1 /*
   2  * Copyright (c) 1999, 2017, 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_VM_ATOMIC_WINDOWS_X86_HPP
  26 #define OS_CPU_WINDOWS_X86_VM_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 inline void Atomic::store    (jbyte    store_value, jbyte*    dest) { *dest = store_value; }
  46 inline void Atomic::store    (jshort   store_value, jshort*   dest) { *dest = store_value; }
  47 inline void Atomic::store    (jint     store_value, jint*     dest) { *dest = store_value; }
  48 
  49 inline void Atomic::store_ptr(intptr_t store_value, intptr_t* dest) { *dest = store_value; }
  50 inline void Atomic::store_ptr(void*    store_value, void*     dest) { *(void**)dest = store_value; }
  51 
  52 inline void Atomic::store    (jbyte    store_value, volatile jbyte*    dest) { *dest = store_value; }
  53 inline void Atomic::store    (jshort   store_value, volatile jshort*   dest) { *dest = store_value; }
  54 inline void Atomic::store    (jint     store_value, volatile jint*     dest) { *dest = store_value; }
  55 
  56 
  57 inline void Atomic::store_ptr(intptr_t store_value, volatile intptr_t* dest) { *dest = store_value; }
  58 inline void Atomic::store_ptr(void*    store_value, volatile void*     dest) { *(void* volatile *)dest = store_value; }
  59 
  60 template<size_t byte_size>
  61 struct Atomic::PlatformAdd
  62   : Atomic::AddAndFetch<Atomic::PlatformAdd<byte_size> >
  63 {
  64   template<typename I, typename D>
  65   D add_and_fetch(I add_value, D volatile* dest) const;
  66 };
  67 
  68 template<>
  69 struct Atomic::PlatformAdd<2>: Atomic::AddShortUsingInt {};
  70 
  71 #ifdef AMD64
  72 inline void Atomic::store    (jlong    store_value, jlong*    dest) { *dest = store_value; }
  73 inline void Atomic::store    (jlong    store_value, volatile jlong*    dest) { *dest = store_value; }
  74 
  75 template<>
  76 template<typename I, typename D>
  77 inline D Atomic::PlatformAdd<4>::add_and_fetch(I add_value, D volatile* dest) const {
  78   return add_using_helper<jint>(os::atomic_add_func, add_value, dest);
  79 }
  80 
  81 template<>
  82 template<typename I, typename D>
  83 inline D Atomic::PlatformAdd<8>::add_and_fetch(I add_value, D volatile* dest) const {
  84   return add_using_helper<intptr_t>(os::atomic_add_ptr_func, add_value, dest);
  85 }
  86 
  87 inline void Atomic::inc    (volatile jint*     dest) {
  88   (void)add    (1, dest);
  89 }
  90 
  91 inline void Atomic::inc_ptr(volatile intptr_t* dest) {
  92   (void)add_ptr(1, dest);
  93 }
  94 
  95 inline void Atomic::inc_ptr(volatile void*     dest) {
  96   (void)add_ptr(1, dest);
  97 }
  98 
  99 inline void Atomic::dec    (volatile jint*     dest) {
 100   (void)add    (-1, dest);
 101 }
 102 
 103 inline void Atomic::dec_ptr(volatile intptr_t* dest) {
 104   (void)add_ptr(-1, dest);
 105 }
 106 
 107 inline void Atomic::dec_ptr(volatile void*     dest) {
 108   (void)add_ptr(-1, dest);
 109 }
 110 
 111 inline jint     Atomic::xchg    (jint     exchange_value, volatile jint*     dest) {
 112   return (jint)(*os::atomic_xchg_func)(exchange_value, dest);
 113 }
 114 
 115 inline intptr_t Atomic::xchg_ptr(intptr_t exchange_value, volatile intptr_t* dest) {
 116   return (intptr_t)(os::atomic_xchg_ptr_func)(exchange_value, dest);
 117 }
 118 
 119 inline void*    Atomic::xchg_ptr(void*    exchange_value, volatile void*     dest) {
 120   return (void *)(os::atomic_xchg_ptr_func)((intptr_t)exchange_value, (volatile intptr_t*)dest);
 121 }
 122 
 123 #define DEFINE_STUB_CMPXCHG(ByteSize, StubType, StubName)               \
 124   template<>                                                            \
 125   template<typename T>                                                  \
 126   inline T Atomic::PlatformCmpxchg<ByteSize>::operator()(T exchange_value, \
 127                                                          T volatile* dest, \
 128                                                          T compare_value, \
 129                                                          cmpxchg_memory_order order) const { \
 130     STATIC_ASSERT(ByteSize == sizeof(T));                               \
 131     return cmpxchg_using_helper<StubType>(StubName, exchange_value, dest, compare_value); \
 132   }
 133 
 134 DEFINE_STUB_CMPXCHG(1, jbyte, os::atomic_cmpxchg_byte_func)
 135 DEFINE_STUB_CMPXCHG(4, jint,  os::atomic_cmpxchg_func)
 136 DEFINE_STUB_CMPXCHG(8, jlong, os::atomic_cmpxchg_long_func)
 137 
 138 #undef DEFINE_STUB_CMPXCHG
 139 
 140 inline jlong Atomic::load(const volatile jlong* src) { return *src; }
 141 
 142 #else // !AMD64
 143 
 144 template<>
 145 template<typename I, typename D>
 146 inline D Atomic::PlatformAdd<4>::add_and_fetch(I add_value, D volatile* dest) const {
 147   STATIC_ASSERT(4 == sizeof(I));
 148   STATIC_ASSERT(4 == sizeof(D));
 149   __asm {
 150     mov edx, dest;
 151     mov eax, add_value;
 152     mov ecx, eax;
 153     lock xadd dword ptr [edx], eax;
 154     add eax, ecx;
 155   }
 156 }
 157 
 158 inline void Atomic::inc    (volatile jint*     dest) {
 159   // alternative for InterlockedIncrement
 160   __asm {
 161     mov edx, dest;
 162     lock add dword ptr [edx], 1;
 163   }
 164 }
 165 
 166 inline void Atomic::inc_ptr(volatile intptr_t* dest) {
 167   inc((volatile jint*)dest);
 168 }
 169 
 170 inline void Atomic::inc_ptr(volatile void*     dest) {
 171   inc((volatile jint*)dest);
 172 }
 173 
 174 inline void Atomic::dec    (volatile jint*     dest) {
 175   // alternative for InterlockedDecrement
 176   __asm {
 177     mov edx, dest;
 178     lock sub dword ptr [edx], 1;
 179   }
 180 }
 181 
 182 inline void Atomic::dec_ptr(volatile intptr_t* dest) {
 183   dec((volatile jint*)dest);
 184 }
 185 
 186 inline void Atomic::dec_ptr(volatile void*     dest) {
 187   dec((volatile jint*)dest);
 188 }
 189 
 190 inline jint     Atomic::xchg    (jint     exchange_value, volatile jint*     dest) {
 191   // alternative for InterlockedExchange
 192   __asm {
 193     mov eax, exchange_value;
 194     mov ecx, dest;
 195     xchg eax, dword ptr [ecx];
 196   }
 197 }
 198 
 199 inline intptr_t Atomic::xchg_ptr(intptr_t exchange_value, volatile intptr_t* dest) {
 200   return (intptr_t)xchg((jint)exchange_value, (volatile jint*)dest);
 201 }
 202 
 203 inline void*    Atomic::xchg_ptr(void*    exchange_value, volatile void*     dest) {
 204   return (void*)xchg((jint)exchange_value, (volatile jint*)dest);
 205 }
 206 
 207 template<>
 208 template<typename T>
 209 inline T Atomic::PlatformCmpxchg<1>::operator()(T exchange_value,
 210                                                 T volatile* dest,
 211                                                 T compare_value,
 212                                                 cmpxchg_memory_order order) const {
 213   STATIC_ASSERT(1 == sizeof(T));
 214   // alternative for InterlockedCompareExchange
 215   __asm {
 216     mov edx, dest
 217     mov cl, exchange_value
 218     mov al, compare_value
 219     lock cmpxchg byte ptr [edx], cl
 220   }
 221 }
 222 
 223 template<>
 224 template<typename T>
 225 inline T Atomic::PlatformCmpxchg<4>::operator()(T exchange_value,
 226                                                 T volatile* dest,
 227                                                 T compare_value,
 228                                                 cmpxchg_memory_order order) const {
 229   STATIC_ASSERT(4 == sizeof(T));
 230   // alternative for InterlockedCompareExchange
 231   __asm {
 232     mov edx, dest
 233     mov ecx, exchange_value
 234     mov eax, compare_value
 235     lock cmpxchg dword ptr [edx], ecx
 236   }
 237 }
 238 
 239 template<>
 240 template<typename T>
 241 inline T Atomic::PlatformCmpxchg<8>::operator()(T exchange_value,
 242                                                 T volatile* dest,
 243                                                 T compare_value,
 244                                                 cmpxchg_memory_order order) const {
 245   STATIC_ASSERT(8 == sizeof(T));
 246   jint ex_lo  = (jint)exchange_value;
 247   jint ex_hi  = *( ((jint*)&exchange_value) + 1 );
 248   jint cmp_lo = (jint)compare_value;
 249   jint cmp_hi = *( ((jint*)&compare_value) + 1 );
 250   __asm {
 251     push ebx
 252     push edi
 253     mov eax, cmp_lo
 254     mov edx, cmp_hi
 255     mov edi, dest
 256     mov ebx, ex_lo
 257     mov ecx, ex_hi
 258     lock cmpxchg8b qword ptr [edi]
 259     pop edi
 260     pop ebx
 261   }
 262 }
 263 
 264 inline jlong Atomic::load(const volatile jlong* src) {
 265   volatile jlong dest;
 266   volatile jlong* pdest = &dest;
 267   __asm {
 268     mov eax, src
 269     fild     qword ptr [eax]
 270     mov eax, pdest
 271     fistp    qword ptr [eax]
 272   }
 273   return dest;
 274 }
 275 
 276 inline void Atomic::store(jlong store_value, volatile jlong* dest) {
 277   volatile jlong* src = &store_value;
 278   __asm {
 279     mov eax, src
 280     fild     qword ptr [eax]
 281     mov eax, dest
 282     fistp    qword ptr [eax]
 283   }
 284 }
 285 
 286 inline void Atomic::store(jlong store_value, jlong* dest) {
 287   Atomic::store(store_value, (volatile jlong*)dest);
 288 }
 289 
 290 #endif // AMD64
 291 
 292 #pragma warning(default: 4035) // Enables warnings reporting missing return statement
 293 
 294 #endif // OS_CPU_WINDOWS_X86_VM_ATOMIC_WINDOWS_X86_HPP