[docs] Update HowToReleaseLLVM documentation.
[llvm-project.git] / compiler-rt / lib / builtins / ppc / gcc_qsub.c
blobac08120be0bd1e15057a8a3e07d57d54e2eae451
1 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
2 // See https://llvm.org/LICENSE.txt for license information.
3 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
5 // long double __gcc_qsub(long double x, long double y);
6 // This file implements the PowerPC 128-bit double-double add operation.
7 // This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!)
9 #include "DD.h"
11 long double __gcc_qsub(long double x, long double y) {
12 static const uint32_t infinityHi = UINT32_C(0x7ff00000);
14 DD dst = {.ld = x}, src = {.ld = y};
16 register double A = dst.s.hi, a = dst.s.lo, B = -src.s.hi, b = -src.s.lo;
18 // If both operands are zero:
19 if ((A == 0.0) && (B == 0.0)) {
20 dst.s.hi = A + B;
21 dst.s.lo = 0.0;
22 return dst.ld;
25 // If either operand is NaN or infinity:
26 const doublebits abits = {.d = A};
27 const doublebits bbits = {.d = B};
28 if ((((uint32_t)(abits.x >> 32) & infinityHi) == infinityHi) ||
29 (((uint32_t)(bbits.x >> 32) & infinityHi) == infinityHi)) {
30 dst.s.hi = A + B;
31 dst.s.lo = 0.0;
32 return dst.ld;
35 // If the computation overflows:
36 // This may be playing things a little bit fast and loose, but it will do for
37 // a start.
38 const double testForOverflow = A + (B + (a + b));
39 const doublebits testbits = {.d = testForOverflow};
40 if (((uint32_t)(testbits.x >> 32) & infinityHi) == infinityHi) {
41 dst.s.hi = testForOverflow;
42 dst.s.lo = 0.0;
43 return dst.ld;
46 double H, h;
47 double T, t;
48 double W, w;
49 double Y;
51 H = B + (A - (A + B));
52 T = b + (a - (a + b));
53 h = A + (B - (A + B));
54 t = a + (b - (a + b));
56 if (local_fabs(A) <= local_fabs(B))
57 w = (a + b) + h;
58 else
59 w = (a + b) + H;
61 W = (A + B) + w;
62 Y = (A + B) - W;
63 Y += w;
65 if (local_fabs(a) <= local_fabs(b))
66 w = t + Y;
67 else
68 w = T + Y;
70 dst.s.hi = Y = W + w;
71 dst.s.lo = (W - Y) + w;
73 return dst.ld;