libatomic/gcas.c

   1 /* Copyright (C) 2012-2022 Free Software Foundation, Inc.
   2    Contributed by Richard Henderson <rth@redhat.com>.
   3
   4    This file is part of the GNU Atomic Library (libatomic).
   5
   6    Libatomic is free software; you can redistribute it and/or modify it
   7    under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 3 of the License, or
   9    (at your option) any later version.
  10
  11    Libatomic is distributed in the hope that it will be useful, but WITHOUT ANY
  12    WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
  13    FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  14    more details.
  15
  16    Under Section 7 of GPL version 3, you are granted additional
  17    permissions described in the GCC Runtime Library Exception, version
  18    3.1, as published by the Free Software Foundation.
  19
  20    You should have received a copy of the GNU General Public License and
  21    a copy of the GCC Runtime Library Exception along with this program;
  22    see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
  23    <http://www.gnu.org/licenses/>.  */
  24
  25 #include "libatomic_i.h"
  26
  27
  28 /* If we natively support the cas, and if we're unconcerned with extra
  29    barriers (e.g. fully in-order cpu for which barriers are a nop), then
  30    go ahead and expand the operation inline.  */
  31 #if !defined(WANT_SPECIALCASE_RELAXED) && !defined(__OPTIMIZE_SIZE__)
  32 # define EXACT_INLINE(N)                                        \
  33   if (C2(HAVE_ATOMIC_CAS_,N))                                   \
  34     return __atomic_compare_exchange_n                          \
  35       (PTR(N,mptr), PTR(N,eptr), *PTR(N,dptr), false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)
  36 #else
  37 # define EXACT_INLINE(N)
  38 #endif
  39
  40 /* ... and if all that fails, invoke the function we generated elsewhere.
  41    Worst case, this will *also* use locks.  */
  42 #define EXACT(N)                                                \
  43   do {                                                          \
  44     if (!C2(HAVE_INT,N)) break;                                 \
  45     if ((uintptr_t)mptr & (N - 1)) break;                       \
  46     EXACT_INLINE (N);                                           \
  47     return C3(local_,compare_exchange_,N)                       \
  48       (PTR(N,mptr), PTR(N,eptr), *PTR(N,dptr), smodel, fmodel); \
  49   } while (0)
  50
  51 #define LARGER(N)                                               \
  52   do {                                                          \
  53     if (!C2(HAVE_INT,N)) break;                                 \
  54     if (!C2(HAVE_ATOMIC_LDST_,N)) break;                        \
  55     if (!C2(MAYBE_HAVE_ATOMIC_CAS_,N)) break;                   \
  56     r = (uintptr_t)mptr & (N - 1);                              \
  57     a = (uintptr_t)mptr & -N;                                   \
  58     if (r + n <= N)                                             \
  59       {                                                         \
  60         pre_barrier (smodel);                                   \
  61         u.C2(i,N) = __atomic_load_n (PTR(N,a), __ATOMIC_RELAXED); \
  62         do {                                                    \
  63           if (memcmp (u.b + r, eptr, n) != 0) goto Lfail;       \
  64           v = u; memcpy (v.b + r, dptr, n);                     \
  65         } while (!(C2(HAVE_ATOMIC_CAS_,N)                       \
  66                    ? __atomic_compare_exchange_n (PTR(N,a),     \
  67                         &u.C2(i,N), v.C2(i,N), true,            \
  68                         __ATOMIC_RELAXED, __ATOMIC_RELAXED)     \
  69                    : C3(local_,compare_exchange_,N) (PTR(N,a),  \
  70                         &u.C2(i,N), v.C2(i,N),                  \
  71                         __ATOMIC_RELAXED, __ATOMIC_RELAXED)));  \
  72         goto Lsucc;                                             \
  73       }                                                         \
  74   } while (0)
  75
  76
  77
  78 bool
  79 libat_compare_exchange (size_t n, void *mptr, void *eptr, void *dptr,
  80                         int smodel, int fmodel)
  81 {
  82   union max_size_u u, v;
  83   uintptr_t r, a;
  84   bool ret;
  85
  86   switch (n)
  87     {
  88     case 0:                             return true;
  89     case 1:             EXACT(1);       goto L4;
  90     case 2:             EXACT(2);       goto L4;
  91     case 4:             EXACT(4);       goto L8;
  92     case 8:             EXACT(8);       goto L16;
  93     case 16:            EXACT(16);      break;
  94
  95     case 3: L4:         LARGER(4);      /* FALLTHRU */
  96     case 5 ... 7: L8:   LARGER(8);      /* FALLTHRU */
  97     case 9 ... 15: L16: LARGER(16);     break;
  98
  99     Lsucc:
 100       post_barrier (smodel);
 101       return true;
 102     Lfail:
 103       post_barrier (fmodel);
 104       memcpy (eptr, u.b + r, n);
 105       return false;
 106     }
 107
 108   pre_seq_barrier (smodel);
 109   libat_lock_n (mptr, n);
 110
 111   ret = memcmp (mptr, eptr, n) == 0;
 112   memcpy ((ret ? mptr : eptr), (ret ? dptr : mptr), n);
 113
 114   libat_unlock_n (mptr, n);
 115   post_seq_barrier (ret ? smodel : fmodel);
 116
 117   return ret;
 118 }
 119
 120 EXPORT_ALIAS (compare_exchange);