about summary refs log tree commit diff stats
diff options
context:
space:
mode:
authorKartik K. Agaram <vc@akkartik.com>2017-10-11 01:38:47 -0700
committerKartik K. Agaram <vc@akkartik.com>2017-10-11 01:38:47 -0700
commitf8831a023d91556eb105edc27e7cbb61bde573ca (patch)
treee36bf6a03e7aaba119044c2929255e5faa7f6b31
parenta7c35665a5b092eab1f4be159bd3f1e43f038283 (diff)
downloadmu-f8831a023d91556eb105edc27e7cbb61bde573ca.tar.gz
4014 - core skeleton for x86 interpreter
-rw-r--r--subx/000organization.cc3
-rw-r--r--subx/001help.cc20
-rw-r--r--subx/010core.cc173
3 files changed, 180 insertions, 16 deletions
diff --git a/subx/000organization.cc b/subx/000organization.cc
index d9f072f9..753ddca1 100644
--- a/subx/000organization.cc
+++ b/subx/000organization.cc
@@ -112,6 +112,9 @@
 
 int main(int argc, char* argv[]) {
   atexit(reset);
+  // run on a 32-bit system
+  assert(sizeof(int) == 4);
+  assert(sizeof(float) == 4);
 
   // End One-time Setup
 
diff --git a/subx/001help.cc b/subx/001help.cc
index ca1cb106..d4c3334f 100644
--- a/subx/001help.cc
+++ b/subx/001help.cc
@@ -2,24 +2,12 @@
 //: This should give you a sense for what to look forward to in later layers.
 
 :(before "End Commandline Parsing")
-if (argc != 1) {
+if (argc <= 1 || is_equal(argv[1], "--help")) {
   //: this is the functionality later layers will provide
   // currently no automated tests for commandline arg parsing
-  return 1;
-}
-
-//: Support for option parsing.
-//: Options always begin with '--' and are always the first arguments. An
-//: option will never follow a non-option.
-:(before "End Commandline Parsing")
-char** arg = &argv[1];
-while (argc > 1 && starts_with(*arg, "--")) {
-  if (false)
-    ;  // no-op branch just so any further additions can consistently always start with 'else'
-  // End Commandline Options(*arg)
-  else
-    cerr << "skipping unknown option " << *arg << '\n';
-  --argc;  ++argv;  ++arg;
+  cerr << "Usage:\n"
+       << "  subx test\n";
+  return 0;
 }
 
 //:: Helper function used by the above fragment of code (and later layers too,
diff --git a/subx/010core.cc b/subx/010core.cc
new file mode 100644
index 00000000..611b3f33
--- /dev/null
+++ b/subx/010core.cc
@@ -0,0 +1,173 @@
+//:: simulated x86 registers
+
+:(before "End Types")
+enum {
+  EAX,
+  ECX,
+  EDX,
+  EBX,
+  ESP,
+  EBP,
+  ESI,
+  EDI,
+  NUM_INT_REGISTERS,
+};
+union reg {
+  int32_t i;
+  uint32_t u;
+};
+:(before "End Globals")
+reg R[NUM_INT_REGISTERS] = { {0} };
+uint32_t EIP = 0;
+:(before "End Reset")
+bzero(R, sizeof(R));
+EIP = 0;
+
+//:: simulated flag registers; just a subset that we care about
+
+:(before "End Globals")
+bool OF=false, ZF=false, SF=false;
+:(before "End Reset")
+OF = ZF = SF = false;
+
+//: how the flag registers are updated after each instruction
+
+:(before "End Includes")
+// beware: no side-effects in args
+#define PERFORM_ARITHMETIC_BINOP(op, arg1, arg2) { \
+  /* arg1 and arg2 must be signed */ \
+  int64_t tmp = arg1 op arg2; \
+  arg1 = arg1 op arg2; \
+  SF = (arg1 < 0); \
+  ZF = (arg1 == 0); \
+  OF = (arg1 != tmp); \
+}
+
+#define PERFORM_BITWISE_BINOP(op, arg1, arg2) { \
+  /* arg1 and arg2 must be unsigned */ \
+  arg1 = arg1 op arg2; \
+  SF = (arg1 >> 31); \
+  ZF = (arg1 == 0); \
+  OF = false; \
+}
+
+//:: simulated RAM
+
+:(before "End Globals")
+vector<uint8_t> Memory;
+:(before "End Reset")
+Memory.clear();
+
+//:: core interpreter loop
+
+:(scenario add_imm32_to_eax)
+# opcode     modrm     sib       displacement      immediate
+  05                                               0a 0b 0c 0d  # add EAX, 0x0d0c0b0a
++load: 05
++load: 0a
++load: 0b
++load: 0c
++load: 0d
++run: add to EAX immediate 0x0d0c0b0a
++reg: storing 0x0d0c0b0a in register EAX
+
+:(code)
+// helper for tests: load a program into memory from a textual representation
+// of its bytes, and run it
+void run(const string& text_bytes) {
+  load_program(text_bytes);
+  EIP = 1;  // preserve null pointer
+  while (EIP < Memory.size())
+    run_one_instruction();
+}
+
+void load_program(const string& text_bytes) {
+  assert(Memory.empty());
+  // initialize address 0
+  Memory.push_back(0);
+  // now, to read the hex bytes in ASCII, we'll use C's strtol
+  // strtol needs a char*, so we grab the buffer backing the string object
+  char* curr = const_cast<char*>(&text_bytes[0]);   // non-standard approach, but blessed by Herb Sutter (http://herbsutter.com/2008/04/07/cringe-not-vectors-are-guaranteed-to-be-contiguous/#comment-483)
+  char* max = curr + strlen(curr);
+  while (true) {
+    if (curr >= max) return;
+    // skip whitespace
+    while (*curr == ' ' || *curr == '\n') ++curr;
+    // skip comments
+    if (*curr == '#') {
+      while (*curr != '\n') {
+        ++curr;
+        if (curr >= max) return;
+      }
+      ++curr;
+      if (curr >= max) return;
+    }
+    Memory.push_back(strtol(curr, &curr, /*hex*/16));
+    trace(99, "load") << HEXBYTE << static_cast<unsigned int>(Memory.back()) << end();  // ugly that iostream doesn't print uint8_t as an integer
+  }
+}
+
+// skeleton of how x86 instructions are decoded
+void run_one_instruction() {
+  uint8_t op=0, op2=0, op3=0;
+  switch(op = next()) {
+  // our first opcode
+  case 0x05: {  // add EAX, imm32
+    int32_t arg2 = imm32();
+    trace(2, "run") << "add to EAX immediate 0x" << HEXWORD << arg2 << end();
+    PERFORM_ARITHMETIC_BINOP(+, R[EAX].i, arg2);
+    trace(98, "reg") << "storing 0x" << HEXWORD << R[EAX].i << " in register EAX" << end();
+    break;
+  }
+  // End Single-Byte Opcodes
+  case 0x0f:
+    switch(op2 = next()) {
+    // End Two-Byte Opcodes Starting With 0x0f
+    default:
+      cerr << "unrecognized second opcode after 0x0f: " << std::hex << op2 << '\n';
+      exit(1);
+    }
+    break;
+  case 0xf3:
+    switch(op2 = next()) {
+    // End Two-Byte Opcodes Starting With 0xf3
+    case 0x0f:
+      switch(op3 = next()) {
+      // End Three-Byte Opcodes Starting With 0xf3 0x0f
+      default:
+        cerr << "unrecognized third opcode after 0xf3 0x0f: " << std::hex << op3 << '\n';
+        exit(1);
+      }
+      break;
+    default:
+      cerr << "unrecognized second opcode after 0xf3: " << std::hex << op2 << '\n';
+      exit(1);
+    }
+    break;
+  case 0xf4:  // hlt
+    EIP = Memory.size();
+    break;
+  default:
+    cerr << "unrecognized opcode: " << std::hex << op << '\n';
+    exit(1);
+  }
+}
+
+uint8_t next(void) {
+  if (EIP >= Memory.size()) return /*hlt*/0xf4;
+  return Memory.at(EIP++);
+}
+
+// read a 32-bit immediate in little-endian order from the instruction stream
+int32_t imm32(void) {
+  int result = next();
+  result |= (next()<<8);
+  result |= (next()<<16);
+  result |= (next()<<24);
+  return result;
+}
+
+:(before "End Includes")
+#include <iomanip>
+#define HEXBYTE  std::hex << std::setw(2) << std::setfill('0')
+#define HEXWORD  std::hex << std::setw(8) << std::setfill('0')