1 /*
   2  * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #ifndef OS_CPU_WINDOWS_X86_ATOMIC_WINDOWS_X86_HPP
  26 #define OS_CPU_WINDOWS_X86_ATOMIC_WINDOWS_X86_HPP
  27 
  28 #include "runtime/os.hpp"
  29 
  30 // The following alternative implementations are needed because
  31 // Windows 95 doesn't support (some of) the corresponding Windows NT
  32 // calls. Furthermore, these versions allow inlining in the caller.
  33 // (More precisely: The documentation for InterlockedExchange says
  34 // it is supported for Windows 95. However, when single-stepping
  35 // through the assembly code we cannot step into the routine and
  36 // when looking at the routine address we see only garbage code.
  37 // Better safe then sorry!). Was bug 7/31/98 (gri).
  38 //
  39 // Performance note: On uniprocessors, the 'lock' prefixes are not
  40 // necessary (and expensive). We should generate separate cases if
  41 // this becomes a performance problem.
  42 
  43 #pragma warning(disable: 4035) // Disables warnings reporting missing return statement
  44 
  45 template<size_t byte_size>
  46 struct Atomic::PlatformAdd
  47   : Atomic::AddAndFetch<Atomic::PlatformAdd<byte_size> >
  48 {
  49   template<typename I, typename D>
  50   D add_and_fetch(I add_value, D volatile* dest, atomic_memory_order order) const;
  51 };
  52 
  53 #ifdef AMD64
  54 template<>
  55 template<typename I, typename D>
  56 inline D Atomic::PlatformAdd<4>::add_and_fetch(I add_value, D volatile* dest,
  57                                                atomic_memory_order order) const {
  58   return add_using_helper<int32_t>(os::atomic_add_func, add_value, dest);
  59 }
  60 
  61 template<>
  62 template<typename I, typename D>
  63 inline D Atomic::PlatformAdd<8>::add_and_fetch(I add_value, D volatile* dest,
  64                                                atomic_memory_order order) const {
  65   return add_using_helper<int64_t>(os::atomic_add_long_func, add_value, dest);
  66 }
  67 
  68 #define DEFINE_STUB_XCHG(ByteSize, StubType, StubName)                  \
  69   template<>                                                            \
  70   template<typename T>                                                  \
  71   inline T Atomic::PlatformXchg<ByteSize>::operator()(T exchange_value, \
  72                                                       T volatile* dest, \
  73                                                       atomic_memory_order order) const { \
  74     STATIC_ASSERT(ByteSize == sizeof(T));                               \
  75     return xchg_using_helper<StubType>(StubName, exchange_value, dest); \
  76   }
  77 
  78 DEFINE_STUB_XCHG(4, int32_t, os::atomic_xchg_func)
  79 DEFINE_STUB_XCHG(8, int64_t, os::atomic_xchg_long_func)
  80 
  81 #undef DEFINE_STUB_XCHG
  82 
  83 #define DEFINE_STUB_CMPXCHG(ByteSize, StubType, StubName)               \
  84   template<>                                                            \
  85   template<typename T>                                                  \
  86   inline T Atomic::PlatformCmpxchg<ByteSize>::operator()(T exchange_value, \
  87                                                          T volatile* dest, \
  88                                                          T compare_value, \
  89                                                          atomic_memory_order order) const { \
  90     STATIC_ASSERT(ByteSize == sizeof(T));                               \
  91     return cmpxchg_using_helper<StubType>(StubName, exchange_value, dest, compare_value); \
  92   }
  93 
  94 DEFINE_STUB_CMPXCHG(1, int8_t,  os::atomic_cmpxchg_byte_func)
  95 DEFINE_STUB_CMPXCHG(4, int32_t, os::atomic_cmpxchg_func)
  96 DEFINE_STUB_CMPXCHG(8, int64_t, os::atomic_cmpxchg_long_func)
  97 
  98 #undef DEFINE_STUB_CMPXCHG
  99 
 100 #else // !AMD64
 101 
 102 template<>
 103 template<typename I, typename D>
 104 inline D Atomic::PlatformAdd<4>::add_and_fetch(I add_value, D volatile* dest,
 105                                                atomic_memory_order order) const {
 106   STATIC_ASSERT(4 == sizeof(I));
 107   STATIC_ASSERT(4 == sizeof(D));
 108   __asm {
 109     mov edx, dest;
 110     mov eax, add_value;
 111     mov ecx, eax;
 112     lock xadd dword ptr [edx], eax;
 113     add eax, ecx;
 114   }
 115 }
 116 
 117 template<>
 118 template<typename T>
 119 inline T Atomic::PlatformXchg<4>::operator()(T exchange_value,
 120                                              T volatile* dest,
 121                                              atomic_memory_order order) const {
 122   STATIC_ASSERT(4 == sizeof(T));
 123   // alternative for InterlockedExchange
 124   __asm {
 125     mov eax, exchange_value;
 126     mov ecx, dest;
 127     xchg eax, dword ptr [ecx];
 128   }
 129 }
 130 
 131 template<>
 132 template<typename T>
 133 inline T Atomic::PlatformCmpxchg<1>::operator()(T exchange_value,
 134                                                 T volatile* dest,
 135                                                 T compare_value,
 136                                                 atomic_memory_order order) const {
 137   STATIC_ASSERT(1 == sizeof(T));
 138   // alternative for InterlockedCompareExchange
 139   __asm {
 140     mov edx, dest
 141     mov cl, exchange_value
 142     mov al, compare_value
 143     lock cmpxchg byte ptr [edx], cl
 144   }
 145 }
 146 
 147 template<>
 148 template<typename T>
 149 inline T Atomic::PlatformCmpxchg<4>::operator()(T exchange_value,
 150                                                 T volatile* dest,
 151                                                 T compare_value,
 152                                                 atomic_memory_order order) const {
 153   STATIC_ASSERT(4 == sizeof(T));
 154   // alternative for InterlockedCompareExchange
 155   __asm {
 156     mov edx, dest
 157     mov ecx, exchange_value
 158     mov eax, compare_value
 159     lock cmpxchg dword ptr [edx], ecx
 160   }
 161 }
 162 
 163 template<>
 164 template<typename T>
 165 inline T Atomic::PlatformCmpxchg<8>::operator()(T exchange_value,
 166                                                 T volatile* dest,
 167                                                 T compare_value,
 168                                                 atomic_memory_order order) const {
 169   STATIC_ASSERT(8 == sizeof(T));
 170   int32_t ex_lo  = (int32_t)exchange_value;
 171   int32_t ex_hi  = *( ((int32_t*)&exchange_value) + 1 );
 172   int32_t cmp_lo = (int32_t)compare_value;
 173   int32_t cmp_hi = *( ((int32_t*)&compare_value) + 1 );
 174   __asm {
 175     push ebx
 176     push edi
 177     mov eax, cmp_lo
 178     mov edx, cmp_hi
 179     mov edi, dest
 180     mov ebx, ex_lo
 181     mov ecx, ex_hi
 182     lock cmpxchg8b qword ptr [edi]
 183     pop edi
 184     pop ebx
 185   }
 186 }
 187 
 188 template<>
 189 template<typename T>
 190 inline T Atomic::PlatformLoad<8>::operator()(T const volatile* src) const {
 191   STATIC_ASSERT(8 == sizeof(T));
 192   volatile T dest;
 193   volatile T* pdest = &dest;
 194   __asm {
 195     mov eax, src
 196     fild     qword ptr [eax]
 197     mov eax, pdest
 198     fistp    qword ptr [eax]
 199   }
 200   return dest;
 201 }
 202 
 203 template<>
 204 template<typename T>
 205 inline void Atomic::PlatformStore<8>::operator()(T store_value,
 206                                                  T volatile* dest) const {
 207   STATIC_ASSERT(8 == sizeof(T));
 208   volatile T* src = &store_value;
 209   __asm {
 210     mov eax, src
 211     fild     qword ptr [eax]
 212     mov eax, dest
 213     fistp    qword ptr [eax]
 214   }
 215 }
 216 
 217 #endif // AMD64
 218 
 219 #pragma warning(default: 4035) // Enables warnings reporting missing return statement
 220 
 221 #endif // OS_CPU_WINDOWS_X86_ATOMIC_WINDOWS_X86_HPP