Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /* Copyright (C) 2007-2015 Free Software Foundation, Inc.
  2.  
  3. This file is part of GCC.
  4.  
  5. GCC is free software; you can redistribute it and/or modify it under
  6. the terms of the GNU General Public License as published by the Free
  7. Software Foundation; either version 3, or (at your option) any later
  8. version.
  9.  
  10. GCC is distributed in the hope that it will be useful, but WITHOUT ANY
  11. WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12. FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  13. for more details.
  14.  
  15. Under Section 7 of GPL version 3, you are granted additional
  16. permissions described in the GCC Runtime Library Exception, version
  17. 3.1, as published by the Free Software Foundation.
  18.  
  19. You should have received a copy of the GNU General Public License and
  20. a copy of the GCC Runtime Library Exception along with this program;
  21. see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
  22. <http://www.gnu.org/licenses/>.  */
  23.  
  24. #ifndef _DIV_MACROS_H_
  25. #define _DIV_MACROS_H_
  26.  
  27. #include "bid_internal.h"
  28.  
  29. #define FENCE __fence
  30. //#define FENCE
  31.  
  32. //#define DOUBLE_EXTENDED_ON
  33.  
  34. #if DOUBLE_EXTENDED_ON
  35.  
  36.  
  37. __BID_INLINE__ void
  38. __div_128_by_128 (UINT128 * pCQ, UINT128 * pCR, UINT128 CX, UINT128 CY) {
  39.   UINT128 CB, CB2, CB4, CB8, CQB, CA;
  40.   int_double d64, dm64, ds;
  41.   int_float t64;
  42.   double dx, dq, dqh;
  43.   BINARY80 lq, lx, ly;
  44.   UINT64 Rh, R, B2, B4, Ph, Ql, Ql2, carry, Qh;
  45.  
  46.   if (!CY.w[1]) {
  47.     pCR->w[1] = 0;
  48.  
  49.     if (!CX.w[1]) {
  50.       pCQ->w[0] = CX.w[0] / CY.w[0];
  51.       pCQ->w[1] = 0;
  52.       pCR->w[1] = 0;
  53.       pCR->w[0] = CX.w[0] - pCQ->w[0] * CY.w[0];
  54.     } else {
  55.  
  56.       // This path works for CX<2^116 only
  57.  
  58.       // 2^64
  59.       d64.i = 0x43f0000000000000;
  60.       // 2^64
  61.       dm64.i = 0x3bf0000000000000;
  62.       // 1.5*2^(-52)
  63.       ds.i = 0x3cb8000000000000;
  64.       dx = (BINARY80) CX.w[1] * d64.d + (BINARY80) CX.w[0];
  65.       dq = dx / (BINARY80) CY.w[0];
  66.       dq -= dq * (ds.d);
  67.       dqh = dq * dm64.d;
  68.       Qh = (UINT64) dqh;
  69.       Ql = (UINT64) (dq - ((double) Qh) * d64.d);
  70.  
  71.       Rh = CX.w[0] - Ql * CY.w[0];
  72.       Ql2 = Rh / CY.w[0];
  73.       pCR->w[0] = Rh - Ql2 * CY.w[0];
  74.       __add_carry_out ((pCQ->w[0]), carry, Ql, Ql2);
  75.       pCQ->w[1] = Qh + carry;
  76.  
  77.     }
  78.     return;
  79.   }
  80.   // now CY.w[1] > 0
  81.  
  82.   // 2^64
  83.   t64.i = 0x5f800000;
  84.   lx = (BINARY80) CX.w[1] * (BINARY80) t64.d + (BINARY80) CX.w[0];
  85.   ly = (BINARY80) CY.w[1] * (BINARY80) t64.d + (BINARY80) CY.w[0];
  86.   lq = lx / ly;
  87.   pCQ->w[0] = (UINT64) lq;
  88.  
  89.   pCQ->w[1] = 0;
  90.  
  91.   if (!pCQ->w[0]) {
  92.     /*if(__unsigned_compare_ge_128(CX,CY))
  93.        {
  94.        pCQ->w[0] = 1;
  95.        __sub_128_128((*pCR), CX, CY);
  96.        }
  97.        else */
  98.     {
  99.       pCR->w[1] = CX.w[1];
  100.       pCR->w[0] = CX.w[0];
  101.     }
  102.     return;
  103.   }
  104.  
  105.   if (CY.w[1] >= 16 || pCQ->w[0] <= 0x1000000000000000ull) {
  106.     pCQ->w[0] = (UINT64) lq - 1;
  107.     __mul_64x128_full (Ph, CQB, (pCQ->w[0]), CY);
  108.     __sub_128_128 (CA, CX, CQB);
  109.     if (__unsigned_compare_ge_128 (CA, CY)) {
  110.       __sub_128_128 (CA, CA, CY);
  111.       pCQ->w[0]++;
  112.       if (__unsigned_compare_ge_128 (CA, CY)) {
  113.         __sub_128_128 (CA, CA, CY);
  114.         pCQ->w[0]++;
  115.       }
  116.     }
  117.     pCR->w[1] = CA.w[1];
  118.     pCR->w[0] = CA.w[0];
  119.   } else {
  120.     pCQ->w[0] = (UINT64) lq - 6;
  121.  
  122.     __mul_64x128_full (Ph, CQB, (pCQ->w[0]), CY);
  123.     __sub_128_128 (CA, CX, CQB);
  124.  
  125.     CB8.w[1] = (CY.w[1] << 3) | (CY.w[0] >> 61);
  126.     CB8.w[0] = CY.w[0] << 3;
  127.     CB4.w[1] = (CY.w[1] << 2) | (CY.w[0] >> 62);
  128.     CB4.w[0] = CY.w[0] << 2;
  129.     CB2.w[1] = (CY.w[1] << 1) | (CY.w[0] >> 63);
  130.     CB2.w[0] = CY.w[0] << 1;
  131.  
  132.     if (__unsigned_compare_ge_128 (CA, CB8)) {
  133.       pCQ->w[0] += 8;
  134.       __sub_128_128 (CA, CA, CB8);
  135.     }
  136.     if (__unsigned_compare_ge_128 (CA, CB4)) {
  137.       pCQ->w[0] += 4;
  138.       __sub_128_128 (CA, CA, CB4);
  139.     }
  140.     if (__unsigned_compare_ge_128 (CA, CB2)) {
  141.       pCQ->w[0] += 2;
  142.       __sub_128_128 (CA, CA, CB2);
  143.     }
  144.     if (__unsigned_compare_ge_128 (CA, CY)) {
  145.       pCQ->w[0] += 1;
  146.       __sub_128_128 (CA, CA, CY);
  147.     }
  148.  
  149.     pCR->w[1] = CA.w[1];
  150.     pCR->w[0] = CA.w[0];
  151.   }
  152. }
  153.  
  154.  
  155.  
  156.  
  157.  
  158.  
  159. __BID_INLINE__ void
  160. __div_256_by_128 (UINT128 * pCQ, UINT256 * pCA4, UINT128 CY) {
  161.   UINT256 CQ2Y;
  162.   UINT128 CQ2, CQ3Y;
  163.   UINT64 Q3, carry64;
  164.   int_double d64;
  165.   BINARY80 lx, ly, lq, l64, l128;
  166.  
  167.   // 2^64
  168.   d64.i = 0x43f0000000000000ull;
  169.   l64 = (BINARY80) d64.d;
  170.   // 2^128
  171.   l128 = l64 * l64;
  172.  
  173.   lx =
  174.     ((BINARY80) (*pCA4).w[3] * l64 +
  175.      (BINARY80) (*pCA4).w[2]) * l128 +
  176.     (BINARY80) (*pCA4).w[1] * l64 + (BINARY80) (*pCA4).w[0];
  177.   ly = (BINARY80) CY.w[1] * l128 + (BINARY80) CY.w[0] * l64;
  178.  
  179.   lq = lx / ly;
  180.   CQ2.w[1] = (UINT64) lq;
  181.   lq = (lq - CQ2.w[1]) * l64;
  182.   CQ2.w[0] = (UINT64) lq;
  183.  
  184.   // CQ2*CY
  185.   __mul_128x128_to_256 (CQ2Y, CY, CQ2);
  186.  
  187.   // CQ2Y <= (*pCA4) ?
  188.   if (CQ2Y.w[3] < (*pCA4).w[3]
  189.       || (CQ2Y.w[3] == (*pCA4).w[3]
  190.           && (CQ2Y.w[2] < (*pCA4).w[2]
  191.               || (CQ2Y.w[2] == (*pCA4).w[2]
  192.                   && (CQ2Y.w[1] < (*pCA4).w[1]
  193.                       || (CQ2Y.w[1] == (*pCA4).w[1]
  194.                           && (CQ2Y.w[0] <= (*pCA4).w[0]))))))) {
  195.  
  196.     // (*pCA4) -CQ2Y, guaranteed below 5*2^49*CY < 5*2^(49+128)
  197.     __sub_borrow_out ((*pCA4).w[0], carry64, (*pCA4).w[0], CQ2Y.w[0]);
  198.     __sub_borrow_in_out ((*pCA4).w[1], carry64, (*pCA4).w[1], CQ2Y.w[1],
  199.                          carry64);
  200.     (*pCA4).w[2] = (*pCA4).w[2] - CQ2Y.w[2] - carry64;
  201.  
  202.     lx = ((BINARY80) (*pCA4).w[2] * l128 +
  203.           ((BINARY80) (*pCA4).w[1] * l64 +
  204.            (BINARY80) (*pCA4).w[0])) * l64;
  205.     lq = lx / ly;
  206.     Q3 = (UINT64) lq;
  207.  
  208.     if (Q3) {
  209.       Q3--;
  210.       __mul_64x128_short (CQ3Y, Q3, CY);
  211.       __sub_borrow_out ((*pCA4).w[0], carry64, (*pCA4).w[0], CQ3Y.w[0]);
  212.       (*pCA4).w[1] = (*pCA4).w[1] - CQ3Y.w[1] - carry64;
  213.  
  214.       if ((*pCA4).w[1] > CY.w[1]
  215.           || ((*pCA4).w[1] == CY.w[1] && (*pCA4).w[0] >= CY.w[0])) {
  216.         Q3++;
  217.         __sub_borrow_out ((*pCA4).w[0], carry64, (*pCA4).w[0], CY.w[0]);
  218.         (*pCA4).w[1] = (*pCA4).w[1] - CY.w[1] - carry64;
  219.         if ((*pCA4).w[1] > CY.w[1]
  220.             || ((*pCA4).w[1] == CY.w[1] && (*pCA4).w[0] >= CY.w[0])) {
  221.           Q3++;
  222.           __sub_borrow_out ((*pCA4).w[0], carry64, (*pCA4).w[0],
  223.                             CY.w[0]);
  224.           (*pCA4).w[1] = (*pCA4).w[1] - CY.w[1] - carry64;
  225.         }
  226.       }
  227.       // add Q3 to Q2
  228.       __add_carry_out (CQ2.w[0], carry64, Q3, CQ2.w[0]);
  229.       CQ2.w[1] += carry64;
  230.     }
  231.   } else {
  232.     // CQ2Y - (*pCA4), guaranteed below 5*2^(49+128)
  233.     __sub_borrow_out ((*pCA4).w[0], carry64, CQ2Y.w[0], (*pCA4).w[0]);
  234.     __sub_borrow_in_out ((*pCA4).w[1], carry64, CQ2Y.w[1], (*pCA4).w[1],
  235.                          carry64);
  236.     (*pCA4).w[2] = CQ2Y.w[2] - (*pCA4).w[2] - carry64;
  237.  
  238.     lx =
  239.       ((BINARY80) (*pCA4).w[2] * l128 +
  240.        (BINARY80) (*pCA4).w[1] * l64 + (BINARY80) (*pCA4).w[0]) * l64;
  241.     lq = lx / ly;
  242.     Q3 = 1 + (UINT64) lq;
  243.  
  244.     __mul_64x128_short (CQ3Y, Q3, CY);
  245.     __sub_borrow_out ((*pCA4).w[0], carry64, CQ3Y.w[0], (*pCA4).w[0]);
  246.     (*pCA4).w[1] = CQ3Y.w[1] - (*pCA4).w[1] - carry64;
  247.  
  248.     if ((SINT64) (*pCA4).w[1] > (SINT64) CY.w[1]
  249.         || ((*pCA4).w[1] == CY.w[1] && (*pCA4).w[0] >= CY.w[0])) {
  250.       Q3--;
  251.       __sub_borrow_out ((*pCA4).w[0], carry64, (*pCA4).w[0], CY.w[0]);
  252.       (*pCA4).w[1] = (*pCA4).w[1] - CY.w[1] - carry64;
  253.     } else if ((SINT64) (*pCA4).w[1] < 0) {
  254.       Q3++;
  255.       __add_carry_out ((*pCA4).w[0], carry64, (*pCA4).w[0], CY.w[0]);
  256.       (*pCA4).w[1] = (*pCA4).w[1] + CY.w[1] + carry64;
  257.     }
  258.     // subtract Q3 from Q2
  259.     __sub_borrow_out (CQ2.w[0], carry64, CQ2.w[0], Q3);
  260.     CQ2.w[1] -= carry64;
  261.   }
  262.  
  263.   // (*pCQ) + CQ2 + carry
  264.   __add_carry_out ((*pCQ).w[0], carry64, CQ2.w[0], (*pCQ).w[0]);
  265.   (*pCQ).w[1] = (*pCQ).w[1] + CQ2.w[1] + carry64;
  266.  
  267.  
  268. }
  269. #else
  270.  
  271. __BID_INLINE__ void
  272. __div_128_by_128 (UINT128 * pCQ, UINT128 * pCR, UINT128 CX0, UINT128 CY) {
  273.   UINT128 CY36, CY51, CQ, A2, CX, CQT;
  274.   UINT64 Q;
  275.   int_double t64, d49, d60;
  276.   double lx, ly, lq;
  277.  
  278.   if (!CX0.w[1] && !CY.w[1]) {
  279.     pCQ->w[0] = CX0.w[0] / CY.w[0];
  280.     pCQ->w[1] = 0;
  281.     pCR->w[1] = pCR->w[0] = 0;
  282.     pCR->w[0] = CX0.w[0] - pCQ->w[0] * CY.w[0];
  283.     return;
  284.   }
  285.  
  286.   CX.w[1] = CX0.w[1];
  287.   CX.w[0] = CX0.w[0];
  288.  
  289.   // 2^64
  290.   t64.i = 0x43f0000000000000ull;
  291.   lx = (double) CX.w[1] * t64.d + (double) CX.w[0];
  292.   ly = (double) CY.w[1] * t64.d + (double) CY.w[0];
  293.   lq = lx / ly;
  294.  
  295.   CY36.w[1] = CY.w[0] >> (64 - 36);
  296.   CY36.w[0] = CY.w[0] << 36;
  297.  
  298.   CQ.w[1] = CQ.w[0] = 0;
  299.  
  300.   // Q >= 2^100 ?
  301.   if (!CY.w[1] && !CY36.w[1] && (CX.w[1] >= CY36.w[0])) {
  302.     // then Q >= 2^100
  303.  
  304.     // 2^(-60)*CX/CY
  305.     d60.i = 0x3c30000000000000ull;
  306.     lq *= d60.d;
  307.     Q = (UINT64) lq - 4ull;
  308.  
  309.     // Q*CY
  310.     __mul_64x64_to_128 (A2, Q, CY.w[0]);
  311.  
  312.     // A2 <<= 60
  313.     A2.w[1] = (A2.w[1] << 60) | (A2.w[0] >> (64 - 60));
  314.     A2.w[0] <<= 60;
  315.  
  316.     __sub_128_128 (CX, CX, A2);
  317.  
  318.     lx = (double) CX.w[1] * t64.d + (double) CX.w[0];
  319.     lq = lx / ly;
  320.  
  321.     CQ.w[1] = Q >> (64 - 60);
  322.     CQ.w[0] = Q << 60;
  323.   }
  324.  
  325.  
  326.   CY51.w[1] = (CY.w[1] << 51) | (CY.w[0] >> (64 - 51));
  327.   CY51.w[0] = CY.w[0] << 51;
  328.  
  329.   if (CY.w[1] < (UINT64) (1 << (64 - 51))
  330.       && (__unsigned_compare_gt_128 (CX, CY51))) {
  331.     // Q > 2^51
  332.  
  333.     // 2^(-49)*CX/CY
  334.     d49.i = 0x3ce0000000000000ull;
  335.     lq *= d49.d;
  336.  
  337.     Q = (UINT64) lq - 1ull;
  338.  
  339.     // Q*CY
  340.     __mul_64x64_to_128 (A2, Q, CY.w[0]);
  341.     A2.w[1] += Q * CY.w[1];
  342.  
  343.     // A2 <<= 49
  344.     A2.w[1] = (A2.w[1] << 49) | (A2.w[0] >> (64 - 49));
  345.     A2.w[0] <<= 49;
  346.  
  347.     __sub_128_128 (CX, CX, A2);
  348.  
  349.     CQT.w[1] = Q >> (64 - 49);
  350.     CQT.w[0] = Q << 49;
  351.     __add_128_128 (CQ, CQ, CQT);
  352.  
  353.     lx = (double) CX.w[1] * t64.d + (double) CX.w[0];
  354.     lq = lx / ly;
  355.   }
  356.  
  357.   Q = (UINT64) lq;
  358.  
  359.   __mul_64x64_to_128 (A2, Q, CY.w[0]);
  360.   A2.w[1] += Q * CY.w[1];
  361.  
  362.   __sub_128_128 (CX, CX, A2);
  363.   if ((SINT64) CX.w[1] < 0) {
  364.     Q--;
  365.     CX.w[0] += CY.w[0];
  366.     if (CX.w[0] < CY.w[0])
  367.       CX.w[1]++;
  368.     CX.w[1] += CY.w[1];
  369.     if ((SINT64) CX.w[1] < 0) {
  370.       Q--;
  371.       CX.w[0] += CY.w[0];
  372.       if (CX.w[0] < CY.w[0])
  373.         CX.w[1]++;
  374.       CX.w[1] += CY.w[1];
  375.     }
  376.   } else if (__unsigned_compare_ge_128 (CX, CY)) {
  377.     Q++;
  378.     __sub_128_128 (CX, CX, CY);
  379.   }
  380.  
  381.   __add_128_64 (CQ, CQ, Q);
  382.  
  383.  
  384.   pCQ->w[1] = CQ.w[1];
  385.   pCQ->w[0] = CQ.w[0];
  386.   pCR->w[1] = CX.w[1];
  387.   pCR->w[0] = CX.w[0];
  388.   return;
  389. }
  390.  
  391.  
  392. __BID_INLINE__ void
  393. __div_256_by_128 (UINT128 * pCQ, UINT256 * pCA4, UINT128 CY) {
  394.   UINT256 CA4, CA2, CY51, CY36;
  395.   UINT128 CQ, A2, A2h, CQT;
  396.   UINT64 Q, carry64;
  397.   int_double t64, d49, d60;
  398.   double lx, ly, lq, d128, d192;
  399.  
  400.   // the quotient is assumed to be at most 113 bits,
  401.   // as needed by BID128 divide routines
  402.  
  403.   // initial dividend
  404.   CA4.w[3] = (*pCA4).w[3];
  405.   CA4.w[2] = (*pCA4).w[2];
  406.   CA4.w[1] = (*pCA4).w[1];
  407.   CA4.w[0] = (*pCA4).w[0];
  408.   CQ.w[1] = (*pCQ).w[1];
  409.   CQ.w[0] = (*pCQ).w[0];
  410.  
  411.   // 2^64
  412.   t64.i = 0x43f0000000000000ull;
  413.   d128 = t64.d * t64.d;
  414.   d192 = d128 * t64.d;
  415.   lx = (double) CA4.w[3] * d192 + ((double) CA4.w[2] * d128 +
  416.                                    ((double) CA4.w[1] * t64.d +
  417.                                     (double) CA4.w[0]));
  418.   ly = (double) CY.w[1] * t64.d + (double) CY.w[0];
  419.   lq = lx / ly;
  420.  
  421.   CY36.w[2] = CY.w[1] >> (64 - 36);
  422.   CY36.w[1] = (CY.w[1] << 36) | (CY.w[0] >> (64 - 36));
  423.   CY36.w[0] = CY.w[0] << 36;
  424.  
  425.   CQ.w[1] = (*pCQ).w[1];
  426.   CQ.w[0] = (*pCQ).w[0];
  427.  
  428.   // Q >= 2^100 ?
  429.   if (CA4.w[3] > CY36.w[2]
  430.       || (CA4.w[3] == CY36.w[2]
  431.           && (CA4.w[2] > CY36.w[1]
  432.               || (CA4.w[2] == CY36.w[1] && CA4.w[1] >= CY36.w[0])))) {
  433.     // 2^(-60)*CA4/CY
  434.     d60.i = 0x3c30000000000000ull;
  435.     lq *= d60.d;
  436.     Q = (UINT64) lq - 4ull;
  437.  
  438.     // Q*CY
  439.     __mul_64x128_to_192 (CA2, Q, CY);
  440.  
  441.     // CA2 <<= 60
  442.     // CA2.w[3] = CA2.w[2] >> (64-60);
  443.     CA2.w[2] = (CA2.w[2] << 60) | (CA2.w[1] >> (64 - 60));
  444.     CA2.w[1] = (CA2.w[1] << 60) | (CA2.w[0] >> (64 - 60));
  445.     CA2.w[0] <<= 60;
  446.  
  447.     // CA4 -= CA2
  448.     __sub_borrow_out (CA4.w[0], carry64, CA4.w[0], CA2.w[0]);
  449.     __sub_borrow_in_out (CA4.w[1], carry64, CA4.w[1], CA2.w[1],
  450.                          carry64);
  451.     CA4.w[2] = CA4.w[2] - CA2.w[2] - carry64;
  452.  
  453.     lx = ((double) CA4.w[2] * d128 +
  454.           ((double) CA4.w[1] * t64.d + (double) CA4.w[0]));
  455.     lq = lx / ly;
  456.  
  457.     CQT.w[1] = Q >> (64 - 60);
  458.     CQT.w[0] = Q << 60;
  459.     __add_128_128 (CQ, CQ, CQT);
  460.   }
  461.  
  462.   CY51.w[2] = CY.w[1] >> (64 - 51);
  463.   CY51.w[1] = (CY.w[1] << 51) | (CY.w[0] >> (64 - 51));
  464.   CY51.w[0] = CY.w[0] << 51;
  465.  
  466.   if (CA4.w[2] > CY51.w[2] || ((CA4.w[2] == CY51.w[2])
  467.                                &&
  468.                                (__unsigned_compare_gt_128 (CA4, CY51))))
  469.   {
  470.     // Q > 2^51
  471.  
  472.     // 2^(-49)*CA4/CY
  473.     d49.i = 0x3ce0000000000000ull;
  474.     lq *= d49.d;
  475.  
  476.     Q = (UINT64) lq - 1ull;
  477.  
  478.     // Q*CY
  479.     __mul_64x64_to_128 (A2, Q, CY.w[0]);
  480.     __mul_64x64_to_128 (A2h, Q, CY.w[1]);
  481.     A2.w[1] += A2h.w[0];
  482.     if (A2.w[1] < A2h.w[0])
  483.       A2h.w[1]++;
  484.  
  485.     // A2 <<= 49
  486.     CA2.w[2] = (A2h.w[1] << 49) | (A2.w[1] >> (64 - 49));
  487.     CA2.w[1] = (A2.w[1] << 49) | (A2.w[0] >> (64 - 49));
  488.     CA2.w[0] = A2.w[0] << 49;
  489.  
  490.     __sub_borrow_out (CA4.w[0], carry64, CA4.w[0], CA2.w[0]);
  491.     __sub_borrow_in_out (CA4.w[1], carry64, CA4.w[1], CA2.w[1],
  492.                          carry64);
  493.     CA4.w[2] = CA4.w[2] - CA2.w[2] - carry64;
  494.  
  495.     CQT.w[1] = Q >> (64 - 49);
  496.     CQT.w[0] = Q << 49;
  497.     __add_128_128 (CQ, CQ, CQT);
  498.  
  499.     lx = ((double) CA4.w[2] * d128 +
  500.           ((double) CA4.w[1] * t64.d + (double) CA4.w[0]));
  501.     lq = lx / ly;
  502.   }
  503.  
  504.   Q = (UINT64) lq;
  505.   __mul_64x64_to_128 (A2, Q, CY.w[0]);
  506.   A2.w[1] += Q * CY.w[1];
  507.  
  508.   __sub_128_128 (CA4, CA4, A2);
  509.   if ((SINT64) CA4.w[1] < 0) {
  510.     Q--;
  511.     CA4.w[0] += CY.w[0];
  512.     if (CA4.w[0] < CY.w[0])
  513.       CA4.w[1]++;
  514.     CA4.w[1] += CY.w[1];
  515.     if ((SINT64) CA4.w[1] < 0) {
  516.       Q--;
  517.       CA4.w[0] += CY.w[0];
  518.       if (CA4.w[0] < CY.w[0])
  519.         CA4.w[1]++;
  520.       CA4.w[1] += CY.w[1];
  521.     }
  522.   } else if (__unsigned_compare_ge_128 (CA4, CY)) {
  523.     Q++;
  524.     __sub_128_128 (CA4, CA4, CY);
  525.   }
  526.  
  527.   __add_128_64 (CQ, CQ, Q);
  528.  
  529.   pCQ->w[1] = CQ.w[1];
  530.   pCQ->w[0] = CQ.w[0];
  531.   pCA4->w[1] = CA4.w[1];
  532.   pCA4->w[0] = CA4.w[0];
  533.   return;
  534.  
  535.  
  536.  
  537. }
  538.  
  539. #endif
  540. #endif
  541.