creole/creole.c

#include "creole.h"

/*************************************************************************
 * Static information
 ************************************************************************/

/* Arguments to opcodes can accept the following:
   * immediate values only (as of now, no values are like this)
   * register values only (push, pop, etc.)
   * either values (math operations)
   * labels (jumps)
   * none (do not give an argument)
 */
enum creole_arg_type {
	TYPE_NONE,
	TYPE_IMM,
	TYPE_REG,
	TYPE_VAL,
	TYPE_LAB,
	CREOLE_ARG_TYPE_LEN
};

/* C99+ allows for designating the array index when initializing arrays:
      [i] = v,
 * in C89 indicies are implicit from 0 to the maximum filled-in value.
 */
#define defop(s, n, a1, a2, a3) {n, {a1, a2, a3}}
static const struct {
	unsigned arglen;
	enum creole_arg_type argtype[CREOLE_MAX_ARG];
} opcode_info[CREOLE_OPCODE_LEN] = {
	defop(NOOP, 0, TYPE_NONE, TYPE_NONE, TYPE_NONE),
	defop(PUSH, 1, TYPE_VAL, TYPE_NONE, TYPE_NONE),
	defop(POP, 1, TYPE_REG, TYPE_NONE, TYPE_NONE),
	defop(ADD, 3, TYPE_REG, TYPE_VAL, TYPE_VAL),
	defop(MUL, 3, TYPE_REG, TYPE_VAL, TYPE_VAL),
	defop(DIV, 3, TYPE_REG, TYPE_VAL, TYPE_VAL),
	defop(JL, 3, TYPE_LAB, TYPE_VAL, TYPE_VAL),
	defop(CLB, 1, TYPE_LAB, TYPE_NONE, TYPE_NONE),
	defop(SYS, 1, TYPE_VAL, TYPE_NONE, TYPE_NONE)
};

/*************************************************************************
 * Reading from the buffer
 ************************************************************************/

static int read(struct creole_reader *r)
{
	if (r->left == 0)
		return -1;
	r->left--;
	return *r->p++;
}

static int read_eof(struct creole_reader *r)
{
	return r->left == 0;
}

/*************************************************************************
 * Pseudo-UTF-8 lexing
 *
 * Pseudo-UTF-8 is based off of UTF-8 but adds more
 * bytes and allows (requires!) overlong encodings.
 *
 * Possible values:
 *   0xxxxxxx                                              (7 bits)
 *   110HHHHx 10xxxxxx                                     (11 bits)
 *   1110HHHH 10xxxxxx 10xxxxxx                            (16 bits)
 *   11110HHH 10Hxxxxx 10xxxxxx 10xxxxxx                   (21 bits)
 *   111110HH 10HHxxxx 10xxxxxx 10xxxxxx 10xxxxxx          (26 bits)
 *   1111110H 10HHHxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx (31 bits)
 *   11111110 10HHHHxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx (36 bits)
 *            10xxxxxx
 ************************************************************************/

/* A Psuedo-UTF-8 sequence can be either
 *
 * * A 1 byte sequence, where the lower 7 bits are the encoded
 * * word (no high bits).

 * * A multi-byte sequence where the 4 MSB are flags, and the
 * * lower bits are the encoded word.
 */
#define MAX_HIGH_BITS 15
struct word {
	int len;
	int high_bits;
	creole_word word;
};

/* Decode a set of continuation bytes directly into the word. This assumes
 * that each continuation byte contains no high words.
 */
static int read_continue(struct creole_reader *r, struct word *w,
                         int to_read)
{
	int i;
	int r_ret;
	unsigned char c;

	for (i = 0; i < to_read; i++) {
		r_ret = read(r);
		if (r_ret < 0) {
			return 0;
		}
		/* Characters might not be 8 bits! */
		c = (unsigned char)(r_ret & 0xFF);
		if (c >> 6 != 0x2) {
			return 0;
		}
		w->word = (w->word << 6) | (c & 0x3F);
	}

	return 1;
}

/* Start bytes must be treated differently. Depending on the scenario,
 * start bytes will contain parts of the encoded word and high-bit flags.
 * In some cases, not all of the high-bit flags are part of the start
 * byte.
 */
#define START_BYTE_NUM 7
static int parse_start_byte(unsigned char c, struct word *w)
{
	static const struct {
		/* The algorithm compares the mask to the start byte
		 * by shifting both to the right by the amount of 'x's
		 * (blank spaces). The array is arranged in reverse
		 * order so that the index indicates the amount of
		 * bits to shift.
		 */
		unsigned char mask;

		/* The word bits, if they exist, always start from the
		 * LSB, so there is no need to shift the bits away. The
		 * word_mask gets the low bits. If there are no bits, set
		 * to 0.
		 */
		unsigned char word_mask;

		/* The high bits may not start from the LSB. There needs
		 * to be a shift to get the bits to the LSB, and a mask
		 * to discard the higher bits.
		 */
		unsigned char high_bit_mask;
		int high_bit_shift;

		/* The amount of NORMAL continuation bytes to read.
		 * This does NOT include continuation bytes that have
		 * high-bit flags in them.
		 */
		int to_read;
	} start_data[START_BYTE_NUM-1] = {
		{0xFE, 0x00, 0x0, 0, 5}, /* 11111110 */
		{0xFC, 0x00, 0x1, 0, 4}, /* 1111110x */
		{0xF8, 0x00, 0x3, 0, 3}, /* 111110xx */
		{0xF0, 0x00, 0x7, 0, 2}, /* 11110xxx */
		{0xE0, 0x00, 0xF, 0, 2}, /* 1110xxxx */
		{0xC0, 0x01, 0xF, 1, 1}  /* 110xxxxx */
	};

	int i;

	for (i = 0; i < START_BYTE_NUM-1; i++) {
		if (c >> i == start_data[i].mask >> i) {
			w->len = START_BYTE_NUM - i;
			w->word = c & start_data[i].word_mask;
			w->high_bits = (c >> start_data[i].high_bit_shift)
			             & start_data[i].high_bit_mask;
			return start_data[i].to_read;
		}
	}
	/* i == 7 */
	if (c >> 7 == 0) {
		w->len = 1;
		w->word = c;
		w->high_bits = 0;
		return 0;
	}

	return -1;
}

/* This parses the first continuation byte if it is special. */
#define SPECIAL_CONTINUE_BYTE_NUM (START_BYTE_NUM - 3)
static int parse_special_byte(unsigned char c, struct word *w)
{
	/* The index denotes the amount of high bits that were in
	 * the start byte. This is the amount that the stored value
	 * must be shifted.
	 *
	 * The amount of bits that must be shifted out in the continue
	 * byte increase with the index. The amount shifted is (i + 2).
	 *
 	 * Each value stored in the array is the mask applied after
	 * shifting the continue byte bits.
	 */
	static const unsigned char mask[SPECIAL_CONTINUE_BYTE_NUM] = {
		0x1, /* 11110HHH 10Hxxxxx */
		0x3, /* 111110HH 10HHxxxx */
		0x7, /* 1111110H 10HHHxxx */
		0xF  /* 11111110 10HHHHxx */
	};
	static const unsigned char wordmask[SPECIAL_CONTINUE_BYTE_NUM] = {
		0x1F, 0xF, 0x7, 0x3
	};

	int i = w->len - 4;
	if (i >= SPECIAL_CONTINUE_BYTE_NUM)
		return 0;

	w->high_bits = (w->high_bits << (i + 1)) | ((c >> (5 - i)) & mask[i]);
	w->word = c & wordmask[i];
	return 1;
}

/* Parse an entire Pseudo-UTF8 sequence. */
static int decode_seq(struct creole_reader *r, struct word *w)
{
	int r_ret;
	int to_read;
	w->high_bits = 0;

	r_ret = read(r);
	if (r_ret < 0)
		return 0;

	to_read = parse_start_byte((unsigned char)(r_ret & 0xFF), w);
	if (to_read < 0)
		return 0;

	/* If to_read is not one less than w->len, that means there are
	 * high bits in the first continuation byte.
	 */
	if (w->len - to_read > 1) {
		r_ret = read(r);
		if (r_ret < 0)
			return 0;
		if (!parse_special_byte((unsigned char)(r_ret & 0xFF), w))
			return 0;
	}

	return read_continue(r, w, to_read);
}

int creole_encode(creole_word i, unsigned encode_to, unsigned high_bits,
                  unsigned char buf[7])
{
	static const struct {
		creole_word max;
		unsigned char b1_mask;
		int high_bit_shift_b1;
		int high_bit_shift_to_right_b1;
		int data_shift_b1;

		int high_bit_mask_b2;
		int high_bit_shift_b2;
		unsigned char b2_data_mask;
	} d[] = {
		{0x7F,       0xC0, 0, 1,  6, 0x0, 0, 0x3F}, /* 2 */
		{0xFFF,      0xE0, 0, 0, 12, 0x0, 0, 0x3F}, /* 3 */
		{0x1FFFF,    0xF0, 1, 0, 17, 0x1, 5, 0x1F}, /* 4 */
		{0x3FFFFF,   0xF8, 2, 0, 22, 0x3, 4, 0x0F}, /* 5 */
		{0x7FFFFFF,  0xFC, 3, 0, 27, 0x7, 3, 0x07}, /* 6 */
		{0xFFFFFFFF, 0xFE, 4, 0, 32, 0xF, 2, 0x03}  /* 7 */
	};
	int lb;
	unsigned j;

	if (encode_to > 8)
		return 0;

	if (encode_to == 1) {
		if (i < 0x80) {
			buf[0] = i;
			return 1;
		}
		return 0;
	}

	lb = encode_to - 2;
	if (i > d[lb].max) {
		return 0;
	}

	buf[0] = (d[lb].b1_mask | (high_bits >> d[lb].high_bit_shift_b1
	                           << d[lb].high_bit_shift_to_right_b1));
	/* shifts greater than or equal to the bit size of a type are
	 * undefined. Data in the first byte is always aligned with the LSB.
	 */
	if (d[lb].data_shift_b1 < sizeof(i) * CHAR_BIT)
	       buf[0] |= i >> d[lb].data_shift_b1;

	buf[1] = 0x80 | ((high_bits & d[lb].high_bit_mask_b2)
	                 << d[lb].high_bit_shift_b2)
	              | ((i >> ((encode_to - 2) * 6))
	                 & d[lb].b2_data_mask);

	for (j = 2; j < encode_to; j++) {
		buf[j] = 0x80 | ((i >> ((encode_to - j - 1) * 6)) & 0x3F);
	}

	return 1;
}

/*************************************************************************
 * Parsing instructions
 *
 * This parses an entire instruction, which is
 *  a single byte sequence,
 *  zero or more multibyte sequences,
 *  one single byte of all zeros.
 *************************************************************************/

enum creole_compiler_ret
creole_parse_line(struct creole_ins *ins, struct creole_reader *r)
{
	struct word w = {0};
	unsigned arg = 0;

	if (!decode_seq(r, &w))
		return CREOLE_OPCODE_READ_ERROR;

	ins->opcode = w.word;
	if (w.word >= CREOLE_ARG_TYPE_LEN || w.len != 1)
		return CREOLE_OPCODE_MALFORMED;

	for (arg = 0; arg < opcode_info[ins->opcode].arglen; arg++) {
		if (!decode_seq(r, &w))
			return CREOLE_ARG_READ_ERROR;
		if (w.len == 1)
			return CREOLE_ARG_MALFORMED;
		ins->w[arg] = w.word;
		ins->w_flags[arg] = w.high_bits;
	}

	if (!decode_seq(r, &w))
		return CREOLE_LAST_READ_ERROR;
	if (w.word != 0 || w.len != 1)
		return CREOLE_LAST_MALFORMED;
	return CREOLE_COMPILE_OK;
}

/**************************************************************************
 * High level compiling interface
 *************************************************************************/

static int typecheck(enum creole_word_flag fl, enum creole_arg_type typ)
{
	switch (typ) {
	case TYPE_IMM: return fl == CREOLE_IMMEDIATE;
	case TYPE_REG: return fl == CREOLE_REGISTER;
	case TYPE_VAL: return fl == CREOLE_IMMEDIATE
	                   || fl == CREOLE_REGISTER;
	case TYPE_LAB: return fl == CREOLE_IMMEDIATE;
	default: return 0;
	}
}

static enum creole_compiler_ret typecheck_ins(struct creole_env *env,
                                              struct creole_ins *ins)
{
	unsigned i;

	for (i = 0; i < opcode_info[ins->opcode].arglen; i++) {
		if (!typecheck(ins->w_flags[i],
		               opcode_info[ins->opcode].argtype[i]))
			return CREOLE_TYPE_ERROR;
	}
	return CREOLE_COMPILE_OK;
}

static void clear_ins(struct creole_ins *i)
{
	i->opcode = 0;
	i->w[0] = i->w[1] = i->w[2] = i->w_flags[0] =
	          i->w_flags[1] = i->w_flags[2] = 0;
}

static enum creole_compiler_ret
handle_compiletime_immediate(struct creole_env *env,
                             struct creole_ins *cur_ins)
{
	switch (cur_ins->opcode) {
	case CREOLE_CLB:
		if (cur_ins->w[0] >= env->lablen)
			return CREOLE_LABEL_OVERFLOW;
		env->lab[cur_ins->w[0]] = env->prgptr;
		/* Delete instruction because it is a compile time
		 * instruction. Place next instruction in its place. */
		clear_ins(cur_ins);
		return CREOLE_COMPILE_CLEARED_INSTRUCTION;
	case CREOLE_NOOP:
		clear_ins(cur_ins);
		return CREOLE_COMPILE_CLEARED_INSTRUCTION;
	default:
		return typecheck_ins(env, cur_ins);
	}
}

enum creole_compiler_ret
creole_compile(struct creole_env *env, struct creole_reader *r)
{
	struct creole_ins *cur_ins = env->prg;
	int rcode;

	while (env->prgptr < env->prglen) {
		rcode = creole_parse_line(cur_ins, r);
		if (rcode != CREOLE_COMPILE_OK)
			return rcode;

		rcode = handle_compiletime_immediate(env, cur_ins);
		switch (rcode) {
		case CREOLE_COMPILE_CLEARED_INSTRUCTION:
			break;
		case CREOLE_COMPILE_OK:
			cur_ins++;
			env->prgptr++;
			break;
		default:
			return rcode;
		}

		if (read_eof(r))
			break;
	}

	if (env->prgptr == env->prglen && !read_eof(r))
		return CREOLE_PROGRAM_OVERFLOW;
	env->prgend = env->prgptr;
	env->prgptr = 0;
	return CREOLE_COMPILE_OK;
}

#if 0

static creole_word read_word(struct creole_ins *ins, int i)
{
	if (env->w_flags[i] == CREOLE_REGISTER)
		return env->reg[env->w[i]];
	else
		return env->w[i];
}

int creole_step(struct creole_env *env)
{
	struct creole_ins *ins = env->prg + env->prgptr;
	creole_word a1, a2;

	if (env->prgptr == env->prgend)
		return CREOLE_STEP_STOP;
	env->prgptr++;

	switch (ins->opcode) {
	case CREOLE_PUSH:
		if (env->stkptr == env->stklen)
			return CREOLE_STACK_OVERFLOW;
		env->stk[env->stkptr++] = env->reg[env->w[0]];
		break;
	case CREOLE_POP:
		if (env->stkptr == 0)
			return CREOLE_STACK_OVERFLOW;
		env->reg[env->w[0]] = env->stk[--env->stkptr];
		break;
	case CREOLE_ADD:
		a1 = read_word(ins, 1);
		a2 = read_word(ins, 2);
		env->reg[env->w[0]] = a1 + a2;
		break;
	case CREOLE_MUL:
		a1 = read_word(ins, 1);
		a2 = read_word(ins, 2);
		env->reg[env->w[0]] = a1 * a2;
		break;
	case CREOLE_DIV:
		a1 = read_word(ins, 1);
		a2 = read_word(ins, 2);
		env->reg[env->w[0]] = a1 / a2;
		break;
	case CREOLE_JL:
		a1 = read_word(ins, 1);
		a2 = read_word(ins, 2);
		if (a1 < a2)
			env->prgptr = env->lab[env->w[0]];
		break;
	case SYS:
		a1 = read_word(ins, 1);
		/* do syscall */
		break;
	}
}
#endif