RVBTCC/boot.c

#include "boot-lib.h"

// lexer

int is_digit(int ch) {
    return '0' <= ch && ch <= '9';
}

int is_id_start(int ch) {
    return 'a' <= ch && ch <= 'z' || 'A' <= ch && ch <= 'Z' || ch == '_';
}

int is_id_cont(int ch) {
    return is_id_start(ch) || is_digit(ch);
}

int token_state;
int token_type;
int token_data;

const int TOKEN_EOF = 0;
const int TOKEN_SEMICOLON = 1;
const int TOKEN_ADD = 2;
const int TOKEN_MINUS = 3;
const int TOKEN_STAR = 4;
const int TOKEN_DIV = 5;
const int TOKEN_REM = 6;
const int TOKEN_ASSIGN = 7;
const int TOKEN_COMMA = 8;
const int TOKEN_DOT = 9;
const int TOKEN_LSHIFT = 10;
const int TOKEN_RSHIFT = 11;
const int TOKEN_AND = 12;
const int TOKEN_OR = 13;
const int TOKEN_XOR = 14;
const int TOKEN_COMPL = 15;
const int TOKEN_NOT = 16;
const int TOKEN_LAND = 17;
const int TOKEN_LOR = 18;
const int TOKEN_ELLIPSIS = 19;
const int TOKEN_INC = 20;
const int TOKEN_DEC = 21;

const int TOKEN_EQ = 40;
const int TOKEN_NE = 41;
const int TOKEN_LT = 42;
const int TOKEN_GT = 43;
const int TOKEN_LE = 44;
const int TOKEN_GE = 45;

const int TOKEN_PAREN_LEFT = 50;
const int TOKEN_PAREN_RIGHT = 51;
const int TOKEN_BRACKET_LEFT = 52;
const int TOKEN_BRACKET_RIGHT = 53;
const int TOKEN_BRACE_LEFT = 54;
const int TOKEN_BRACE_RIGHT = 55;

const int TOKEN_STRING = 99;
const int TOKEN_NUMBER = 100;
const int TOKEN_ID = 101;
const int TOKEN_IF = 102;
const int TOKEN_ELSE = 103;
const int TOKEN_WHILE = 104;
const int TOKEN_FOR = 105;
const int TOKEN_DO = 106;
const int TOKEN_BREAK = 107;
const int TOKEN_CONTINUE = 108;
const int TOKEN_RETURN = 109;

const int TOKEN_CONST = 127;
const int TOKEN_VOID = 128;
const int TOKEN_INT = 129;
const int TOKEN_CHAR = 130;

const int TYPE_VOID = 0;
const int TYPE_INT = 1;
const int TYPE_CHAR = 2;
const int TYPE_VOID_PTR = 16;
const int TYPE_INT_PTR = 17;
const int TYPE_CHAR_PTR = 18;

const int TYPE_PTR_MASK = 16;
const int TYPE_FUNC_MASK = 32;
const int TYPE_TOKEN_MASK = 128;

int parse_int(int ch) {
    int num = ch - '0';
    while (is_digit(ch = getchar())) {
        num = num * 10;
        num = num + ch - '0';
    }
    ungetchar(ch);
    return num;
}

int get_escaped_char() {
    int ch = getchar();
    if (ch == 'n') {
        ch = '\n';
    } else if (ch == 't') {
        ch = '\t';
    } else if (ch == 'r') {
        ch = '\r';
    } else if (ch == '0') {
        ch = '\0';
    } else if (ch == '\\') {
        ch = '\\';
    } else if (ch == '\'') {
        ch = '\'';
    } else if (ch == '\"') {
        ch = '\"';
    } else {
        eprintf("unexpected escaped character: %c\n", ch);
        exit(1);
    }
    return ch;
}

int streq(const char* s1, const char* s2) {
    while (*s1 && *s2 && *s1 == *s2) {
        s1++;
        s2++;
    }
    return *s1 == *s2;
}

char string_table[65536];
int string_offset;
int string_lut[4096];
int string_lut_size;
int parse_string() {
    int offset = string_offset;
    int ch;
    while ((ch = getchar()) != '"') {
        if (ch == -1 || ch == '\n') {
            eprintf("expecting '\"'\n");
            exit(1);
        }
        if (ch == '\\') {
            ch = get_escaped_char();
        }
        string_table[string_offset++] = ch;
    }
    string_table[string_offset++] = 0;
    string_lut[string_lut_size] = offset;
    return string_lut_size++;
}

char id_table[65536];
int id_offset;
int id_lut[4096];
int id_lut_size;
int parse_id(int ch) {
    int offset = id_offset;
    id_table[id_offset++] = ch;
    while (is_id_cont(ch = getchar())) {
        id_table[id_offset++] = ch;
    }
    ungetchar(ch);
    id_table[id_offset++] = 0;
    id_lut[id_lut_size] = offset;
    return id_lut_size++;
}

void rewind_id(int new_data) {
    id_offset = id_lut[token_data];
    token_data = new_data;
    --id_lut_size;
}

void dedup_id() {
    int last_id = id_lut_size - 1;
    char* latest = id_table + id_lut[last_id];
    for (int i = 0; i < last_id; i++) {
        char* candidate = id_table + id_lut[i];
        if (streq(candidate, latest)) {
            rewind_id(i);
            return;
        }
    }
}

void parse_id_like(int ch) {
    token_type = TOKEN_ID;
    token_data = parse_id(ch);
    char* id = id_table + id_lut[token_data];
    if (streq(id, "int")) {
        token_type = TOKEN_INT;
    } else if (streq(id, "if")) {
        token_type = TOKEN_IF;
    } else if (streq(id, "else")) {
        token_type = TOKEN_ELSE;
    } else if (streq(id, "while")) {
        token_type = TOKEN_WHILE;
    } else if (streq(id, "break")) {
        token_type = TOKEN_BREAK;
    } else if (streq(id, "continue")) {
        token_type = TOKEN_CONTINUE;
    } else if (streq(id, "return")) {
        token_type = TOKEN_RETURN;
    } else if (streq(id, "void")) {
        token_type = TOKEN_VOID;
    } else if (streq(id, "const")) {
        token_type = TOKEN_CONST;
    } else if (streq(id, "char")) {
        token_type = TOKEN_CHAR;
    } else if (streq(id, "for")) {
        token_type = TOKEN_FOR;
    } else if (streq(id, "do")) {
        token_type = TOKEN_DO;
    }
    if (token_type != TOKEN_ID) {
        rewind_id(0);
    } else {
        dedup_id();
    }
}

void unget_token() {
    token_state = 1;
}

void next_token() {
    if (token_state) {
        token_state = 0;
        return;
    }
    int ch = getchar();
    while (ch == ' ' || ch == '\t' || ch == '\r' || ch == '\n') {
        ch = getchar();
    }
    if (ch == -1) {
        token_type = TOKEN_EOF;
    } else if (ch == '(') {
        token_type = TOKEN_PAREN_LEFT;
    } else if (ch == ')') {
        token_type = TOKEN_PAREN_RIGHT;
    } else if (ch == '[') {
        token_type = TOKEN_BRACKET_LEFT;
    } else if (ch == ']') {
        token_type = TOKEN_BRACKET_RIGHT;
    } else if (ch == '{') {
        token_type = TOKEN_BRACE_LEFT;
    } else if (ch == '}') {
        token_type = TOKEN_BRACE_RIGHT;
    } else if (ch == '+') {
        int ch2 = getchar();
        if (ch2 == '+') {
            token_type = TOKEN_INC;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_ADD;
        }
    } else if (ch == '-') {
        int ch2 = getchar();
        if (ch2 == '-') {
            token_type = TOKEN_DEC;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_MINUS;
        }
    } else if (ch == '*') {
        token_type = TOKEN_STAR;
    } else if (ch == '/') {
        int ch2 = getchar();
        if (ch2 == '/') {
            do ch = getchar(); while (ch != -1 && ch != '\n');
            next_token();
            return;
        } else if (ch2 == '*') {
            while (1) {
                ch = getchar();
                if (ch == -1) {
                    eprintf("expecting '*/'\n");
                    exit(1);
                }
                if (ch == '*') {
                    ch = getchar();
                    if (ch == '/') {
                        break;
                    }
                }
            }
            next_token();
            return;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_DIV;
        }
    } else if (ch == '%') {
        token_type = TOKEN_REM;
    } else if (ch == ';') {
        token_type = TOKEN_SEMICOLON;
    } else if (ch == ',') {
        token_type = TOKEN_COMMA;
    } else if (ch == '<') {
        int ch2 = getchar();
        if (ch2 == '=') {
            token_type = TOKEN_LE;
        } else if (ch2 == '<') {
            token_type = TOKEN_LSHIFT;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_LT;
        }
    } else if (ch == '>') {
        int ch2 = getchar();
        if (ch2 == '=') {
            token_type = TOKEN_GE;
        } else if (ch2 == '>') {
            token_type = TOKEN_RSHIFT;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_GT;
        }
    } else if (ch == '=') {
        int ch2 = getchar();
        if (ch2 == '=') {
            token_type = TOKEN_EQ;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_ASSIGN;
        }
    } else if (ch == '!') {
        int ch2 = getchar();
        if (ch2 == '=') {
            token_type = TOKEN_NE;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_NOT;
        }
    } else if (ch == '&') {
        int ch2 = getchar();
        if (ch2 == '&') {
            token_type = TOKEN_LAND;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_AND;
        }
    } else if (ch == '|') {
        int ch2 = getchar();
        if (ch2 == '|') {
            token_type = TOKEN_LOR;
        } else {
            ungetchar(ch2);
            token_type = TOKEN_OR;
        }
    } else if (ch == '^') {
        token_type = TOKEN_XOR;
    } else if (ch == '~') {
        token_type = TOKEN_COMPL;
    } else if (ch == '\'') {
        token_type = TOKEN_NUMBER;
        token_data = getchar();
        if (token_data == '\\') {
            token_data = get_escaped_char();
        }
        if (getchar() != '\'') {
            eprintf("expecting '\n");
            exit(1);
        }
    } else if (ch == '"') {
        token_type = TOKEN_STRING;
        token_data = parse_string();
    } else if (ch == '.') {
        int ch2 = getchar();
        if (ch2 == '.') {
            int ch3 = getchar();
            if (ch3 == '.') {
                token_type = TOKEN_ELLIPSIS;
            } else {
                eprintf("unexpected character: %c\n", ch3);
                exit(1);
            }
        } else {
            eprintf("unexpected character: %c\n", ch2);
            exit(1);
        }
    } else if (is_digit(ch)) {
        token_type = TOKEN_NUMBER;
        token_data = parse_int(ch);
    } else if (is_id_start(ch)) {
        parse_id_like(ch);
    } else {
        eprintf("unexpected character: %c(%d)\n", ch, ch);
        exit(1);
    }
    eprintf("token: %d\n", token_type);
    if (token_type == TOKEN_ID) {
        const char* name = id_table + id_lut[token_data];
        eprintf("  id: %s\n", name);
    } else if (token_type == TOKEN_NUMBER) {
        eprintf("  number: %d\n", token_data);
    }
}

void expect_token(int expected_type) {
    next_token();
    if (token_type != expected_type) {
        eprintf("unexpected token: %d, should be %d\n", token_type, expected_type);
        exit(1);
    }
}

void ignore_const() {
    if (token_type == TOKEN_CONST) {
        next_token();
    }
}

int parse_type() {
    ignore_const();
    if (token_type == TOKEN_INT || token_type == TOKEN_CHAR || token_type == TOKEN_VOID) {
        int type = token_type & ~TYPE_TOKEN_MASK;
        next_token();
        ignore_const();
        if (token_type == TOKEN_STAR) {
            ignore_const();
            return type | TYPE_PTR_MASK;
        }
        unget_token();
        return type;
    } else {
        return -1;
    }
}

// asm

int epilog_label;

int local_table[4096]; // id -> local id
int next_local_id = 2;
int max_local_id = 2;

const int MARKER_TEMP = 0;
const int MARKER_SCALAR = 1;
const int MARKER_ARRAY = 2;

int local_marker[4096];
int global_marker[4096];
int local_type[4096];
int global_type[4096];
int indirection[4096];

void reset_local() {
    next_local_id = 2;
    max_local_id = 2;
    for (int i = 0; i < 4096; ++i) {
        local_table[i] = 0;
        local_marker[i] = MARKER_TEMP;
        local_type[i] = TYPE_VOID;
        indirection[i] = 0;
    }
}

void reset_temp() {
    while (next_local_id > 2 && local_marker[next_local_id - 1] == MARKER_TEMP) {
        --next_local_id;
    }
}

int next_reg(int type) {
    int reg = next_local_id++;
    local_type[reg] = type;
    indirection[reg] = 0;
    if (next_local_id > max_local_id) {
        max_local_id = next_local_id;
    }
    return reg;
}

int declare_local(int id, int type) {
    if (local_table[id] != 0) return local_table[id];
    int reg = next_reg(type);
    local_marker[reg] = MARKER_SCALAR;
    return local_table[id] = reg;
}

int declare_local_array(int id, int type, int size) {
    if (local_table[id] != 0) return local_table[id];
    int reg;
    for (int i = 0; i < size; ++i) local_marker[reg = next_reg(type)] = MARKER_ARRAY;
    return local_table[id] = reg;
}

void declare_global(int id, int marker, int type) {
    global_marker[id] = marker;
    global_type[id] = type;
}

int check_itype_immediate(int value) {
    return value >= -2048 && value <= 2047;
}

void asm_ld(const char* rd, int imm, const char* rs) {
    if (check_itype_immediate(imm)) {
        printf("  ld %s, %d(%s)\n", rd, imm, rs);
    } else {
        printf("  li t0, %d\n", imm);
        printf("  add t0, %s, t0\n", rs);
        printf("  ld %s, 0(t0)\n", rd);
    }
}

void asm_sd(const char* rs1, int imm, const char* rs2) {
    if (check_itype_immediate(imm)) {
        printf("  sd %s, %d(%s)\n", rs1, imm, rs2);
    } else {
        printf("  li t0, %d\n", imm);
        printf("  add t0, %s, t0\n", rs2);
        printf("  sd %s, 0(t0)\n", rs1);
    }
}

void asm_addi(const char* rd, const char* rs, int imm) {
    if (check_itype_immediate(imm)) {
        printf("  addi %s, %s, %d\n", rd, rs, imm);
    } else {
        printf("  li t0, %d\n", imm);
        printf("  add %s, %s, t0\n", rd, rs);
    }
}

void load_address(int rd, int id) {
    if (id == -1) {
        eprintf("void cannot be arithmetically operated\n");
        exit(1);
    }
    int offset = -id * 8 - 8;
    if (indirection[id]) {
        if (check_itype_immediate(offset)) {
            printf("  ld t%d, %d(fp) # indirection\n", rd, offset);
        } else {
            printf("  li t%d, %d\n", rd, offset);
            printf("  add t%d, fp, t%d\n", rd, rd);
            printf("  ld t%d, 0(t%d)  # indirection\n", rd, rd);
        }
    } else {
        if (check_itype_immediate(offset)) {
            printf("  addi t%d, fp, %d\n", rd, offset);
        } else {
            printf("  li t%d, %d\n", rd, offset);
            printf("  add t%d, fp, t%d\n", rd, rd);
        }
    }
}

void load(int rd, int id) {
    load_address(rd, id);
    int type = local_type[id];
    const char* op = "lw"; // int
    if (type == TYPE_CHAR) {
        op = "lb";
    } else if (type & TYPE_PTR_MASK || type & TYPE_FUNC_MASK) {
        op = "ld";
    }
    printf("  %s t%d, 0(t%d) # id: type %d\n", op, rd, rd, type);
}

void store_t0(int id) {
    load_address(1, id);
    int type = local_type[id];
    const char* op = "sw"; // int
    if (type == TYPE_CHAR) {
        op = "sb";
    } else if (type & TYPE_PTR_MASK || type & TYPE_FUNC_MASK) {
        op = "sd";
    }
    printf("  %s t0, 0(t1) # id: type %d\n", op, type);
}

int materialize_t0(int type) {
    int reg = next_reg(type);
    store_t0(reg);
    return reg;
}

int dereference(int reg) {
    local_type[reg] = local_type[reg] & ~TYPE_PTR_MASK;
    indirection[reg] = 1;
    return reg;
}

int lookup(int id) {
    int local = local_table[id];
    if (local) {
        if (local_marker[local] == MARKER_ARRAY) {
            load_address(0, local);
            return materialize_t0(local_type[local] | TYPE_PTR_MASK);
        }
        return local;   
    }
    const char* name = id_table + id_lut[id];
    if (global_marker[id]) {
        printf("  la t0, %s # id: %d\n", name, id);
        int reg = materialize_t0(global_type[id] | TYPE_PTR_MASK);
        if (global_marker[id] == MARKER_SCALAR) {
            reg = dereference(reg);
        }
        return reg;
    }
    eprintf("unresolved identifier: %s\n", name);
    exit(1);
}


int next_label_id = 0;
int next_label() {
    return next_label_id++;
}

int asm_label(int label) {
    printf("L%d:\n", label);
    return label;
}

int is_not_reusable(int rs1, int expected_type) {
    return indirection[rs1] || local_marker[rs1] != MARKER_TEMP || local_type[rs1] != expected_type;
}

int asm_r(const char* op, int rs1) {
    load(0, rs1);
    printf("  %s t0, t0\n", op);
    int rd = rs1;
    if (is_not_reusable(rs1, TYPE_INT)) {
        rd = next_reg(TYPE_INT);
    }
    store_t0(rd);
    return rd;
}

int asm_r_arith(const char* op, int rs1) {
    if (local_type[rs1] & TYPE_PTR_MASK || local_type[rs1] & TYPE_FUNC_MASK) {
        eprintf("pointer cannot be arithmetically operated by %s\n", op);
        exit(1);
    }
    return asm_r(op, rs1);
}

int asm_rr(const char* op, int rs1, int rs2) {
    load(0, rs1);
    load(1, rs2);
    printf("  %s t0, t0, t1\n", op);
    int rd = rs1;
    if (is_not_reusable(rd, TYPE_INT)) { 
        rd = rs2;
        if (is_not_reusable(rd, TYPE_INT)) { 
            rd = next_reg(TYPE_INT);
        }
    }
    store_t0(rd);
    return rd;
}

int asm_rr_arith(const char* op, int rs1, int rs2) {
    if (local_type[rs1] & TYPE_PTR_MASK || local_type[rs2] & TYPE_PTR_MASK 
        || local_type[rs1] & TYPE_FUNC_MASK || local_type[rs2] & TYPE_FUNC_MASK) {
        eprintf("pointer cannot be arithmetically operated by %s\n", op);
        exit(1);
    }
    return asm_rr(op, rs1, rs2);
}

int asm_rr_cmp(const char* op, int rs1, int rs2) {
    // since NULL is virtually 0, it is considered valid example of a pointer comparing with an integer
    return asm_rr(op, rs1, rs2);
}

void asm_beqz(int rs1, int label) {
    load(0, rs1);
    printf("  beqz t0, L%d\n", label);
}

void asm_bnez(int rs1, int label) {
    load(0, rs1);
    printf("  bnez t0, L%d\n", label);
}

void asm_j(int label) {
    printf("  j L%d\n", label);
}

int break_label_stack[4096];
int cont_label_stack[4096];
int break_label_stack_size;
int cont_label_stack_size;

int asm_get_break_label() {
    return break_label_stack[break_label_stack_size - 1];
}

int asm_get_cont_label() {
    return cont_label_stack[cont_label_stack_size - 1];
}

void asm_push_label(int break_label, int cont_label) {
    break_label_stack[break_label_stack_size++] = break_label;
    cont_label_stack[cont_label_stack_size++] = cont_label;
}

void asm_pop_label() {
    --break_label_stack_size;
    --cont_label_stack_size;
}

int step_of(int type) {
    if (type == TYPE_INT_PTR) {
        return 4;
    }
    return 1;
}

void asm_shift_t0(const char* op, int type) {
    if (type == TYPE_INT_PTR) {
        printf("  %s t0, t0, 2\n", op);
    }
}

int asm_add(int lhs, int rhs) {
    int type1 = local_type[lhs] & TYPE_PTR_MASK;
    int type2 = local_type[rhs] & TYPE_PTR_MASK;
    if (type1 != type2) {
        int ptr;
        int idx;
        if (type1) {
            ptr = lhs;
            idx = rhs;
        } else {
            ptr = rhs;
            idx = lhs;
        }
        int ptr_type = local_type[ptr];
        if (ptr_type == TYPE_VOID_PTR) {
            eprintf("void pointer cannot be arithmetically operated\n");
            exit(1);
        }
        load(0, idx);
        load(1, ptr);
        asm_shift_t0("slli", ptr_type);
        printf("  add t0, t0, t1\n");
        return materialize_t0(ptr_type);
    }
    if (type1 && type2) {
        eprintf("operands of addition cannot be both pointers\n");
        exit(1);
    }
    return asm_rr("add", lhs, rhs);
}

int asm_sub(int lhs, int rhs) {
    int lhs_type = local_type[lhs];
    int rhs_type = local_type[rhs];
    int type1 = lhs_type & TYPE_PTR_MASK;
    int type2 = rhs_type & TYPE_PTR_MASK;
    if (type1 && type2) {
        if (lhs_type != rhs_type) {
            eprintf("pointer type mismatch\n");
            exit(1);
        }
        if (lhs_type == TYPE_VOID_PTR) {
            eprintf("void pointer cannot be arithmetically operated\n");
            exit(1);
        }
        load(0, lhs);
        load(1, rhs);
        printf("  sub t0, t0, t1\n");
        asm_shift_t0("srli", lhs_type);
        return materialize_t0(TYPE_INT);
    }
    if (type1) {
        int neg = asm_r_arith("neg", rhs);
        return asm_add(lhs, neg);
    }
    return asm_rr("sub", lhs, rhs);
}

// parser
int parse_expr();

int parse_primary_expr() {
    next_token();
    if (token_type == TOKEN_EOF) {
        exit(1);
    } else if (token_type == TOKEN_NUMBER) {
        printf("  li t0, %d\n", token_data);
        return materialize_t0(TYPE_INT);
    } else if (token_type == TOKEN_ID) {
        return lookup(token_data);
    } else if (token_type == TOKEN_STRING) {
        printf("  la t0, .LC%d\n", token_data);
        return materialize_t0(TYPE_CHAR_PTR);
    } else if (token_type == TOKEN_PAREN_LEFT) {
        int reg = parse_expr();
        expect_token(TOKEN_PAREN_RIGHT);
        return reg;
    } else {
        eprintf("unexpected token in primary expression: %d\n", token_type);
        exit(1);
    }
}

int parse_function_call(int func) {
    int arg = 0;
    int args[8];
    while (1) {
        next_token();
        if (token_type == TOKEN_PAREN_RIGHT) {
            break;
        }
        unget_token();
        if (arg >= 8) {
            eprintf("too many arguments\n");
            exit(1);
        }
        args[arg++] = parse_expr();
        next_token();
        if (token_type == TOKEN_COMMA) {
            // continue;
        } else if (token_type == TOKEN_PAREN_RIGHT) {
            break;
        } else {
            eprintf("expecting ',' or ')'\n");
            exit(1);
        }
    }
    for (int i = 0; i < arg; ++i) {
        load(0, args[i]);
        printf("  mv a%d, t0\n", i);
    }
    load_address(0, func);
    printf("  jalr t0\n");
    int type = local_type[func];
    if (type != TYPE_VOID) {
        printf("  mv t0, a0\n");
        return materialize_t0(type);
    }
    return -1;
}

int parse_postfix_expr() {
    int lhs = parse_primary_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_INC) {
            int type = local_type[lhs];
            int reg = next_reg(type);
            load(0, lhs);
            store_t0(reg);
            printf("  addi t0, t0, %d\n", step_of(type));
            store_t0(lhs);
            lhs = reg;
        } else if (token_type == TOKEN_DEC) {
            int type = local_type[lhs];
            int reg = next_reg(type);
            load(0, lhs);
            store_t0(reg);
            printf("  addi t0, t0, -%d\n", step_of(type));
            store_t0(lhs);
            lhs = reg;
        } else if (token_type == TOKEN_BRACKET_LEFT) {
            int rhs = parse_expr();
            expect_token(TOKEN_BRACKET_RIGHT);
            lhs = dereference(asm_add(lhs, rhs));
        } else if (token_type == TOKEN_PAREN_LEFT) {
            lhs = parse_function_call(lhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_prefix_expr() {
    next_token();
    if (token_type == TOKEN_AND) {
        int reg = parse_postfix_expr();
        int type = local_type[reg];
        if (type & TYPE_PTR_MASK) {
            eprintf("cannot take address of a pointer\n");
            exit(1);
        }
        load_address(0, reg);
        return materialize_t0(type | TYPE_PTR_MASK);
    } else if (token_type == TOKEN_STAR) {
        int reg = parse_postfix_expr();
        int type = local_type[reg];
        if (!(type & TYPE_PTR_MASK)) {
            eprintf("cannot dereference a non-pointer\n");
            exit(1);
        }
        if (type == TYPE_VOID_PTR) {
            eprintf("cannot dereference void pointer\n");
            exit(1);
        }
        load(0, reg);
        return dereference(materialize_t0(type));
    } else if (token_type == TOKEN_MINUS) {
        int reg = parse_postfix_expr();
        return asm_r_arith("neg", reg);
    } else if (token_type == TOKEN_COMPL) {
        int reg = parse_postfix_expr();
        return asm_r_arith("not", reg);
    } else if (token_type == TOKEN_NOT) {
        int reg = parse_postfix_expr();
        return asm_r("seqz", reg);
    } else if (token_type == TOKEN_INC) {
        int reg = parse_postfix_expr();
        load(0, reg);
        printf("  addi t0, t0, %d\n", step_of(local_type[reg]));
        store_t0(reg);
        return reg;
    } else if (token_type == TOKEN_DEC) {
        int reg = parse_postfix_expr();
        load(0, reg);
        printf("  addi t0, t0, -%d\n", step_of(local_type[reg]));
        store_t0(reg);
        return reg;
    } else {
        unget_token();
        return parse_postfix_expr();
    }
}

int parse_mul_expr() {
    int lhs = parse_prefix_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_STAR) {
            int rhs = parse_prefix_expr();
            lhs = asm_rr_arith("mul", lhs, rhs);
        } else if (token_type == TOKEN_DIV) {
            int rhs =  parse_prefix_expr();
            lhs = asm_rr_arith("div", lhs, rhs);
        } else if (token_type == TOKEN_REM) {
            int rhs =  parse_prefix_expr();
            lhs = asm_rr_arith("rem", lhs, rhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_add_expr() {
    int lhs = parse_mul_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_ADD) {
            int rhs = parse_mul_expr();
            lhs = asm_add(lhs, rhs);
        } else if (token_type == TOKEN_MINUS) {
            int rhs = parse_mul_expr();
            lhs = asm_sub(lhs, rhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_shift_expr() {
    int lhs = parse_add_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_LSHIFT) {
            int rhs = parse_add_expr();
            lhs = asm_rr_arith("sll", lhs, rhs);
        } else if (token_type == TOKEN_RSHIFT) {
            int rhs = parse_add_expr();
            lhs = asm_rr_arith("sra", lhs, rhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_cmp_expr() {
    int lhs = parse_shift_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_LT) {
            int rhs = parse_shift_expr();
            lhs = asm_rr_cmp("slt", lhs, rhs);
        } else if (token_type == TOKEN_GT) {
            int rhs = parse_shift_expr();
            lhs = asm_rr_cmp("sgt", lhs, rhs);
        } else if (token_type == TOKEN_LE) {
            int rhs = parse_shift_expr();
            int sgt = asm_rr_cmp("sgt", lhs, rhs);
            lhs = asm_r("seqz", sgt);
        } else if (token_type == TOKEN_GE) {
            int rhs = parse_shift_expr();
            int slt = asm_rr_cmp("slt", lhs, rhs);
            lhs = asm_r("seqz", slt);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_eq_expr() {
    int lhs = parse_cmp_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_EQ) {
            int rhs = parse_cmp_expr();
            int xor0 = asm_rr_cmp("xor", lhs, rhs);
            lhs = asm_r("seqz", xor0);
        } else if (token_type == TOKEN_NE) {
            int rhs = parse_cmp_expr();
            int xor0 = asm_rr_cmp("xor", lhs, rhs);
            lhs = asm_r("snez", xor0);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_bitwise_and_expr() {
    int lhs = parse_eq_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_AND) {
            int rhs = parse_eq_expr();
            lhs = asm_rr_arith("and", lhs, rhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}


int parse_bitwise_xor_expr() {
    int lhs = parse_bitwise_and_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_XOR) {
            int rhs = parse_bitwise_and_expr();
            lhs = asm_rr_arith("xor", lhs, rhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_bitwise_or_expr() {
    int lhs = parse_bitwise_xor_expr();
    while (1) {
        next_token();
        if (token_type == TOKEN_OR) {
            int rhs = parse_bitwise_xor_expr();
            lhs = asm_rr_arith("or", lhs, rhs);
        } else {
            unget_token();
            break;
        }
    }
    return lhs;
}

int parse_logical_and_expr() {
    int lhs = parse_bitwise_or_expr();
    int label = next_label();
    int label_used = 0;
    while (1) {
        next_token();
        if (token_type == TOKEN_LAND) {
            lhs = asm_r("snez", lhs);
            asm_beqz(lhs, label);
            int rhs = parse_bitwise_or_expr();
            rhs = asm_r("snez", rhs);
            lhs = asm_rr("and", lhs, rhs);
            label_used = 1;
        } else {
            unget_token();
            break;
        }
    }
    if (label_used) {
        asm_label(label);
    }
    return lhs;
}

int parse_logical_or_expr() {
    int lhs = parse_logical_and_expr();
    int label = next_label();
    int label_used = 0;
    while (1) {
        next_token();
        if (token_type == TOKEN_LOR) {
            lhs = asm_r("snez", lhs);
            asm_bnez(lhs, label);
            int rhs = parse_logical_and_expr();
            rhs = asm_r("snez", rhs);
            lhs = asm_rr("or", lhs, rhs);
            label_used = 1;
        } else {
            unget_token();
            break;
        }
    }
    if (label_used) {
        asm_label(label);
    }
    return lhs;
}

int parse_assign_expr() {
    int lhs = parse_logical_or_expr();
    next_token();
    if (token_type == TOKEN_ASSIGN) {
        int rhs = parse_assign_expr();
        load(0, rhs);
        store_t0(lhs);
        return lhs;
    } else {
        unget_token();
        return lhs;
    }
}

int parse_expr() {
    return parse_assign_expr();
}

void parse_local_variable(int type) {
    if (type == TYPE_VOID) {
        eprintf("local variable of void type is not supported\n");
        exit(1);
    }
    expect_token(TOKEN_ID);
    int id = token_data;
    next_token();
    if (token_type == TOKEN_BRACKET_LEFT) {
        if (type & TYPE_PTR_MASK) {
            eprintf("local variable of array of pointers is not supported\n");
            exit(1);
        }
        expect_token(TOKEN_NUMBER);
        int size = token_data;
        expect_token(TOKEN_BRACKET_RIGHT);
        declare_local_array(id, type, size);
        next_token();
    } else {
        declare_local(id, type);
    }
    if (token_type == TOKEN_SEMICOLON) {
        unget_token();
        return;
    }
    unget_token();
    expect_token(TOKEN_ASSIGN);
    int reg = parse_expr();
    load(0, reg);
    store_t0(local_table[id]);
}

void parse_stmt();

void parse_if() {
    expect_token(TOKEN_PAREN_LEFT);
    int cond = parse_expr();
    int label1 = next_label();
    int label2 = next_label();
    asm_beqz(cond, label1);
    reset_temp();
    expect_token(TOKEN_PAREN_RIGHT);
    parse_stmt();
    asm_j(label2);
    asm_label(label1);
    next_token();
    if (token_type == TOKEN_ELSE) {
        parse_stmt();
    } else {
        unget_token();
    }
    asm_label(label2);
}

void parse_while() {
    expect_token(TOKEN_PAREN_LEFT);
    int break_label = next_label();
    int cont_label = next_label();
    asm_push_label(break_label, cont_label);
    asm_label(cont_label);
    int cond = parse_expr();
    asm_beqz(cond, break_label);
    reset_temp();
    expect_token(TOKEN_PAREN_RIGHT);
    parse_stmt();
    asm_j(cont_label);
    asm_label(break_label);
    asm_pop_label();
}

void parse_for() {
    expect_token(TOKEN_PAREN_LEFT);
    int cont_label = next_label();
    int break_label = next_label();
    int cond_label = next_label();
    int body_label = next_label();
    asm_push_label(break_label, cont_label);
    parse_stmt(); // init
    asm_label(cond_label);
    int cond = parse_expr();
    asm_beqz(cond, break_label);
    asm_j(body_label);
    reset_temp();
    expect_token(TOKEN_SEMICOLON);
    asm_label(cont_label);
    parse_expr(); // update
    reset_temp();
    expect_token(TOKEN_PAREN_RIGHT);
    asm_j(cond_label);
    asm_label(body_label);
    parse_stmt(); // body
    asm_j(cont_label);
    asm_label(break_label);
    asm_pop_label();
}

void parse_do_while() {
    int cont_label = next_label();
    int break_label = next_label();
    asm_push_label(break_label, cont_label);
    asm_label(cont_label);
    parse_stmt(); // body
    expect_token(TOKEN_WHILE);
    expect_token(TOKEN_PAREN_LEFT);
    int cond = parse_expr();
    asm_bnez(cond, cont_label);
    expect_token(TOKEN_PAREN_RIGHT);
    asm_label(break_label);
    asm_pop_label();
}

void parse_stmt() {
    next_token();
    int decl_type;
    if (token_type == TOKEN_IF) {
        parse_if();
        return;
    } else if (token_type == TOKEN_WHILE) {
        parse_while();
        return;
    } else if (token_type == TOKEN_FOR) {
        parse_for();
        return;
    } else if (token_type == TOKEN_DO) {
        parse_do_while();
    } else if (token_type == TOKEN_BRACE_LEFT) {
        while (1) {
            next_token();
            if (token_type == TOKEN_BRACE_RIGHT) {
                break;
            }
            unget_token();
            parse_stmt();
        }
        return;
    } else if (token_type == TOKEN_RETURN) {
        next_token();
        if (token_type == TOKEN_SEMICOLON) {
            asm_j(epilog_label);
            return;
        }
        unget_token();
        int reg = parse_expr();
        load(0, reg);
        printf("  mv a0, t0\n");
        asm_j(epilog_label);
    } else if (token_type == TOKEN_BREAK) {
        int label = asm_get_break_label();
        asm_j(label);
    } else if (token_type == TOKEN_CONTINUE) {
        int label = asm_get_cont_label();
        asm_j(label);
    } else if (token_type == TOKEN_SEMICOLON) {
        unget_token();
    } else if ((decl_type = parse_type()) >= 0) {
        parse_local_variable(decl_type);
    } else {
        unget_token();
        parse_expr();
    }
    expect_token(TOKEN_SEMICOLON);
    reset_temp();
}

void parse_function(const char* name) {
    reset_local();
    int arg = 0;
    int args[8];
    while (1) {
        next_token();
        if (token_type == TOKEN_PAREN_RIGHT) {
            break;
        }
        if (token_type == TOKEN_ELLIPSIS) {
            expect_token(TOKEN_PAREN_RIGHT);
            break;
        }
        if (token_type == TOKEN_VOID) {
            if (arg != 0) {
                eprintf("void should be the only argument\n");
                exit(1);
            }
            expect_token(TOKEN_PAREN_RIGHT);
            break;
        }
        int arg_type = parse_type();
        if (arg_type < 0 || arg_type == TYPE_VOID) {
            eprintf("unexpected a non-void argument type");
            exit(1);
        }
        expect_token(TOKEN_ID);
        int arg_name = token_data;
        next_token();
        if (token_type == TOKEN_BRACKET_LEFT) {
            expect_token(TOKEN_BRACKET_RIGHT);
            next_token();
            if (arg_type & TYPE_PTR_MASK) {
                eprintf("local variable of array of pointers is not supported\n");
                exit(1);
            }
            arg_type = arg_type | TYPE_PTR_MASK;
        }
        args[arg++] = declare_local(token_data, arg_type);
        if (token_type == TOKEN_COMMA) {
            // continue;
        } else if (token_type == TOKEN_PAREN_RIGHT) {
            break;
        } else {
            eprintf("expecting ',' or ')'\n");
            exit(1);
        }
    }
    next_token();
    if (token_type == TOKEN_SEMICOLON) {
        return;
    }
    unget_token();
    expect_token(TOKEN_BRACE_LEFT);
    printf(".text\n");
    printf(".global %s\n", name);
    printf("%s:\n", name);
    int label = next_label();
    int prolog_label = next_label();
    epilog_label = next_label();
    asm_j(prolog_label);
    asm_label(label);
    while (1) {
        next_token();
        if (token_type == TOKEN_BRACE_RIGHT) {
            break;
        }
        unget_token();
        parse_stmt();
    }
    asm_j(epilog_label);
    int frame_size = max_local_id * 8;
    if (frame_size % 16 != 0) {
        frame_size = frame_size + 8;
    }
    // prolog
    asm_label(prolog_label);
    asm_addi("sp", "sp", -frame_size);
    asm_sd("ra", frame_size - 8, "sp");
    asm_sd("fp", frame_size - 16, "sp");
    asm_addi("fp", "sp", frame_size);
    for (int i = 0; i < arg; ++i) {
        printf("  mv t0, a%d\n", i);
        store_t0(args[i]);
    }
    asm_j(label);
    // epilog
    asm_label(epilog_label);
    asm_ld("fp", frame_size - 16, "sp");
    asm_ld("ra", frame_size - 8, "sp");
    asm_addi("sp", "sp", frame_size);
    printf("  ret\n");
}

void parse_global_variable(int id, const char* name, int type) {
    if (type == TYPE_VOID) {
        eprintf("global variable of void type is not supported\n");
        exit(1);
    }
    if (type & TYPE_PTR_MASK) {
        eprintf("global variable of pointer is not supported\n");
        exit(1);
    }
    printf(".data\n");
    printf(".globl %s\n", name);
    printf(".align 5\n");
    printf("%s:\n", name);
    if (token_type == TOKEN_ASSIGN) {
        expect_token(TOKEN_NUMBER);
        printf("  .word %d\n", token_data);
    } else if (token_type == TOKEN_BRACKET_LEFT) {
        expect_token(TOKEN_NUMBER);
        int size = token_data;
        expect_token(TOKEN_BRACKET_RIGHT);
        int array_size = 4 * size;
        if (type == TYPE_CHAR) {
            array_size = size;
        }
        printf("  .zero %d\n", array_size);
        declare_global(id, MARKER_ARRAY, type);
    } else {
        printf("  .zero %d\n", 4);
        unget_token();
    }
    expect_token(TOKEN_SEMICOLON);
}

void parse_global_declaration() {
    int type = parse_type();
    if (type < 0) {
        eprintf("expecting type for global declaration\n");
        exit(1);
    }
    expect_token(TOKEN_ID);
    int id = token_data;
    char* name = id_table + id_lut[id];
    next_token();
    if (token_type == TOKEN_PAREN_LEFT) {
        declare_global(id, MARKER_SCALAR, type | TYPE_FUNC_MASK); 
        parse_function(name);
    } else {
        declare_global(id, MARKER_SCALAR, type); 
        parse_global_variable(id, name, type);
    }
}

void parse_top_level() {
    next_token();
    if (token_type == TOKEN_EOF) 
        return;
    parse_global_declaration();
    parse_top_level();
}

void dump_string_table() {
    printf(".data\n");
    for (int i = 0; i < string_lut_size; ++i) {
        printf(".LC%d: .string \"", i);
        int offset = string_lut[i];
        int ch; 
        while ((ch = string_table[offset++]) != 0) {
            if (ch == '\n') {
                printf("\\n");
            } else if (ch == '\t') {
                printf("\\t");
            } else if (ch == '\r') {
                printf("\\r");
            } else if (ch == '\0') {
                printf("\\0");
            } else if (ch == '\\') {
                printf("\\\\");
            } else if (ch == '\'') {
                printf("\\'");
            } else if (ch == '\"') {
                printf("\\\"");
            } else {
                printf("%c", ch);
            }
        }
        printf("\"\n");
    }
}

int main() {
    parse_top_level();
    dump_string_table();
    return 0;
}