aboutsummaryrefslogtreecommitdiffstats
path: root/gcc-4.9/libgcc/config/libbid/bid128_string.c
diff options
context:
space:
mode:
Diffstat (limited to 'gcc-4.9/libgcc/config/libbid/bid128_string.c')
-rw-r--r--gcc-4.9/libgcc/config/libbid/bid128_string.c672
1 files changed, 672 insertions, 0 deletions
diff --git a/gcc-4.9/libgcc/config/libbid/bid128_string.c b/gcc-4.9/libgcc/config/libbid/bid128_string.c
new file mode 100644
index 000000000..68f2a3d8b
--- /dev/null
+++ b/gcc-4.9/libgcc/config/libbid/bid128_string.c
@@ -0,0 +1,672 @@
+/* Copyright (C) 2007-2014 Free Software Foundation, Inc.
+
+This file is part of GCC.
+
+GCC is free software; you can redistribute it and/or modify it under
+the terms of the GNU General Public License as published by the Free
+Software Foundation; either version 3, or (at your option) any later
+version.
+
+GCC is distributed in the hope that it will be useful, but WITHOUT ANY
+WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+for more details.
+
+Under Section 7 of GPL version 3, you are granted additional
+permissions described in the GCC Runtime Library Exception, version
+3.1, as published by the Free Software Foundation.
+
+You should have received a copy of the GNU General Public License and
+a copy of the GCC Runtime Library Exception along with this program;
+see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
+<http://www.gnu.org/licenses/>. */
+
+/*****************************************************************************
+ * BID128_to_string
+ ****************************************************************************/
+
+#define BID_128RES
+#include <stdio.h>
+#include "bid_internal.h"
+#include "bid128_2_str.h"
+#include "bid128_2_str_macros.h"
+
+extern int bid128_coeff_2_string (UINT64 X_hi, UINT64 X_lo,
+ char *char_ptr);
+
+#if DECIMAL_CALL_BY_REFERENCE
+
+void
+bid128_to_string (char *str,
+ UINT128 *
+ px _EXC_FLAGS_PARAM _EXC_MASKS_PARAM
+ _EXC_INFO_PARAM) {
+ UINT128 x;
+#else
+
+void
+bid128_to_string (char *str, UINT128 x
+ _EXC_FLAGS_PARAM _EXC_MASKS_PARAM _EXC_INFO_PARAM) {
+#endif
+ UINT64 x_sign;
+ UINT64 x_exp;
+ int exp; // unbiased exponent
+ // Note: C1.w[1], C1.w[0] represent x_signif_hi, x_signif_lo (all are UINT64)
+ int ind;
+ UINT128 C1;
+ unsigned int k = 0; // pointer in the string
+ unsigned int d0, d123;
+ UINT64 HI_18Dig, LO_18Dig, Tmp;
+ UINT32 MiDi[12], *ptr;
+ char *c_ptr_start, *c_ptr;
+ int midi_ind, k_lcv, len;
+
+#if DECIMAL_CALL_BY_REFERENCE
+ x = *px;
+#endif
+
+ BID_SWAP128(x);
+ // check for NaN or Infinity
+ if ((x.w[1] & MASK_SPECIAL) == MASK_SPECIAL) {
+ // x is special
+ if ((x.w[1] & MASK_NAN) == MASK_NAN) { // x is NAN
+ if ((x.w[1] & MASK_SNAN) == MASK_SNAN) { // x is SNAN
+ // set invalid flag
+ str[0] = ((SINT64)x.w[1]<0)? '-':'+';
+ str[1] = 'S';
+ str[2] = 'N';
+ str[3] = 'a';
+ str[4] = 'N';
+ str[5] = '\0';
+ } else { // x is QNaN
+ str[0] = ((SINT64)x.w[1]<0)? '-':'+';
+ str[1] = 'Q';
+ str[2] = 'N';
+ str[3] = 'a';
+ str[4] = 'N';
+ str[5] = '\0';
+ }
+ } else { // x is not a NaN, so it must be infinity
+ if ((x.w[1] & MASK_SIGN) == 0x0ull) { // x is +inf
+ str[0] = '+';
+ str[1] = 'I';
+ str[2] = 'n';
+ str[3] = 'f';
+ str[4] = '\0';
+ } else { // x is -inf
+ str[0] = '-';
+ str[1] = 'I';
+ str[2] = 'n';
+ str[3] = 'f';
+ str[4] = '\0';
+ }
+ }
+ return;
+ } else if (((x.w[1] & MASK_COEFF) == 0x0ull) && (x.w[0] == 0x0ull)) {
+ // x is 0
+ len = 0;
+
+ //determine if +/-
+ if (x.w[1] & MASK_SIGN)
+ str[len++] = '-';
+ else
+ str[len++] = '+';
+ str[len++] = '0';
+ str[len++] = 'E';
+
+ // extract the exponent and print
+ exp = (int) (((x.w[1] & MASK_EXP) >> 49) - 6176);
+ if(exp > (((0x5ffe)>>1) - (6176))) {
+ exp = (int) ((((x.w[1]<<2) & MASK_EXP) >> 49) - 6176);
+ }
+ if (exp >= 0) {
+ str[len++] = '+';
+ len += sprintf (str + len, "%u", exp);// should not use sprintf (should
+ // use sophisticated algorithm, since we know range of exp is limited)
+ str[len++] = '\0';
+ } else {
+ len += sprintf (str + len, "%d", exp);// should not use sprintf (should
+ // use sophisticated algorithm, since we know range of exp is limited)
+ str[len++] = '\0';
+ }
+ return;
+ } else { // x is not special and is not zero
+ // unpack x
+ x_sign = x.w[1] & MASK_SIGN;// 0 for positive, MASK_SIGN for negative
+ x_exp = x.w[1] & MASK_EXP;// biased and shifted left 49 bit positions
+ if ((x.w[1] & 0x6000000000000000ull) == 0x6000000000000000ull)
+ x_exp = (x.w[1]<<2) & MASK_EXP;// biased and shifted left 49 bit positions
+ C1.w[1] = x.w[1] & MASK_COEFF;
+ C1.w[0] = x.w[0];
+ exp = (x_exp >> 49) - 6176;
+
+ // determine sign's representation as a char
+ if (x_sign)
+ str[k++] = '-';// negative number
+ else
+ str[k++] = '+';// positive number
+
+ // determine coefficient's representation as a decimal string
+
+ // if zero or non-canonical, set coefficient to '0'
+ if ((C1.w[1] > 0x0001ed09bead87c0ull) ||
+ (C1.w[1] == 0x0001ed09bead87c0ull &&
+ (C1.w[0] > 0x378d8e63ffffffffull)) ||
+ ((x.w[1] & 0x6000000000000000ull) == 0x6000000000000000ull) ||
+ ((C1.w[1] == 0) && (C1.w[0] == 0))) {
+ str[k++] = '0';
+ } else {
+ /* ****************************************************
+ This takes a bid coefficient in C1.w[1],C1.w[0]
+ and put the converted character sequence at location
+ starting at &(str[k]). The function returns the number
+ of MiDi returned. Note that the character sequence
+ does not have leading zeros EXCEPT when the input is of
+ zero value. It will then output 1 character '0'
+ The algorithm essentailly tries first to get a sequence of
+ Millenial Digits "MiDi" and then uses table lookup to get the
+ character strings of these MiDis.
+ **************************************************** */
+ /* Algorithm first decompose possibly 34 digits in hi and lo
+ 18 digits. (The high can have at most 16 digits). It then
+ uses macro that handle 18 digit portions.
+ The first step is to get hi and lo such that
+ 2^(64) C1.w[1] + C1.w[0] = hi * 10^18 + lo, 0 <= lo < 10^18.
+ We use a table lookup method to obtain the hi and lo 18 digits.
+ [C1.w[1],C1.w[0]] = c_8 2^(107) + c_7 2^(101) + ... + c_0 2^(59) + d
+ where 0 <= d < 2^59 and each c_j has 6 bits. Because d fits in
+ 18 digits, we set hi = 0, and lo = d to begin with.
+ We then retrieve from a table, for j = 0, 1, ..., 8
+ that gives us A and B where c_j 2^(59+6j) = A * 10^18 + B.
+ hi += A ; lo += B; After each accumulation into lo, we normalize
+ immediately. So at the end, we have the decomposition as we need. */
+
+ Tmp = C1.w[0] >> 59;
+ LO_18Dig = (C1.w[0] << 5) >> 5;
+ Tmp += (C1.w[1] << 5);
+ HI_18Dig = 0;
+ k_lcv = 0;
+ // Tmp = {C1.w[1]{49:0}, C1.w[0]{63:59}}
+ // Lo_18Dig = {C1.w[0]{58:0}}
+
+ while (Tmp) {
+ midi_ind = (int) (Tmp & 0x000000000000003FLL);
+ midi_ind <<= 1;
+ Tmp >>= 6;
+ HI_18Dig += mod10_18_tbl[k_lcv][midi_ind++];
+ LO_18Dig += mod10_18_tbl[k_lcv++][midi_ind];
+ __L0_Normalize_10to18 (HI_18Dig, LO_18Dig);
+ }
+ ptr = MiDi;
+ if (HI_18Dig == 0LL) {
+ __L1_Split_MiDi_6_Lead (LO_18Dig, ptr);
+ } else {
+ __L1_Split_MiDi_6_Lead (HI_18Dig, ptr);
+ __L1_Split_MiDi_6 (LO_18Dig, ptr);
+ }
+ len = ptr - MiDi;
+ c_ptr_start = &(str[k]);
+ c_ptr = c_ptr_start;
+
+ /* now convert the MiDi into character strings */
+ __L0_MiDi2Str_Lead (MiDi[0], c_ptr);
+ for (k_lcv = 1; k_lcv < len; k_lcv++) {
+ __L0_MiDi2Str (MiDi[k_lcv], c_ptr);
+ }
+ k = k + (c_ptr - c_ptr_start);
+ }
+
+ // print E and sign of exponent
+ str[k++] = 'E';
+ if (exp < 0) {
+ exp = -exp;
+ str[k++] = '-';
+ } else {
+ str[k++] = '+';
+ }
+
+ // determine exponent's representation as a decimal string
+ // d0 = exp / 1000;
+ // Use Property 1
+ d0 = (exp * 0x418a) >> 24;// 0x418a * 2^-24 = (10^(-3))RP,15
+ d123 = exp - 1000 * d0;
+
+ if (d0) { // 1000 <= exp <= 6144 => 4 digits to return
+ str[k++] = d0 + 0x30;// ASCII for decimal digit d0
+ ind = 3 * d123;
+ str[k++] = char_table3[ind];
+ str[k++] = char_table3[ind + 1];
+ str[k++] = char_table3[ind + 2];
+ } else { // 0 <= exp <= 999 => d0 = 0
+ if (d123 < 10) { // 0 <= exp <= 9 => 1 digit to return
+ str[k++] = d123 + 0x30;// ASCII
+ } else if (d123 < 100) { // 10 <= exp <= 99 => 2 digits to return
+ ind = 2 * (d123 - 10);
+ str[k++] = char_table2[ind];
+ str[k++] = char_table2[ind + 1];
+ } else { // 100 <= exp <= 999 => 3 digits to return
+ ind = 3 * d123;
+ str[k++] = char_table3[ind];
+ str[k++] = char_table3[ind + 1];
+ str[k++] = char_table3[ind + 2];
+ }
+ }
+ str[k] = '\0';
+
+ }
+ return;
+
+}
+
+
+#define MAX_FORMAT_DIGITS_128 34
+#define MAX_STRING_DIGITS_128 100
+#define MAX_SEARCH MAX_STRING_DIGITS_128-MAX_FORMAT_DIGITS_128-1
+
+
+#if DECIMAL_CALL_BY_REFERENCE
+
+void
+bid128_from_string (UINT128 * pres,
+ char *ps _RND_MODE_PARAM _EXC_FLAGS_PARAM
+ _EXC_MASKS_PARAM _EXC_INFO_PARAM) {
+#else
+
+UINT128
+bid128_from_string (char *ps _RND_MODE_PARAM _EXC_FLAGS_PARAM
+ _EXC_MASKS_PARAM _EXC_INFO_PARAM) {
+#endif
+ UINT128 CX, res;
+ UINT64 sign_x, coeff_high, coeff_low, coeff2, coeff_l2, carry = 0x0ull,
+ scale_high, right_radix_leading_zeros;
+ int ndigits_before, ndigits_after, ndigits_total, dec_expon, sgn_exp,
+ i, d2, rdx_pt_enc;
+ char c, buffer[MAX_STRING_DIGITS_128];
+ int save_rnd_mode;
+ int save_fpsf;
+
+#if DECIMAL_CALL_BY_REFERENCE
+#if !DECIMAL_GLOBAL_ROUNDING
+ _IDEC_round rnd_mode = *prnd_mode;
+#endif
+#endif
+
+ save_rnd_mode = rnd_mode; // dummy
+ save_fpsf = *pfpsf; // dummy
+
+ right_radix_leading_zeros = rdx_pt_enc = 0;
+
+ // if null string, return NaN
+ if (!ps) {
+ res.w[1] = 0x7c00000000000000ull;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+ // eliminate leading white space
+ while ((*ps == ' ') || (*ps == '\t'))
+ ps++;
+
+ // c gets first character
+ c = *ps;
+
+
+ // if c is null or not equal to a (radix point, negative sign,
+ // positive sign, or number) it might be SNaN, sNaN, Infinity
+ if (!c
+ || (c != '.' && c != '-' && c != '+'
+ && ((unsigned) (c - '0') > 9))) {
+ res.w[0] = 0;
+ // Infinity?
+ if ((tolower_macro (ps[0]) == 'i' && tolower_macro (ps[1]) == 'n'
+ && tolower_macro (ps[2]) == 'f')
+ && (!ps[3]
+ || (tolower_macro (ps[3]) == 'i'
+ && tolower_macro (ps[4]) == 'n'
+ && tolower_macro (ps[5]) == 'i'
+ && tolower_macro (ps[6]) == 't'
+ && tolower_macro (ps[7]) == 'y' && !ps[8])
+ )) {
+ res.w[1] = 0x7800000000000000ull;
+ BID_RETURN (res);
+ }
+ // return sNaN
+ if (tolower_macro (ps[0]) == 's' && tolower_macro (ps[1]) == 'n' &&
+ tolower_macro (ps[2]) == 'a' && tolower_macro (ps[3]) == 'n') {
+ // case insensitive check for snan
+ res.w[1] = 0x7e00000000000000ull;
+ BID_RETURN (res);
+ } else {
+ // return qNaN
+ res.w[1] = 0x7c00000000000000ull;
+ BID_RETURN (res);
+ }
+ }
+ // if +Inf, -Inf, +Infinity, or -Infinity (case insensitive check for inf)
+ if ((tolower_macro (ps[1]) == 'i' && tolower_macro (ps[2]) == 'n' &&
+ tolower_macro (ps[3]) == 'f') && (!ps[4] ||
+ (tolower_macro (ps[4]) == 'i' && tolower_macro (ps[5]) == 'n' &&
+ tolower_macro (ps[6]) == 'i' && tolower_macro (ps[7]) == 't' &&
+ tolower_macro (ps[8]) == 'y' && !ps[9]))) { // ci check for infinity
+ res.w[0] = 0;
+
+ if (c == '+')
+ res.w[1] = 0x7800000000000000ull;
+ else if (c == '-')
+ res.w[1] = 0xf800000000000000ull;
+ else
+ res.w[1] = 0x7c00000000000000ull;
+
+ BID_RETURN (res);
+ }
+ // if +sNaN, +SNaN, -sNaN, or -SNaN
+ if (tolower_macro (ps[1]) == 's' && tolower_macro (ps[2]) == 'n'
+ && tolower_macro (ps[3]) == 'a' && tolower_macro (ps[4]) == 'n') {
+ res.w[0] = 0;
+ if (c == '-')
+ res.w[1] = 0xfe00000000000000ull;
+ else
+ res.w[1] = 0x7e00000000000000ull;
+ BID_RETURN (res);
+ }
+ // set up sign_x to be OR'ed with the upper word later
+ if (c == '-')
+ sign_x = 0x8000000000000000ull;
+ else
+ sign_x = 0;
+
+ // go to next character if leading sign
+ if (c == '-' || c == '+')
+ ps++;
+
+ c = *ps;
+
+ // if c isn't a decimal point or a decimal digit, return NaN
+ if (c != '.' && ((unsigned) (c - '0') > 9)) {
+ res.w[1] = 0x7c00000000000000ull | sign_x;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+ // detect zero (and eliminate/ignore leading zeros)
+ if (*(ps) == '0') {
+
+ // if all numbers are zeros (with possibly 1 radix point, the number is zero
+ // should catch cases such as: 000.0
+ while (*ps == '0') {
+
+ ps++;
+
+ // for numbers such as 0.0000000000000000000000000000000000001001,
+ // we want to count the leading zeros
+ if (rdx_pt_enc) {
+ right_radix_leading_zeros++;
+ }
+ // if this character is a radix point, make sure we haven't already
+ // encountered one
+ if (*(ps) == '.') {
+ if (rdx_pt_enc == 0) {
+ rdx_pt_enc = 1;
+ // if this is the first radix point, and the next character is NULL,
+ // we have a zero
+ if (!*(ps + 1)) {
+ res.w[1] =
+ (0x3040000000000000ull -
+ (right_radix_leading_zeros << 49)) | sign_x;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+ ps = ps + 1;
+ } else {
+ // if 2 radix points, return NaN
+ res.w[1] = 0x7c00000000000000ull | sign_x;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+ } else if (!*(ps)) {
+ //res.w[1] = 0x3040000000000000ull | sign_x;
+ res.w[1] =
+ (0x3040000000000000ull -
+ (right_radix_leading_zeros << 49)) | sign_x;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+ }
+ }
+
+ c = *ps;
+
+ // initialize local variables
+ ndigits_before = ndigits_after = ndigits_total = 0;
+ sgn_exp = 0;
+ // pstart_coefficient = ps;
+
+ if (!rdx_pt_enc) {
+ // investigate string (before radix point)
+ while ((unsigned) (c - '0') <= 9
+ && ndigits_before < MAX_STRING_DIGITS_128) {
+ buffer[ndigits_before] = c;
+ ps++;
+ c = *ps;
+ ndigits_before++;
+ }
+
+ ndigits_total = ndigits_before;
+ if (c == '.') {
+ ps++;
+ if ((c = *ps)) {
+
+ // investigate string (after radix point)
+ while ((unsigned) (c - '0') <= 9
+ && ndigits_total < MAX_STRING_DIGITS_128) {
+ buffer[ndigits_total] = c;
+ ps++;
+ c = *ps;
+ ndigits_total++;
+ }
+ ndigits_after = ndigits_total - ndigits_before;
+ }
+ }
+ } else {
+ // we encountered a radix point while detecting zeros
+ //if (c = *ps){
+
+ c = *ps;
+ ndigits_total = 0;
+ // investigate string (after radix point)
+ while ((unsigned) (c - '0') <= 9
+ && ndigits_total < MAX_STRING_DIGITS_128) {
+ buffer[ndigits_total] = c;
+ ps++;
+ c = *ps;
+ ndigits_total++;
+ }
+ ndigits_after = ndigits_total - ndigits_before;
+ }
+
+ // get exponent
+ dec_expon = 0;
+ if (ndigits_total < MAX_STRING_DIGITS_128) {
+ if (c) {
+ if (c != 'e' && c != 'E') {
+ // return NaN
+ res.w[1] = 0x7c00000000000000ull;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+ ps++;
+ c = *ps;
+
+ if (((unsigned) (c - '0') > 9)
+ && ((c != '+' && c != '-') || (unsigned) (ps[1] - '0') > 9)) {
+ // return NaN
+ res.w[1] = 0x7c00000000000000ull;
+ res.w[0] = 0;
+ BID_RETURN (res);
+ }
+
+ if (c == '-') {
+ sgn_exp = -1;
+ ps++;
+ c = *ps;
+ } else if (c == '+') {
+ ps++;
+ c = *ps;
+ }
+
+ dec_expon = c - '0';
+ i = 1;
+ ps++;
+ c = *ps - '0';
+ while (((unsigned) c) <= 9 && i < 7) {
+ d2 = dec_expon + dec_expon;
+ dec_expon = (d2 << 2) + d2 + c;
+ ps++;
+ c = *ps - '0';
+ i++;
+ }
+ }
+
+ dec_expon = (dec_expon + sgn_exp) ^ sgn_exp;
+ }
+
+
+ if (ndigits_total <= MAX_FORMAT_DIGITS_128) {
+ dec_expon +=
+ DECIMAL_EXPONENT_BIAS_128 - ndigits_after -
+ right_radix_leading_zeros;
+ if (dec_expon < 0) {
+ res.w[1] = 0 | sign_x;
+ res.w[0] = 0;
+ }
+ if (ndigits_total == 0) {
+ CX.w[0] = 0;
+ CX.w[1] = 0;
+ } else if (ndigits_total <= 19) {
+ coeff_high = buffer[0] - '0';
+ for (i = 1; i < ndigits_total; i++) {
+ coeff2 = coeff_high + coeff_high;
+ coeff_high = (coeff2 << 2) + coeff2 + buffer[i] - '0';
+ }
+ CX.w[0] = coeff_high;
+ CX.w[1] = 0;
+ } else {
+ coeff_high = buffer[0] - '0';
+ for (i = 1; i < ndigits_total - 17; i++) {
+ coeff2 = coeff_high + coeff_high;
+ coeff_high = (coeff2 << 2) + coeff2 + buffer[i] - '0';
+ }
+ coeff_low = buffer[i] - '0';
+ i++;
+ for (; i < ndigits_total; i++) {
+ coeff_l2 = coeff_low + coeff_low;
+ coeff_low = (coeff_l2 << 2) + coeff_l2 + buffer[i] - '0';
+ }
+ // now form the coefficient as coeff_high*10^19+coeff_low+carry
+ scale_high = 100000000000000000ull;
+ __mul_64x64_to_128_fast (CX, coeff_high, scale_high);
+
+ CX.w[0] += coeff_low;
+ if (CX.w[0] < coeff_low)
+ CX.w[1]++;
+ }
+ get_BID128 (&res, sign_x, dec_expon, CX,&rnd_mode,pfpsf);
+ BID_RETURN (res);
+ } else {
+ // simply round using the digits that were read
+
+ dec_expon +=
+ ndigits_before + DECIMAL_EXPONENT_BIAS_128 -
+ MAX_FORMAT_DIGITS_128 - right_radix_leading_zeros;
+
+ if (dec_expon < 0) {
+ res.w[1] = 0 | sign_x;
+ res.w[0] = 0;
+ }
+
+ coeff_high = buffer[0] - '0';
+ for (i = 1; i < MAX_FORMAT_DIGITS_128 - 17; i++) {
+ coeff2 = coeff_high + coeff_high;
+ coeff_high = (coeff2 << 2) + coeff2 + buffer[i] - '0';
+ }
+ coeff_low = buffer[i] - '0';
+ i++;
+ for (; i < MAX_FORMAT_DIGITS_128; i++) {
+ coeff_l2 = coeff_low + coeff_low;
+ coeff_low = (coeff_l2 << 2) + coeff_l2 + buffer[i] - '0';
+ }
+ switch(rnd_mode) {
+ case ROUNDING_TO_NEAREST:
+ carry = ((unsigned) ('4' - buffer[i])) >> 31;
+ if ((buffer[i] == '5' && !(coeff_low & 1)) || dec_expon < 0) {
+ if (dec_expon >= 0) {
+ carry = 0;
+ i++;
+ }
+ for (; i < ndigits_total; i++) {
+ if (buffer[i] > '0') {
+ carry = 1;
+ break;
+ }
+ }
+ }
+ break;
+
+ case ROUNDING_DOWN:
+ if(sign_x)
+ for (; i < ndigits_total; i++) {
+ if (buffer[i] > '0') {
+ carry = 1;
+ break;
+ }
+ }
+ break;
+ case ROUNDING_UP:
+ if(!sign_x)
+ for (; i < ndigits_total; i++) {
+ if (buffer[i] > '0') {
+ carry = 1;
+ break;
+ }
+ }
+ break;
+ case ROUNDING_TO_ZERO:
+ carry=0;
+ break;
+ case ROUNDING_TIES_AWAY:
+ carry = ((unsigned) ('4' - buffer[i])) >> 31;
+ if (dec_expon < 0) {
+ for (; i < ndigits_total; i++) {
+ if (buffer[i] > '0') {
+ carry = 1;
+ break;
+ }
+ }
+ }
+ break;
+
+
+ }
+ // now form the coefficient as coeff_high*10^17+coeff_low+carry
+ scale_high = 100000000000000000ull;
+ if (dec_expon < 0) {
+ if (dec_expon > -MAX_FORMAT_DIGITS_128) {
+ scale_high = 1000000000000000000ull;
+ coeff_low = (coeff_low << 3) + (coeff_low << 1);
+ dec_expon--;
+ }
+ if (dec_expon == -MAX_FORMAT_DIGITS_128
+ && coeff_high > 50000000000000000ull)
+ carry = 0;
+ }
+
+ __mul_64x64_to_128_fast (CX, coeff_high, scale_high);
+
+ coeff_low += carry;
+ CX.w[0] += coeff_low;
+ if (CX.w[0] < coeff_low)
+ CX.w[1]++;
+
+
+ get_BID128(&res, sign_x, dec_expon, CX, &rnd_mode, pfpsf);
+ BID_RETURN (res);
+ }
+}