/*

 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

 *

 * This code is free software; you can redistribute it and/or modify it

 * under the terms of the GNU General Public License version 2 only, as

 * published by the Free Software Foundation.

 *

 * This code 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

 * version 2 for more details (a copy is included in the LICENSE file that

 * accompanied this code).

 *

 * You should have received a copy of the GNU General Public License version

 * 2 along with this work; if not, write to the Free Software Foundation,

 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

 *

 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

 * or visit www.oracle.com if you need additional information or have any

 * questions.

 *

 */

#ifndef SHARE_VM_RUNTIME_ORDERACCESS_HPP

#define SHARE_VM_RUNTIME_ORDERACCESS_HPP

#include "memory/allocation.hpp"

//                Memory Access Ordering Model

//

//

// In the following, the terms 'previous', 'subsequent', 'before',

// 'after', 'preceding' and 'succeeding' refer to program order.  The

// terms 'down' and 'below' refer to forward load or store motion

// relative to program order, while 'up' and 'above' refer to backward

// motion.

//

// We define four primitive memory barrier operations.

//

// LoadLoad:   Load1(s); LoadLoad; Load2

//

// Ensures that Load1 completes (obtains the value it loads from memory)

// before Load2 and any subsequent load operations.  Loads before Load1

// may *not* float below Load2 and any subsequent load operations.

//

// StoreStore: Store1(s); StoreStore; Store2

//

// Ensures that Store1 completes (the effect on memory of Store1 is made

// visible to other processors) before Store2 and any subsequent store

// operations.  Stores before Store1 may *not* float below Store2 and any

// subsequent store operations.

//

// LoadStore:  Load1(s); LoadStore; Store2

//

// Ensures that Load1 completes before Store2 and any subsequent store

// operations.  Loads before Load1 may *not* float below Store2 and any

// subsequent store operations.

//

// StoreLoad:  Store1(s); StoreLoad; Load2

//

// Ensures that Store1 completes before Load2 and any subsequent load

// operations.  Stores before Store1 may *not* float below Load2 and any

// subsequent load operations.

//

//

//

//

//

//

//

//

// which implies a membar #StoreLoad between the store and load under

//

// Conventional usage is to issue a load_acquire for ordered loads.  Use

// release_store for ordered stores when you care only that prior stores

// are visible before the release_store, but don't care exactly when the

// store associated with the release_store becomes visible.  Use

// release_store_fence to update values like the thread state, where we

// don't want the current thread to continue until all our prior memory

// accesses (including the new thread state) are visible to other threads.

//

//

//

//                os::is_MP Considered Redundant

//

// Callers of this interface do not need to test os::is_MP() before

// issuing an operation. The test is taken care of by the implementation

// of the interface (depending on the vm version and platform, the test

// may or may not be actually done by the implementation).

//

//

//                A Note on Memory Ordering and Cache Coherency

//

// Cache coherency and memory ordering are orthogonal concepts, though they

// interact.  E.g., all existing itanium machines are cache-coherent, but

// the hardware can freely reorder loads wrt other loads unless it sees a

// load-acquire instruction.  All existing sparc machines are cache-coherent

// and, unlike itanium, TSO guarantees that the hardware orders loads wrt

// loads and stores, and stores wrt to each other.

//

// Consider the implementation of loadload.  *If* your platform *isn't*

// cache-coherent, then loadload must not only prevent hardware load

// instruction reordering, but it must *also* ensure that subsequent

// loads from addresses that could be written by other processors (i.e.,

// that are broadcast by other processors) go all the way to the first

// level of memory shared by those processors and the one issuing

// the loadload.

//

// So if we have a MP that has, say, a per-processor D$ that doesn't see

// writes by other processors, and has a shared E$ that does, the loadload

// barrier would have to make sure that either

//

// 1. cache lines in the issuing processor's D$ that contained data from

// addresses that could be written by other processors are invalidated, so

// subsequent loads from those addresses go to the E$, (it could do this

// by tagging such cache lines as 'shared', though how to tell the hardware

// to do the tagging is an interesting problem), or

//

// 2. there never are such cache lines in the issuing processor's D$, which

// means all references to shared data (however identified: see above)

// bypass the D$ (i.e., are satisfied from the E$).

//

// If your machine doesn't have an E$, substitute 'main memory' for 'E$'.

//

// Either of these alternatives is a pain, so no current machine we know of

// has incoherent caches.

//

// If loadload didn't have these properties, the store-release sequence for

// publishing a shared data structure wouldn't work, because a processor

// trying to read data newly published by another processor might go to

// its own incoherent caches to satisfy the read instead of to the newly

// written shared memory.

//

//

//                NOTE WELL!!

//

//                A Note on MutexLocker and Friends

//

// See mutexLocker.hpp.  We assume throughout the VM that MutexLocker's

// and friends' constructors do a fence, a lock and an acquire *in that

// order*.  And that their destructors do a release and unlock, in *that*

// order.  If their implementations change such that these assumptions

// are violated, a whole lot of code will break.

class OrderAccess : AllStatic {

 public:

  static void     loadload();

  static void     storestore();

  static void     loadstore();

  static void     storeload();

  static void     acquire();

  static void     release();

  static void     fence();

  static jbyte    load_acquire(volatile jbyte*   p);

  static jshort   load_acquire(volatile jshort*  p);

  static jint     load_acquire(volatile jint*    p);

  static jlong    load_acquire(volatile jlong*   p);

  static jubyte   load_acquire(volatile jubyte*  p);

  static jushort  load_acquire(volatile jushort* p);

  static juint    load_acquire(volatile juint*   p);

  static julong   load_acquire(volatile julong*  p);

  static jfloat   load_acquire(volatile jfloat*  p);

  static jdouble  load_acquire(volatile jdouble* p);

  static intptr_t load_ptr_acquire(volatile intptr_t*   p);

  static void*    load_ptr_acquire(volatile void*       p);

  static void*    load_ptr_acquire(const volatile void* p);

  static void     release_store(volatile jbyte*   p, jbyte   v);

  static void     release_store(volatile jshort*  p, jshort  v);

  static void     release_store(volatile jint*    p, jint    v);

  static void     release_store(volatile jlong*   p, jlong   v);

  static void     release_store(volatile jubyte*  p, jubyte  v);

  static void     release_store(volatile jushort* p, jushort v);

  static void     release_store(volatile juint*   p, juint   v);

  static void     release_store(volatile julong*  p, julong  v);

  static void     release_store(volatile jfloat*  p, jfloat  v);

  static void     release_store(volatile jdouble* p, jdouble v);

  static void     release_store_ptr(volatile intptr_t* p, intptr_t v);

  static void     release_store_ptr(volatile void*     p, void*    v);

  static void     release_store_fence(volatile jbyte*   p, jbyte   v);

  static void     release_store_fence(volatile jshort*  p, jshort  v);

  static void     release_store_fence(volatile jint*    p, jint    v);

  static void     release_store_fence(volatile jlong*   p, jlong   v);

  static void     release_store_fence(volatile jubyte*  p, jubyte  v);

  static void     release_store_fence(volatile jushort* p, jushort v);

  static void     release_store_fence(volatile juint*   p, juint   v);

  static void     release_store_fence(volatile julong*  p, julong  v);

  static void     release_store_fence(volatile jfloat*  p, jfloat  v);

  static void     release_store_fence(volatile jdouble* p, jdouble v);

  static void     release_store_ptr_fence(volatile intptr_t* p, intptr_t v);

  static void     release_store_ptr_fence(volatile void*     p, void*    v);

 private:

  // This is a helper that invokes the StubRoutines::fence_entry()

  // routine if it exists, It should only be used by platforms that

  // don't have another way to do the inline assembly.

  static void StubRoutines_fence();

};

#endif // SHARE_VM_RUNTIME_ORDERACCESS_HPP