Revert "Remove gmp from AOS"

This reverts commit f37c97684f0910a3f241394549392f00145ab0f7.

We need gmp for SymEngine for symbolicmanipultion in C++

Change-Id: Ia13216d1715cf96944f7b4f422b7a799f921d4a4
Signed-off-by: Austin Schuh <austin.linux@gmail.com>
diff --git a/third_party/gmp/mpn/generic/divis.c b/third_party/gmp/mpn/generic/divis.c
new file mode 100644
index 0000000..f989ddb
--- /dev/null
+++ b/third_party/gmp/mpn/generic/divis.c
@@ -0,0 +1,194 @@
+/* mpn_divisible_p -- mpn by mpn divisibility test
+
+   THE FUNCTIONS IN THIS FILE ARE FOR INTERNAL USE ONLY.  THEY'RE ALMOST
+   CERTAIN TO BE SUBJECT TO INCOMPATIBLE CHANGES OR DISAPPEAR COMPLETELY IN
+   FUTURE GNU MP RELEASES.
+
+Copyright 2001, 2002, 2005, 2009, 2014, 2017, 2018 Free Software
+Foundation, Inc.
+
+This file is part of the GNU MP Library.
+
+The GNU MP Library is free software; you can redistribute it and/or modify
+it under the terms of either:
+
+  * the GNU Lesser General Public License as published by the Free
+    Software Foundation; either version 3 of the License, or (at your
+    option) any later version.
+
+or
+
+  * the GNU General Public License as published by the Free Software
+    Foundation; either version 2 of the License, or (at your option) any
+    later version.
+
+or both in parallel, as here.
+
+The GNU MP Library 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.
+
+You should have received copies of the GNU General Public License and the
+GNU Lesser General Public License along with the GNU MP Library.  If not,
+see https://www.gnu.org/licenses/.  */
+
+#include "gmp-impl.h"
+#include "longlong.h"
+
+
+/* Determine whether A={ap,an} is divisible by D={dp,dn}.  Must have both
+   operands normalized, meaning high limbs non-zero, except that an==0 is
+   allowed.
+
+   There usually won't be many low zero bits on D, but the checks for this
+   are fast and might pick up a few operand combinations, in particular they
+   might reduce D to fit the single-limb mod_1/modexact_1 code.
+
+   Future:
+
+   Getting the remainder limb by limb would make an early exit possible on
+   finding a non-zero.  This would probably have to be bdivmod style so
+   there's no addback, but it would need a multi-precision inverse and so
+   might be slower than the plain method (on small sizes at least).
+
+   When D must be normalized (shifted to low bit set), it's possible to
+   suppress the bit-shifting of A down, as long as it's already been checked
+   that A has at least as many trailing zero bits as D.  */
+
+int
+mpn_divisible_p (mp_srcptr ap, mp_size_t an,
+		 mp_srcptr dp, mp_size_t dn)
+{
+  mp_limb_t  alow, dlow, dmask;
+  mp_ptr     qp, rp, tp;
+  mp_limb_t di;
+  unsigned  twos;
+  int c;
+  TMP_DECL;
+
+  ASSERT (an >= 0);
+  ASSERT (an == 0 || ap[an-1] != 0);
+  ASSERT (dn >= 1);
+  ASSERT (dp[dn-1] != 0);
+  ASSERT_MPN (ap, an);
+  ASSERT_MPN (dp, dn);
+
+  /* When a<d only a==0 is divisible.
+     Notice this test covers all cases of an==0. */
+  if (an < dn)
+    return (an == 0);
+
+  /* Strip low zero limbs from d, requiring a==0 on those. */
+  for (;;)
+    {
+      alow = *ap;
+      dlow = *dp;
+
+      if (dlow != 0)
+	break;
+
+      if (alow != 0)
+	return 0;  /* a has fewer low zero limbs than d, so not divisible */
+
+      /* a!=0 and d!=0 so won't get to n==0 */
+      an--; ASSERT (an >= 1);
+      dn--; ASSERT (dn >= 1);
+      ap++;
+      dp++;
+    }
+
+  /* a must have at least as many low zero bits as d */
+  dmask = LOW_ZEROS_MASK (dlow);
+  if ((alow & dmask) != 0)
+    return 0;
+
+  if (dn == 1)
+    {
+      if (ABOVE_THRESHOLD (an, BMOD_1_TO_MOD_1_THRESHOLD))
+	return mpn_mod_1 (ap, an, dlow) == 0;
+
+      count_trailing_zeros (twos, dlow);
+      dlow >>= twos;
+      return mpn_modexact_1_odd (ap, an, dlow) == 0;
+    }
+
+  count_trailing_zeros (twos, dlow);
+  if (dn == 2)
+    {
+      mp_limb_t  dsecond = dp[1];
+      if (dsecond <= dmask)
+	{
+	  dlow = (dlow >> twos) | (dsecond << (GMP_NUMB_BITS-twos));
+	  ASSERT_LIMB (dlow);
+	  return MPN_MOD_OR_MODEXACT_1_ODD (ap, an, dlow) == 0;
+	}
+    }
+
+  /* Should we compute Q = A * D^(-1) mod B^k,
+                       R = A - Q * D  mod B^k
+     here, for some small values of k?  Then check if R = 0 (mod B^k).  */
+
+  /* We could also compute A' = A mod T and D' = D mod P, for some
+     P = 3 * 5 * 7 * 11 ..., and then check if any prime factor from P
+     dividing D' also divides A'.  */
+
+  TMP_MARK;
+
+  TMP_ALLOC_LIMBS_2 (rp, an + 1,
+		     qp, an - dn + 1); /* FIXME: Could we avoid this? */
+
+  if (twos != 0)
+    {
+      tp = TMP_ALLOC_LIMBS (dn);
+      ASSERT_NOCARRY (mpn_rshift (tp, dp, dn, twos));
+      dp = tp;
+
+      ASSERT_NOCARRY (mpn_rshift (rp, ap, an, twos));
+    }
+  else
+    {
+      MPN_COPY (rp, ap, an);
+    }
+  if (rp[an - 1] >= dp[dn - 1])
+    {
+      rp[an] = 0;
+      an++;
+    }
+  else if (an == dn)
+    {
+      TMP_FREE;
+      return 0;
+    }
+
+  ASSERT (an > dn);		/* requirement of functions below */
+
+  if (BELOW_THRESHOLD (dn, DC_BDIV_QR_THRESHOLD) ||
+      BELOW_THRESHOLD (an - dn, DC_BDIV_QR_THRESHOLD))
+    {
+      binvert_limb (di, dp[0]);
+      mpn_sbpi1_bdiv_qr (qp, rp, an, dp, dn, -di);
+      rp += an - dn;
+    }
+  else if (BELOW_THRESHOLD (dn, MU_BDIV_QR_THRESHOLD))
+    {
+      binvert_limb (di, dp[0]);
+      mpn_dcpi1_bdiv_qr (qp, rp, an, dp, dn, -di);
+      rp += an - dn;
+    }
+  else
+    {
+      tp = TMP_ALLOC_LIMBS (mpn_mu_bdiv_qr_itch (an, dn));
+      mpn_mu_bdiv_qr (qp, rp, rp, an, dp, dn, tp);
+    }
+
+  /* In general, bdiv may return either R = 0 or R = D when D divides
+     A. But R = 0 can happen only when A = 0, which we already have
+     excluded. Furthermore, R == D (mod B^{dn}) implies no carry, so
+     we don't need to check the carry returned from bdiv. */
+
+  MPN_CMP (c, rp, dp, dn);
+
+  TMP_FREE;
+  return c == 0;
+}