jerryscript/jerry-core/ecma/base/ecma-helpers-conversion.c
Akos Kiss d501c92f96 Eliminate TODO and FIXME macros
Those macros are legacy and are not used consitently throughout the
code base. This patch eliminates their definitions and rewrites
their remaining occurrences to TODO comments.

All occurrences have been checked and made sure that the comments
used a consistent style.

JerryScript-DCO-1.0-Signed-off-by: Akos Kiss akiss@inf.u-szeged.hu
2016-04-06 14:40:43 +02:00

1484 lines
40 KiB
C

/* Copyright 2014-2016 Samsung Electronics Co., Ltd.
* Copyright 2016 University of Szeged.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/** \addtogroup ecma ECMA
* @{
*
* \addtogroup ecmahelpers Helpers for operations with ECMA data types
* @{
*/
#include "ecma-globals.h"
#include "ecma-helpers.h"
#include "jrt-libc-includes.h"
#include "lit-char-helpers.h"
#include "lit-magic-strings.h"
/*
* \addtogroup ecmahelpersbigintegers Helpers for operations intermediate 128-bit integers
* @{
*/
/**
* Check that parts of 128-bit integer are 32-bit.
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT(name) \
{ \
JERRY_ASSERT (name[0] == (uint32_t) name[0]); \
JERRY_ASSERT (name[1] == (uint32_t) name[1]); \
JERRY_ASSERT (name[2] == (uint32_t) name[2]); \
JERRY_ASSERT (name[3] == (uint32_t) name[3]); \
}
/**
* Declare 128-bit integer.
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER(name) uint64_t name[4] = { 0, 0, 0, 0 }
/**
* Declare 128-bit in-out argument integer.
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ARG(name) uint64_t name[4]
/**
* Initialize 128-bit integer with given 32-bit parts
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INIT(name, high, mid_high, mid_low, low) \
{ \
name[3] = high; \
name[2] = mid_high; \
name[1] = mid_low; \
name[0] = low; \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
}
/**
* Copy specified 128-bit integer
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_COPY(name_copy_to, name_copy_from) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name_copy_to); \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name_copy_from); \
\
name_copy_to[0] = name_copy_from[0]; \
name_copy_to[1] = name_copy_from[1]; \
name_copy_to[2] = name_copy_from[2]; \
name_copy_to[3] = name_copy_from[3]; \
}
/**
* Copy high and middle parts of 128-bit integer to specified uint64_t variable
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ROUND_HIGH_AND_MIDDLE_TO_UINT64(name, uint64_var) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
uint64_var = ((name[3] << 32u) | (name[2])) + (((name[1] >> 31u) != 0 ? 1 : 0)); \
}
/**
* Copy middle and low parts of 128-bit integer to specified uint64_t variable
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ROUND_MIDDLE_AND_LOW_TO_UINT64(name, uint64_var) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
uint64_var = (name[1] << 32u) | (name[0]); \
}
/**
* Check if specified 128-bit integers are equal
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ARE_EQUAL(name1, name2) \
((name1)[0] == (name2[0]) \
&& (name1)[1] == (name2[1]) \
&& (name1)[2] == (name2[2]) \
&& (name1)[3] == (name2[3]))
/**
* Check if bits [lowest_bit, 128) are zero
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO(name, lowest_bit) \
((lowest_bit) >= 96 ? ((name[3] >> ((lowest_bit) - 96)) == 0) : \
((lowest_bit) >= 64 ? (name[3] == 0 \
&& ((name[2] >> ((lowest_bit) - 64)) == 0)) : \
((lowest_bit) >= 32 ? (name[3] == 0 \
&& name[2] == 0 && ((name[1] >> ((lowest_bit) - 32)) == 0)) : \
(name[3] == 0 && name[2] == 0 && name[1] == 0 && ((name[0] >> (lowest_bit)) == 0)))))
/**
* Check if bits [0, highest_bit] are zero
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_LOW_BIT_MASK_ZERO(name, highest_bit) \
((highest_bit >= 96) ? (name[2] == 0 && name[1] == 0 && name[0] == 0 \
&& (((uint32_t) name[3] << (127 - (highest_bit))) == 0)) : \
((highest_bit >= 64) ? (name[1] == 0 && name[0] == 0 \
&& (((uint32_t) name[2] << (95 - (highest_bit))) == 0)) : \
((highest_bit >= 32) ? (name[0] == 0 \
&& (((uint32_t) name[1] << (63 - (highest_bit))) == 0)) : \
(((uint32_t) name[0] << (31 - (highest_bit))) == 0))))
/**
* Check if 128-bit integer is zero
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO(name) \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (name, 0)
/**
* Shift 128-bit integer one bit left
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT(name) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
\
name[3] = (uint32_t) (name[3] << 1u); \
name[3] |= name[2] >> 31u; \
name[2] = (uint32_t) (name[2] << 1u); \
name[2] |= name[1] >> 31u; \
name[1] = (uint32_t) (name[1] << 1u); \
name[1] |= name[0] >> 31u; \
name[0] = (uint32_t) (name[0] << 1u); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
}
/**
* Shift 128-bit integer one bit right
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT(name) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
\
name[0] >>= 1u; \
name[0] |= (uint32_t) (name[1] << 31u); \
name[1] >>= 1u; \
name[1] |= (uint32_t) (name[2] << 31u); \
name[2] >>= 1u; \
name[2] |= (uint32_t) (name[3] << 31u); \
name[3] >>= 1u; \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
}
/**
* Increment 128-bit integer
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INC(name) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
\
name[0] += 1ull; \
name[1] += (name[0] >> 32u); \
name[0] = (uint32_t) name[0]; \
name[2] += (name[1] >> 32u); \
name[1] = (uint32_t) name[1]; \
name[3] += (name[2] >> 32u); \
name[2] = (uint32_t) name[2]; \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
}
/**
* Add 128-bit integer
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ADD(name_add_to, name_to_add) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name_add_to); \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name_to_add); \
\
name_add_to[0] += name_to_add[0]; \
name_add_to[1] += name_to_add[1]; \
name_add_to[2] += name_to_add[2]; \
name_add_to[3] += name_to_add[3]; \
\
name_add_to[1] += (name_add_to[0] >> 32u); \
name_add_to[0] = (uint32_t) name_add_to[0]; \
name_add_to[2] += (name_add_to[1] >> 32u); \
name_add_to[1] = (uint32_t) name_add_to[1]; \
name_add_to[3] += (name_add_to[2] >> 32u); \
name_add_to[2] = (uint32_t) name_add_to[2]; \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name_add_to); \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name_to_add); \
}
/**
* Multiply 128-bit integer by 10
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10(name) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (name); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER (name ## _tmp); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_COPY (name ## _tmp, name); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (name ## _tmp); \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (name ## _tmp); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ADD (name, name ## _tmp); \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
}
/**
* Divide 128-bit integer by 10
*/
#define ECMA_NUMBER_CONVERSION_128BIT_INTEGER_DIV_10(name) \
{ \
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
\
/* estimation of reciprocal of 10 */ \
const uint64_t div10_p_low = 0x9999999aul; \
const uint64_t div10_p_mid = 0x99999999ul; \
const uint64_t div10_p_high = 0x19999999ul; \
\
uint64_t intermediate[8] = { 0, 0, 0, 0, 0, 0, 0, 0 }; \
uint64_t l0, l1, l2, l3, m0, m1, m2, m3, h0, h1, h2, h3; \
l0 = name[0] * div10_p_low; \
l1 = name[1] * div10_p_low; \
l2 = name[2] * div10_p_low; \
l3 = name[3] * div10_p_low; \
m0 = name[0] * div10_p_mid; \
m1 = name[1] * div10_p_mid; \
m2 = name[2] * div10_p_mid; \
m3 = name[3] * div10_p_mid; \
h0 = name[0] * div10_p_high; \
h1 = name[1] * div10_p_high; \
h2 = name[2] * div10_p_high; \
h3 = name[3] * div10_p_high; \
intermediate[0] += (uint32_t) l0; \
intermediate[1] += l0 >> 32u; \
\
intermediate[1] += (uint32_t) l1; \
intermediate[2] += l1 >> 32u; \
intermediate[1] += (uint32_t) m0; \
intermediate[2] += m0 >> 32u; \
\
intermediate[2] += (uint32_t) l2; \
intermediate[3] += l2 >> 32u; \
intermediate[2] += (uint32_t) m1; \
intermediate[3] += m1 >> 32u; \
intermediate[2] += (uint32_t) m0; \
intermediate[3] += m0 >> 32u; \
\
intermediate[3] += (uint32_t) l3; \
intermediate[4] += l3 >> 32u; \
intermediate[3] += (uint32_t) m2; \
intermediate[4] += m2 >> 32u; \
intermediate[3] += (uint32_t) m1; \
intermediate[4] += m1 >> 32u; \
intermediate[3] += (uint32_t) h0; \
intermediate[4] += h0 >> 32u; \
\
intermediate[4] += (uint32_t) m3; \
intermediate[5] += m3 >> 32u; \
intermediate[4] += (uint32_t) m2; \
intermediate[5] += m2 >> 32u; \
intermediate[4] += (uint32_t) h1; \
intermediate[5] += h1 >> 32u; \
\
intermediate[5] += (uint32_t) m3; \
intermediate[6] += m3 >> 32u; \
intermediate[5] += (uint32_t) h2; \
intermediate[6] += h2 >> 32u; \
\
intermediate[6] += (uint32_t) h3; \
intermediate[7] += h3 >> 32u; \
\
intermediate[1] += intermediate[0] >> 32u; \
intermediate[0] = (uint32_t) intermediate[0]; \
intermediate[2] += intermediate[1] >> 32u; \
intermediate[1] = (uint32_t) intermediate[1]; \
intermediate[3] += intermediate[2] >> 32u; \
intermediate[2] = (uint32_t) intermediate[2]; \
intermediate[4] += intermediate[3] >> 32u; \
intermediate[3] = (uint32_t) intermediate[3]; \
intermediate[5] += intermediate[4] >> 32u; \
intermediate[4] = (uint32_t) intermediate[4]; \
intermediate[6] += intermediate[5] >> 32u; \
intermediate[5] = (uint32_t) intermediate[5]; \
intermediate[7] += intermediate[6] >> 32u; \
intermediate[6] = (uint32_t) intermediate[6]; \
\
name[0] = intermediate[4]; \
name[1] = intermediate[5]; \
name[2] = intermediate[6]; \
name[3] = intermediate[7]; \
\
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_CHECK_PARTS_ARE_32BIT (name); \
}
/**
* @}
*/
/**
* ECMA-defined conversion of string to Number.
*
* See also:
* ECMA-262 v5, 9.3.1
*
* @return ecma-number
*/
ecma_number_t
ecma_utf8_string_to_number (const lit_utf8_byte_t *str_p, /**< utf-8 string */
lit_utf8_size_t str_size) /**< string size */
{
/* TODO: Check license issues */
const lit_utf8_byte_t dec_digits_range[10] = { '0', '9' };
const lit_utf8_byte_t hex_lower_digits_range[10] = { 'a', 'f' };
const lit_utf8_byte_t hex_upper_digits_range[10] = { 'A', 'F' };
const lit_utf8_byte_t hex_x_chars[2] = { 'x', 'X' };
const lit_utf8_byte_t e_chars[2] = { 'e', 'E' };
const lit_utf8_byte_t plus_char = '+';
const lit_utf8_byte_t minus_char = '-';
const lit_utf8_byte_t dot_char = '.';
if (str_size == 0)
{
return ECMA_NUMBER_ZERO;
}
lit_utf8_byte_t *str_curr_p = (lit_utf8_byte_t *) str_p;
const lit_utf8_byte_t *str_end_p = str_p + str_size;
ecma_char_t code_unit;
while (str_curr_p < str_end_p)
{
code_unit = lit_utf8_peek_next (str_curr_p);
if (lit_char_is_white_space (code_unit) || lit_char_is_line_terminator (code_unit))
{
lit_utf8_incr (&str_curr_p);
}
else
{
break;
}
}
const lit_utf8_byte_t *begin_p = str_curr_p;
str_curr_p = (lit_utf8_byte_t *) str_end_p;
while (str_curr_p > str_p)
{
code_unit = lit_utf8_peek_prev (str_curr_p);
if (lit_char_is_white_space (code_unit) || lit_char_is_line_terminator (code_unit))
{
lit_utf8_decr (&str_curr_p);
}
else
{
break;
}
}
const lit_utf8_byte_t *end_p = str_curr_p - 1;
if (begin_p > end_p)
{
return ECMA_NUMBER_ZERO;
}
if ((end_p >= begin_p + 2)
&& begin_p[0] == dec_digits_range[0]
&& (begin_p[1] == hex_x_chars[0]
|| begin_p[1] == hex_x_chars[1]))
{
/* Hex literal handling */
begin_p += 2;
ecma_number_t num = ECMA_NUMBER_ZERO;
for (const lit_utf8_byte_t * iter_p = begin_p;
iter_p <= end_p;
iter_p++)
{
int32_t digit_value;
if (*iter_p >= dec_digits_range[0]
&& *iter_p <= dec_digits_range[1])
{
digit_value = (*iter_p - dec_digits_range[0]);
}
else if (*iter_p >= hex_lower_digits_range[0]
&& *iter_p <= hex_lower_digits_range[1])
{
digit_value = 10 + (*iter_p - hex_lower_digits_range[0]);
}
else if (*iter_p >= hex_upper_digits_range[0]
&& *iter_p <= hex_upper_digits_range[1])
{
digit_value = 10 + (*iter_p - hex_upper_digits_range[0]);
}
else
{
return ecma_number_make_nan ();
}
num = num * 16 + (ecma_number_t) digit_value;
}
return num;
}
bool sign = false; /* positive */
if (*begin_p == plus_char)
{
begin_p++;
}
else if (*begin_p == minus_char)
{
sign = true; /* negative */
begin_p++;
}
if (begin_p > end_p)
{
return ecma_number_make_nan ();
}
/* Checking if significant part of parse string is equal to "Infinity" */
const lit_utf8_byte_t *infinity_zt_str_p = lit_get_magic_string_utf8 (LIT_MAGIC_STRING_INFINITY_UL);
for (const lit_utf8_byte_t *iter_p = begin_p, *iter_infinity_p = infinity_zt_str_p;
;
iter_infinity_p++, iter_p++)
{
if (*iter_p != *iter_infinity_p)
{
break;
}
if (iter_p == end_p)
{
return ecma_number_make_infinity (sign);
}
}
uint64_t fraction_uint64 = 0;
uint32_t digits = 0;
int32_t e = 0;
/* Parsing digits before dot (or before end of digits part if there is no dot in number) */
while (begin_p <= end_p)
{
int32_t digit_value;
if (*begin_p >= dec_digits_range[0]
&& *begin_p <= dec_digits_range[1])
{
digit_value = (*begin_p - dec_digits_range[0]);
}
else
{
break;
}
if (digits != 0 || digit_value != 0)
{
if (digits < ECMA_NUMBER_MAX_DIGITS)
{
fraction_uint64 = fraction_uint64 * 10 + (uint32_t) digit_value;
digits++;
}
else if (e <= 100000) /* Some limit to not overflow exponent value
(so big exponent anyway will make number
rounded to infinity) */
{
e++;
}
}
begin_p++;
}
if (begin_p <= end_p
&& *begin_p == dot_char)
{
begin_p++;
/* Parsing number's part that is placed after dot */
while (begin_p <= end_p)
{
int32_t digit_value;
if (*begin_p >= dec_digits_range[0]
&& *begin_p <= dec_digits_range[1])
{
digit_value = (*begin_p - dec_digits_range[0]);
}
else
{
break;
}
if (digits < ECMA_NUMBER_MAX_DIGITS)
{
if (digits != 0 || digit_value != 0)
{
fraction_uint64 = fraction_uint64 * 10 + (uint32_t) digit_value;
digits++;
}
e--;
}
begin_p++;
}
}
/* Parsing exponent literal */
int32_t e_in_lit = 0;
bool e_in_lit_sign = false;
if (begin_p <= end_p
&& (*begin_p == e_chars[0]
|| *begin_p == e_chars[1]))
{
begin_p++;
if (*begin_p == plus_char)
{
begin_p++;
}
else if (*begin_p == minus_char)
{
e_in_lit_sign = true;
begin_p++;
}
if (begin_p > end_p)
{
return ecma_number_make_nan ();
}
while (begin_p <= end_p)
{
int32_t digit_value;
if (*begin_p >= dec_digits_range[0]
&& *begin_p <= dec_digits_range[1])
{
digit_value = (*begin_p - dec_digits_range[0]);
}
else
{
return ecma_number_make_nan ();
}
e_in_lit = e_in_lit * 10 + digit_value;
begin_p++;
}
}
/* Adding value of exponent literal to exponent value */
if (e_in_lit_sign)
{
e -= e_in_lit;
}
else
{
e += e_in_lit;
}
bool e_sign;
if (e < 0)
{
e_sign = true;
e = -e;
}
else
{
e_sign = false;
}
if (begin_p <= end_p)
{
return ecma_number_make_nan ();
}
JERRY_ASSERT (begin_p == end_p + 1);
if (fraction_uint64 == 0)
{
return sign ? -ECMA_NUMBER_ZERO : ECMA_NUMBER_ZERO;
}
#if CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT64
int32_t binary_exponent = 33;
/*
* 128-bit mantissa storage
*
* Normalized: |4 bits zero|124-bit mantissa with highest bit set to 1 if mantissa is non-zero|
*/
ECMA_NUMBER_CONVERSION_128BIT_INTEGER (fraction_uint128);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INIT (fraction_uint128,
0ull,
fraction_uint64 >> 32u,
(uint32_t) fraction_uint64,
0ull);
/* Normalizing mantissa */
JERRY_ASSERT (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124));
while (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 123))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (fraction_uint128);
binary_exponent--;
JERRY_ASSERT (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128));
}
if (!e_sign)
{
/* positive or zero decimal exponent */
JERRY_ASSERT (e >= 0);
while (e > 0)
{
JERRY_ASSERT (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124));
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10 (fraction_uint128);
e--;
/* Normalizing mantissa */
while (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128);
binary_exponent++;
}
while (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 123))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (fraction_uint128);
binary_exponent--;
JERRY_ASSERT (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128));
}
}
}
else
{
/* negative decimal exponent */
JERRY_ASSERT (e != 0);
while (e > 0)
{
/* Denormalizing mantissa, moving highest 1 to 95-bit */
while (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 127))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (fraction_uint128);
binary_exponent--;
JERRY_ASSERT (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128));
}
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_DIV_10 (fraction_uint128);
e--;
}
/* Normalizing mantissa */
while (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128);
binary_exponent++;
}
while (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 123))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (fraction_uint128);
binary_exponent--;
JERRY_ASSERT (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128));
}
}
JERRY_ASSERT (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128));
JERRY_ASSERT (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124));
/*
* Preparing mantissa for conversion to 52-bit representation, converting it to:
*
* |12 zero bits|116 mantissa bits|
*/
while (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 116 + 1))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128);
binary_exponent++;
}
while (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 116))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (fraction_uint128);
binary_exponent--;
JERRY_ASSERT (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128));
}
JERRY_ASSERT (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 116 + 1));
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ROUND_HIGH_AND_MIDDLE_TO_UINT64 (fraction_uint128, fraction_uint64);
return ecma_number_make_from_sign_mantissa_and_exponent (sign,
fraction_uint64,
binary_exponent);
#elif CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT32
/* Less precise conversion */
ecma_number_t num = (ecma_number_t) (uint32_t) fraction_uint64;
ecma_number_t m = e_sign ? (ecma_number_t) 0.1 : (ecma_number_t) 10.0;
while (e)
{
if (e % 2)
{
num *= m;
}
m *= m;
e /= 2;
}
return num;
#endif /* CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT32 */
} /* ecma_utf8_string_to_number */
/**
* ECMA-defined conversion of UInt32 to String (zero-terminated).
*
* See also:
* ECMA-262 v5, 9.8.1
*
* @return number of bytes copied to buffer
*/
lit_utf8_size_t
ecma_uint32_to_utf8_string (uint32_t value, /**< value to convert */
lit_utf8_byte_t *out_buffer_p, /**< buffer for string */
lit_utf8_size_t buffer_size) /**< size of buffer */
{
const lit_utf8_byte_t digits[10] = { '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' };
lit_utf8_byte_t *p = out_buffer_p + buffer_size - 1;
lit_utf8_size_t bytes_copied = 0;
do
{
JERRY_ASSERT (p >= out_buffer_p);
*p-- = digits[value % 10];
value /= 10;
bytes_copied++;
}
while (value != 0);
p++;
JERRY_ASSERT (p >= out_buffer_p);
if (likely (p != out_buffer_p))
{
memmove (out_buffer_p, p, bytes_copied);
}
return bytes_copied;
} /* ecma_uint32_to_utf8_string */
/**
* ECMA-defined conversion of Number value to UInt32 value
*
* See also:
* ECMA-262 v5, 9.6
*
* @return 32-bit unsigned integer - result of conversion.
*/
uint32_t
ecma_number_to_uint32 (ecma_number_t num) /**< ecma-number */
{
if (ecma_number_is_nan (num)
|| ecma_number_is_zero (num)
|| ecma_number_is_infinity (num))
{
return 0;
}
const bool sign = ecma_number_is_negative (num);
const ecma_number_t abs_num = sign ? ecma_number_negate (num) : num;
// 2 ^ 32
const uint64_t uint64_2_pow_32 = (1ull << 32);
const ecma_number_t num_2_pow_32 = (float) uint64_2_pow_32;
ecma_number_t num_in_uint32_range;
if (abs_num >= num_2_pow_32)
{
num_in_uint32_range = ecma_number_calc_remainder (abs_num,
num_2_pow_32);
}
else
{
num_in_uint32_range = abs_num;
}
// Check that the floating point value can be represented with uint32_t
JERRY_ASSERT (num_in_uint32_range < uint64_2_pow_32);
uint32_t uint32_num = (uint32_t) num_in_uint32_range;
const uint32_t ret = sign ? -uint32_num : uint32_num;
#ifndef JERRY_NDEBUG
if (sign
&& uint32_num != 0)
{
JERRY_ASSERT (ret == uint64_2_pow_32 - uint32_num);
}
else
{
JERRY_ASSERT (ret == uint32_num);
}
#endif /* !JERRY_NDEBUG */
return ret;
} /* ecma_number_to_uint32 */
/**
* ECMA-defined conversion of Number value to Int32 value
*
* See also:
* ECMA-262 v5, 9.5
*
* @return 32-bit signed integer - result of conversion.
*/
int32_t
ecma_number_to_int32 (ecma_number_t num) /**< ecma-number */
{
uint32_t uint32_num = ecma_number_to_uint32 (num);
// 2 ^ 32
const int64_t int64_2_pow_32 = (1ll << 32);
// 2 ^ 31
const uint32_t uint32_2_pow_31 = (1ull << 31);
int32_t ret;
if (uint32_num >= uint32_2_pow_31)
{
ret = (int32_t) (uint32_num - int64_2_pow_32);
}
else
{
ret = (int32_t) uint32_num;
}
#ifndef JERRY_NDEBUG
int64_t int64_num = uint32_num;
JERRY_ASSERT (int64_num >= 0);
if (int64_num >= uint32_2_pow_31)
{
JERRY_ASSERT (ret == int64_num - int64_2_pow_32);
}
else
{
JERRY_ASSERT (ret == int64_num);
}
#endif /* !JERRY_NDEBUG */
return ret;
} /* ecma_number_to_int32 */
#if CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT64
/**
* Perform conversion of 128-bit binary representation of number
* to decimal representation with decimal exponent.
*/
static void
ecma_number_helper_binary_to_decimal (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ARG (fraction_uint128), /**< mantissa */
int32_t binary_exponent, /**< binary exponent */
int32_t *out_decimal_exp_p) /**< [out] decimal exponent */
{
int32_t decimal_exp = 0;
if (binary_exponent > 0)
{
while (binary_exponent > 0)
{
if (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INC (fraction_uint128);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128);
binary_exponent++;
}
else
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER (fraction_uint128_tmp);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_COPY (fraction_uint128_tmp, fraction_uint128);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_DIV_10 (fraction_uint128_tmp);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10 (fraction_uint128_tmp);
if (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ARE_EQUAL (fraction_uint128, fraction_uint128_tmp)
&& ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 123))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_LEFT_SHIFT (fraction_uint128);
binary_exponent--;
}
else
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_DIV_10 (fraction_uint128);
decimal_exp++;
}
}
}
}
else if (binary_exponent < 0)
{
while (binary_exponent < 0)
{
if (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_LOW_BIT_MASK_ZERO (fraction_uint128, 0)
|| !ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128, 124))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128);
binary_exponent++;
}
else
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10 (fraction_uint128);
decimal_exp--;
}
}
}
*out_decimal_exp_p = decimal_exp;
} /* ecma_number_helper_binary_to_decimal */
#endif /* CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT64 */
/**
* Perform conversion of ecma-number to decimal representation with decimal exponent
*
* Note:
* The calculated values correspond to s, n, k parameters in ECMA-262 v5, 9.8.1, item 5:
* - s represents digits of the number;
* - k is the number of digits;
* - n is the decimal exponent.
*/
void
ecma_number_to_decimal (ecma_number_t num, /**< ecma-number */
uint64_t *out_digits_p, /**< [out] digits */
int32_t *out_digits_num_p, /**< [out] number of digits */
int32_t *out_decimal_exp_p) /**< [out] decimal exponent */
{
JERRY_ASSERT (!ecma_number_is_nan (num));
JERRY_ASSERT (!ecma_number_is_zero (num));
JERRY_ASSERT (!ecma_number_is_infinity (num));
#if CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT64
ecma_number_t num_m1 = ecma_number_get_prev (num);
ecma_number_t num_p1 = ecma_number_get_next (num);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER (fraction_uint128);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER (fraction_uint128_m1);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER (fraction_uint128_p1);
uint64_t fraction_uint64, fraction_uint64_m1, fraction_uint64_p1;
int32_t binary_exponent, binary_exponent_m1, binary_exponent_p1;
int32_t decimal_exp, decimal_exp_m1, decimal_exp_p1;
int32_t dot_shift, dot_shift_m1, dot_shift_p1;
dot_shift_m1 = ecma_number_get_fraction_and_exponent (num_m1, &fraction_uint64_m1, &binary_exponent_m1);
dot_shift = ecma_number_get_fraction_and_exponent (num, &fraction_uint64, &binary_exponent);
dot_shift_p1 = ecma_number_get_fraction_and_exponent (num_p1, &fraction_uint64_p1, &binary_exponent_p1);
binary_exponent_m1 -= dot_shift_m1;
binary_exponent -= dot_shift;
binary_exponent_p1 -= dot_shift_p1;
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INIT (fraction_uint128,
0ull,
0ull,
(fraction_uint64) >> 32u,
((fraction_uint64) << 32u) >> 32u);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INIT (fraction_uint128_m1,
0ull,
0ull,
(fraction_uint64_m1) >> 32u,
((fraction_uint64_m1) << 32u) >> 32u);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_INIT (fraction_uint128_p1,
0ull,
0ull,
(fraction_uint64_p1) >> 32u,
((fraction_uint64_p1) << 32u) >> 32u);
ecma_number_helper_binary_to_decimal (fraction_uint128, binary_exponent, &decimal_exp);
ecma_number_helper_binary_to_decimal (fraction_uint128_m1, binary_exponent_m1, &decimal_exp_m1);
ecma_number_helper_binary_to_decimal (fraction_uint128_p1, binary_exponent_p1, &decimal_exp_p1);
if (ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_ZERO (fraction_uint128_m1))
{
decimal_exp_m1 = decimal_exp;
}
while (decimal_exp != decimal_exp_m1
|| decimal_exp != decimal_exp_p1)
{
while (decimal_exp > decimal_exp_m1
|| decimal_exp > decimal_exp_p1)
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10 (fraction_uint128);
decimal_exp--;
}
while (decimal_exp_m1 > decimal_exp
|| decimal_exp_m1 > decimal_exp_p1)
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10 (fraction_uint128_m1);
decimal_exp_m1--;
}
while (decimal_exp_p1 > decimal_exp
|| decimal_exp_p1 > decimal_exp_m1)
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_MUL_10 (fraction_uint128_p1);
decimal_exp_p1--;
}
}
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ADD (fraction_uint128_m1, fraction_uint128);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128_m1);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ADD (fraction_uint128_p1, fraction_uint128);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_RIGHT_SHIFT (fraction_uint128_p1);
/* While fraction doesn't fit to integer, divide it by 10
and simultaneously increment decimal exponent */
uint64_t digits_min, digits_max;
while (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128_m1, 63))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_DIV_10 (fraction_uint128_m1);
decimal_exp_m1++;
}
while (!ECMA_NUMBER_CONVERSION_128BIT_INTEGER_IS_HIGH_BIT_MASK_ZERO (fraction_uint128_p1, 63))
{
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_DIV_10 (fraction_uint128_p1);
decimal_exp_p1++;
}
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ROUND_MIDDLE_AND_LOW_TO_UINT64 (fraction_uint128_m1, digits_min);
ECMA_NUMBER_CONVERSION_128BIT_INTEGER_ROUND_MIDDLE_AND_LOW_TO_UINT64 (fraction_uint128_p1, digits_max);
digits_min++;
if (decimal_exp_m1 < decimal_exp_p1)
{
JERRY_ASSERT (decimal_exp_m1 == decimal_exp_p1 - 1);
digits_min /= 10;
decimal_exp_m1++;
}
else if (decimal_exp_m1 > decimal_exp_p1)
{
JERRY_ASSERT (decimal_exp_m1 == decimal_exp_p1 + 1);
digits_max /= 10;
decimal_exp_p1++;
}
JERRY_ASSERT (digits_max >= digits_min);
while (digits_min / 10 != digits_max / 10)
{
digits_min /= 10;
digits_max /= 10;
decimal_exp_m1++;
decimal_exp_p1++;
}
uint64_t digits = (digits_min + digits_max + 1) / 2;
int32_t digits_num = 0;
uint64_t t = digits;
while (t != 0)
{
t /= 10;
digits_num++;
}
JERRY_ASSERT (digits_num > 0);
*out_digits_p = digits;
*out_digits_num_p = digits_num;
*out_decimal_exp_p = decimal_exp_p1 + digits_num;
#elif CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT32
/* Less precise conversion */
uint64_t fraction_uint64;
uint32_t fraction;
int32_t exponent;
int32_t dot_shift;
int32_t decimal_exp = 0;
dot_shift = ecma_number_get_fraction_and_exponent (num, &fraction_uint64, &exponent);
fraction = (uint32_t) fraction_uint64;
JERRY_ASSERT (fraction == fraction_uint64);
if (exponent != 0)
{
ecma_number_t t = 1.0f;
bool do_divide;
if (exponent < 0)
{
do_divide = true;
while (exponent <= 0)
{
t *= 2.0f;
exponent++;
if (t >= 10.0f)
{
t /= 10.0f;
decimal_exp--;
}
JERRY_ASSERT (t < 10.0f);
}
while (t > 1.0f)
{
exponent--;
t /= 2.0f;
}
}
else
{
do_divide = false;
while (exponent >= 0)
{
t *= 2.0f;
exponent--;
if (t >= 10.0f)
{
t /= 10.0f;
decimal_exp++;
}
JERRY_ASSERT (t < 10.0f);
}
while (t > 2.0f)
{
exponent++;
t /= 2.0f;
}
}
if (do_divide)
{
fraction = (uint32_t) ((ecma_number_t) fraction / t);
}
else
{
fraction = (uint32_t) ((ecma_number_t) fraction * t);
}
}
uint32_t s;
int32_t n;
int32_t k;
if (exponent > 0)
{
fraction <<= exponent;
}
else
{
fraction >>= -exponent;
}
const int32_t int_part_shift = dot_shift;
const uint32_t frac_part_mask = ((((uint32_t) 1) << int_part_shift) - 1);
uint32_t int_part = fraction >> int_part_shift;
uint32_t frac_part = fraction & frac_part_mask;
s = int_part;
k = 1;
n = decimal_exp + 1;
JERRY_ASSERT (int_part < 10);
while (k < ECMA_NUMBER_MAX_DIGITS
&& frac_part != 0)
{
frac_part *= 10;
uint32_t new_frac_part = frac_part & frac_part_mask;
uint32_t digit = (frac_part - new_frac_part) >> int_part_shift;
s = s * 10 + digit;
k++;
frac_part = new_frac_part;
}
JERRY_ASSERT (k > 0);
*out_digits_p = s;
*out_digits_num_p = k;
*out_decimal_exp_p = n;
#endif /* CONFIG_ECMA_NUMBER_TYPE == CONFIG_ECMA_NUMBER_FLOAT32 */
} /* ecma_number_to_decimal */
/**
* Convert ecma-number to zero-terminated string
*
* See also:
* ECMA-262 v5, 9.8.1
*
*
* @return size of utf-8 string
*/
lit_utf8_size_t
ecma_number_to_utf8_string (ecma_number_t num, /**< ecma-number */
lit_utf8_byte_t *buffer_p, /**< buffer for utf-8 string */
lit_utf8_size_t buffer_size) /**< size of buffer */
{
const lit_utf8_byte_t digits[10] = { '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' };
const lit_utf8_byte_t e_chars[2] = { 'e', 'E' };
const lit_utf8_byte_t plus_char = '+';
const lit_utf8_byte_t minus_char = '-';
const lit_utf8_byte_t dot_char = '.';
lit_utf8_size_t size;
if (ecma_number_is_nan (num))
{
// 1.
lit_copy_magic_string_to_buffer (LIT_MAGIC_STRING_NAN, buffer_p, buffer_size);
size = lit_get_magic_string_size (LIT_MAGIC_STRING_NAN);
}
else
{
lit_utf8_byte_t *dst_p = buffer_p;
if (ecma_number_is_zero (num))
{
// 2.
*dst_p++ = digits[0];
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
size = (lit_utf8_size_t) (dst_p - buffer_p);
}
else if (ecma_number_is_negative (num))
{
// 3.
*dst_p++ = minus_char;
lit_utf8_size_t new_buffer_size = (lit_utf8_size_t) ((buffer_p + buffer_size) - dst_p);
size = 1 + ecma_number_to_utf8_string (ecma_number_negate (num), dst_p, new_buffer_size);
}
else if (ecma_number_is_infinity (num))
{
// 4.
dst_p = lit_copy_magic_string_to_buffer (LIT_MAGIC_STRING_INFINITY_UL, buffer_p, buffer_size);
size = (lit_utf8_size_t) (dst_p - buffer_p);
}
else
{
ecma_number_t p = ecma_number_get_prev (num);
ecma_number_t q = ecma_number_get_next (p);
JERRY_ASSERT (q == num);
// 5.
uint32_t num_uint32 = ecma_number_to_uint32 (num);
if (((ecma_number_t) num_uint32) == num)
{
size = ecma_uint32_to_utf8_string (num_uint32, dst_p, buffer_size);
}
else
{
/* mantissa */
uint64_t s;
/* decimal exponent */
int32_t n;
/* number of digits in k */
int32_t k;
ecma_number_to_decimal (num, &s, &k, &n);
// 6.
if (k <= n && n <= 21)
{
dst_p += n;
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
size = (lit_utf8_size_t) (dst_p - buffer_p);
for (int32_t i = 0; i < n - k; i++)
{
*--dst_p = digits[0];
}
for (int32_t i = 0; i < k; i++)
{
*--dst_p = digits[s % 10];
s /= 10;
}
}
else if (0 < n && n <= 21)
{
// 7.
dst_p += k + 1;
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
size = (lit_utf8_size_t) (dst_p - buffer_p);
for (int32_t i = 0; i < k - n; i++)
{
*--dst_p = digits[s % 10];
s /= 10;
}
*--dst_p = dot_char;
for (int32_t i = 0; i < n; i++)
{
*--dst_p = digits[s % 10];
s /= 10;
}
}
else if (-6 < n && n <= 0)
{
// 8.
dst_p += k - n + 1 + 1;
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
size = (lit_utf8_size_t) (dst_p - buffer_p);
for (int32_t i = 0; i < k; i++)
{
*--dst_p = digits[s % 10];
s /= 10;
}
for (int32_t i = 0; i < -n; i++)
{
*--dst_p = digits[0];
}
*--dst_p = dot_char;
*--dst_p = digits[0];
}
else
{
if (k == 1)
{
// 9.
JERRY_ASSERT (1 <= buffer_size);
size = 1;
*dst_p++ = digits[s % 10];
s /= 10;
}
else
{
// 10.
dst_p += k + 1;
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
for (int32_t i = 0; i < k - 1; i++)
{
*--dst_p = digits[s % 10];
s /= 10;
}
*--dst_p = dot_char;
*--dst_p = digits[s % 10];
s /= 10;
dst_p += k + 1;
}
// 9., 10.
JERRY_ASSERT (dst_p + 2 <= buffer_p + buffer_size);
*dst_p++ = e_chars[0];
*dst_p++ = (n >= 1) ? plus_char : minus_char;
int32_t t = (n >= 1) ? (n - 1) : -(n - 1);
if (t == 0)
{
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
*dst_p++ = digits[0];
}
else
{
int32_t t_mod = 1000000000u;
while ((t / t_mod) == 0)
{
t_mod /= 10;
JERRY_ASSERT (t != 0);
}
while (t_mod != 0)
{
JERRY_ASSERT (dst_p + 1 <= buffer_p + buffer_size);
*dst_p++ = digits[t / t_mod];
t -= (t / t_mod) * t_mod;
t_mod /= 10;
}
}
JERRY_ASSERT (dst_p <= buffer_p + buffer_size);
size = (lit_utf8_size_t) (dst_p - buffer_p);
}
JERRY_ASSERT (s == 0);
}
}
}
return size;
} /* ecma_number_to_utf8_string */
/**
* @}
* @}
*/