Linux C编译器从零开发三

AST语法树

BNF抽象

expr       = equality
equality   = relational ("==" relational | "!=" relational)*
relational = add ("<" add | "<=" add | ">" add | ">=" add)*
add        = mul ("+" mul | "-" mul)*
mul        = unary ("*" unary | "/" unary)*
unary      = ("+" | "-")? primary
primary    = num | "(" expr ")"

AST构建

#include "chibicc.h"

static Node *expr(Token **rest, Token *tok);
static Node *equality(Token **rest, Token *tok);
static Node *relational(Token **rest, Token *tok);
static Node *add(Token **rest, Token *tok);
static Node *mul(Token **rest, Token *tok);
static Node *unary(Token **rest, Token *tok);
static Node *primary(Token **rest, Token *tok);

static Node *new_node(NodeKind kind) {
  Node *node = calloc(1, sizeof(Node));
  node->kind = kind;
  return node;
}

static Node *new_binary(NodeKind kind, Node *lhs, Node *rhs) {
  Node *node = new_node(kind);
  node->lhs = lhs;
  node->rhs = rhs;
  return node;
}

static Node *new_unary(NodeKind kind, Node *expr) {
  Node *node = new_node(kind);
  node->lhs = expr;
  return node;
}

static Node *new_num(int val) {
  Node *node = new_node(ND_NUM);
  node->val = val;
  return node;
}

// expr = equality
static Node *expr(Token **rest, Token *tok) {
  return equality(rest, tok);
}

// equality = relational ("==" relational | "!=" relational)*
static Node *equality(Token **rest, Token *tok) {
  Node *node = relational(&tok, tok);

  for (;;) {
    if (equal(tok, "==")) {
      node = new_binary(ND_EQ, node, relational(&tok, tok->next));
      continue;
    }

    if (equal(tok, "!=")) {
      node = new_binary(ND_NE, node, relational(&tok, tok->next));
      continue;
    }

    *rest = tok;
    return node;
  }
}

// relational = add ("<" add | "<=" add | ">" add | ">=" add)*
static Node *relational(Token **rest, Token *tok) {
  Node *node = add(&tok, tok);

  for (;;) {
    if (equal(tok, "<")) {
      node = new_binary(ND_LT, node, add(&tok, tok->next));
      continue;
    }

    if (equal(tok, "<=")) {
      node = new_binary(ND_LE, node, add(&tok, tok->next));
      continue;
    }

    if (equal(tok, ">")) {
      node = new_binary(ND_LT, add(&tok, tok->next), node);
      continue;
    }

    if (equal(tok, ">=")) {
      node = new_binary(ND_LE, add(&tok, tok->next), node);
      continue;
    }

    *rest = tok;
    return node;
  }
}

// add = mul ("+" mul | "-" mul)*
static Node *add(Token **rest, Token *tok) {
  Node *node = mul(&tok, tok);

  for (;;) {
    if (equal(tok, "+")) {
      node = new_binary(ND_ADD, node, mul(&tok, tok->next));
      continue;
    }

    if (equal(tok, "-")) {
      node = new_binary(ND_SUB, node, mul(&tok, tok->next));
      continue;
    }

    *rest = tok;
    return node;
  }
}

// mul = unary ("*" unary | "/" unary)*
static Node *mul(Token **rest, Token *tok) {
  Node *node = unary(&tok, tok);

  for (;;) {
    if (equal(tok, "*")) {
      node = new_binary(ND_MUL, node, unary(&tok, tok->next));
      continue;
    }

    if (equal(tok, "/")) {
      node = new_binary(ND_DIV, node, unary(&tok, tok->next));
      continue;
    }

    *rest = tok;
    return node;
  }
}

// unary = ("+" | "-") unary
//       | primary
static Node *unary(Token **rest, Token *tok) {
  if (equal(tok, "+"))
    return unary(rest, tok->next);

  if (equal(tok, "-"))
    return new_unary(ND_NEG, unary(rest, tok->next));

  return primary(rest, tok);
}

// primary = "(" expr ")" | num
static Node *primary(Token **rest, Token *tok) {
  if (equal(tok, "(")) {
    Node *node = expr(&tok, tok->next);
    *rest = skip(tok, ")");
    return node;
  }

  if (tok->kind == TK_NUM) {
    Node *node = new_num(tok->val);
    *rest = tok->next;
    return node;
  }

  error_tok(tok, "expected an expression");
}

Node *parse(Token *tok) {
  Node *node = expr(&tok, tok);
  if (tok->kind != TK_EOF)
    error_tok(tok, "extra token");
  return node;
}

代码生成 

#include "chibicc.h"

static int depth;

static void push(void) {
  printf("  push %%rax\n");
  depth++;
}

static void pop(char *arg) {
  printf("  pop %s\n", arg);
  depth--;
}

static void gen_expr(Node *node) {
  switch (node->kind) {
  case ND_NUM:
    printf("  mov $%d, %%rax\n", node->val);
    return;
  case ND_NEG:
    gen_expr(node->lhs);
    printf("  neg %%rax\n");
    return;
  }

  gen_expr(node->rhs);
  push();
  gen_expr(node->lhs);
  pop("%rdi");

  switch (node->kind) {
  case ND_ADD:
    printf("  add %%rdi, %%rax\n");
    return;
  case ND_SUB:
    printf("  sub %%rdi, %%rax\n");
    return;
  case ND_MUL:
    printf("  imul %%rdi, %%rax\n");
    return;
  case ND_DIV:
    printf("  cqo\n");
    printf("  idiv %%rdi\n");
    return;
  case ND_EQ:
  case ND_NE:
  case ND_LT:
  case ND_LE:
    printf("  cmp %%rdi, %%rax\n");

    if (node->kind == ND_EQ)
      printf("  sete %%al\n");
    else if (node->kind == ND_NE)
      printf("  setne %%al\n");
    else if (node->kind == ND_LT)
      printf("  setl %%al\n");
    else if (node->kind == ND_LE)
      printf("  setle %%al\n");

    printf("  movzb %%al, %%rax\n");
    return;
  }

  error("invalid expression");
}

void codegen(Node *node) {
  printf("  .globl main\n");
  printf("main:\n");

  gen_expr(node);
  printf("  ret\n");

  assert(depth == 0);
}

#include "chibicc.h"

int main(int argc, char **argv) {
  if (argc != 2)
    error("%s: invalid number of arguments", argv[0]);

  Token *tok = tokenize(argv[1]);
  Node *node = parse(tok);
  codegen(node);
  return 0;
}

 ./chibicc "1 + -2* 3 -(4+5/6) > tmp.s
cc -o tmp tmp.s

./tmp

echo $?

247

---

创作不易，小小的支持一下吧！