: "s + reg);
+  val_reg = index1->second;
+ }
+
+ result |= (val_reg << 3);
+
+ // rm
  auto index2{ IndexOfRegister.find(rm) };
  if (index2 == IndexOfRegister.end())
   throw std::runtime_error("Unknown register for arg2: "s + rm);
@@ -62,6 +73,15 @@ uint8_t ModRM(const std::string& reg, const std::string& rm) {
  return result;
 }
 
+uint8_t RegNo(const std::string& reg)
+{
+ auto index{ IndexOfRegister.find(reg) };
+ if (index == IndexOfRegister.end())
+  throw std::runtime_error("Reg: Unknown register for arg: "s + reg);
+
+ return index->second;
+}
+
 #if 0
  prefixes{
   "lock", 0xf0,
diff --git a/asm/intel64/codes.h b/asm/intel64/codes.h
index 0ff17f1..112eef4 100644
--- a/asm/intel64/codes.h
+++ b/asm/intel64/codes.h
@@ -10,3 +10,6 @@ std::vector<uint8_t> REX(const std::string& s);
 // Manual, page 530
 // Reg + Reg/Memory
 uint8_t ModRM(const std::string& reg, const std::string& rm);
+
+// Just the number of reg, e.g. for encoding inside primary opcode
+uint8_t RegNo(const std::string& reg);
diff --git a/asm/intel64/mov.cpp b/asm/intel64/mov.cpp
index 40a48f8..8603fc9 100644
--- a/asm/intel64/mov.cpp
+++ b/asm/intel64/mov.cpp
@@ -15,6 +15,8 @@ Op_mov::Op_mov(Asm::Args& args)
   // r/m8, r8: ModRM:r/m (w), ModRM:reg (r)
   machine_code = std::vector<uint8_t>{ 0x88 } +
    ModRM(std::any_cast<Asm::Args::Register8>(args[1]).name(), std::any_cast<Asm::Args::Register8>(args[0]).name());
+ } else if (args[0].type() == typeid(Asm::Args::Register32) && args[1].type() == typeid(Asm::Args::Immediate32)) { // mov reg32, imm32
+  machine_code = std::vector<uint8_t>{ static_cast<uint8_t>(0xB8 + RegNo(std::any_cast<Asm::Args::Register32>(args[0]).name())) } + std::any_cast<Asm::Args::Immediate32>(args[1]).getCode();
  } else {
   throw std::runtime_error("Unimplemented: mov "s + args[0].type().name() + " "s + args[1].type().name());
  }
@@ -25,6 +27,9 @@ namespace {
 bool registered {
  registerOp(mangleName<Asm::Args::Register8, Asm::Args::Register8>("mov"), [](Asm::Args& args) -> std::shared_ptr<Op>{
                return std::make_shared<Op_mov>(args);
+            }) &&
+ registerOp(mangleName<Asm::Args::Register32, Asm::Args::Immediate32>("mov"), [](Asm::Args& args) -> std::shared_ptr<Op>{
+               return std::make_shared<Op_mov>(args);
             })
 };