libatomic/gexch.c

   1 /* Copyright (C) 2012-2024 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 #define LAT_GEXCH
  26 #include "libatomic_i.h"
  27
  28
  29 /* If we natively support the exchange, and if we're unconcerned with extra
  30    barriers (e.g. fully in-order cpu for which barriers are a nop), then
  31    go ahead and expand the operation inline.  */
  32 #if !defined(WANT_SPECIALCASE_RELAXED) && !defined(__OPTIMIZE_SIZE__)
  33 # define EXACT_INLINE(N)                                        \
  34   if (C2(HAVE_ATOMIC_EXCHANGE_,N))                              \
  35     {                                                           \
  36       *PTR(N,rptr) = __atomic_exchange_n                        \
  37         (PTR(N,mptr), *PTR(N,vptr), __ATOMIC_SEQ_CST);          \
  38       return;                                                   \
  39     }
  40 #else
  41 # define EXACT_INLINE(N)
  42 #endif
  43
  44
  45 #define EXACT(N)                                                \
  46   do {                                                          \
  47     if (!C2(HAVE_INT,N)) break;                                 \
  48     if ((uintptr_t)mptr & (N - 1)) break;                       \
  49     EXACT_INLINE (N);                                           \
  50     *PTR(N,rptr) = C3(local_,exchange_,N)                       \
  51       (PTR(N,mptr), *PTR(N,vptr), smodel);                      \
  52     return;                                                     \
  53   } while (0)
  54
  55
  56 #define LARGER(N)                                               \
  57   do {                                                          \
  58     if (!C2(HAVE_INT,N)) break;                                 \
  59     if (!C2(MAYBE_HAVE_ATOMIC_CAS_,N)) break;                   \
  60     r = (uintptr_t)mptr & (N - 1);                              \
  61     a = (uintptr_t)mptr & -N;                                   \
  62     if (r + n <= N)                                             \
  63       {                                                         \
  64         pre_barrier (smodel);                                   \
  65         u.C2(i,N) = *PTR(N,a);                                  \
  66         do {                                                    \
  67           v = u;                                                \
  68           memcpy (v.b + r, vptr, n);                            \
  69         } while (!(C2(HAVE_ATOMIC_CAS_,N)                       \
  70                    ? __atomic_compare_exchange_n (PTR(N,a),     \
  71                         &u.C2(i,N), v.C2(i,N), true,            \
  72                         __ATOMIC_RELAXED, __ATOMIC_RELAXED)     \
  73                    : C3(local_,compare_exchange_,N) (PTR(N,a),  \
  74                         &u.C2(i,N), v.C2(i,N),                  \
  75                         __ATOMIC_RELAXED, __ATOMIC_RELAXED)));  \
  76         goto Lfinish;                                           \
  77       }                                                         \
  78   } while (0)
  79
  80
  81 static void __attribute__((noinline))
  82 libat_exchange_large_inplace (size_t n, void *mptr, void *vptr)
  83 {
  84 #define BUF     1024
  85
  86   char temp[BUF];
  87   size_t i = 0;
  88
  89   for (i = 0; n >= BUF; i += BUF, n -= BUF)
  90     {
  91       memcpy (temp, mptr + i, BUF);
  92       memcpy (mptr + i, vptr + i, BUF);
  93       memcpy (vptr + i, temp, BUF);
  94     }
  95   if (n > 0)
  96     {
  97       memcpy (temp, mptr + i, n);
  98       memcpy (mptr + i, vptr + i, n);
  99       memcpy (vptr + i, temp, n);
 100     }
 101
 102 #undef BUF
 103 }
 104
 105 void
 106 libat_exchange (size_t n, void *mptr, void *vptr, void *rptr, int smodel)
 107 {
 108   union max_size_u u, v;
 109   uintptr_t r, a;
 110
 111   switch (n)
 112     {
 113     case 0:                             return;
 114     case 1:             EXACT(1);       goto L4;
 115     case 2:             EXACT(2);       goto L4;
 116     case 4:             EXACT(4);       goto L8;
 117     case 8:             EXACT(8);       goto L16;
 118     case 16:            EXACT(16);      break;
 119
 120     case 3: L4:         LARGER(4);      /* FALLTHRU */
 121     case 5 ... 7: L8:   LARGER(8);      /* FALLTHRU */
 122     case 9 ... 15: L16: LARGER(16);     break;
 123
 124     Lfinish:
 125       post_barrier (smodel);
 126       memcpy (rptr, u.b + r, n);
 127       return;
 128     }
 129
 130   pre_seq_barrier (smodel);
 131   libat_lock_n (mptr, n);
 132
 133   if (vptr != rptr)
 134     {
 135       memcpy (rptr, mptr, n);
 136       memcpy (mptr, vptr, n);
 137     }
 138   else
 139     libat_exchange_large_inplace (n, mptr, vptr);
 140
 141   libat_unlock_n (mptr, n);
 142   post_seq_barrier (smodel);
 143 }
 144
 145 EXPORT_ALIAS (exchange);
 146 #undef LAT_GEXCH