/* * Copyright (c) 2013 * The President and Fellows of Harvard College. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef _MEMBAR_H_ #define _MEMBAR_H_ /* * Memory barriers. These create ordering barriers in CPU memory * accesses as actually issued by the CPU to the cache and memory * system. Because superscalar CPUs can execute many instructions at * once, they can potentially be retired in a different order from * what's written in your code. Normally this doesn't matter, but * sometimes it does (e.g. when writing to device registers) and in * those cases you need to insert memory barrier instructions to * create ordering guarantees. * * membar_load_load creates an ordering barrier between preceding * loads (from memory to registers) and subsequent loads, but has * (potentially) no effect on stores. This is what some people call a * "load fence". * * membar_store_store creates an ordering barrier between preceding * stores (from registers to memory) and subsequent stores, but has * (potentially) no effect on loads. This is what some people call a * "store" or "write fence". * * membar_store_any creates an ordering barrier between preceding * stores and subsequent stores *and* loads. Preceding loads may be * delayed past the barrier. This is the behavior needed for * operations comparable to spinlock_acquire(). * * membar_any_store creates an ordering barrier between preceding * loads and stores and subsequent stores. Following loads may be * executed before the barrier. This is the behavior needed for * operations comparable to spinlock_release(). * * membar_any_any creates a full ordering barrier, between preceding * loads and stores and following loads and stores. * * In OS/161 we assume that the spinlock operations include any memory * barrier instructions they require. (On many CPUs the synchronized/ * locked instructions used to implement spinlocks are themselves * implicit memory barriers.) You do not need to use membar_store_any * and membar_any_store unless rolling your own lock-like objects, * using atomic operations, implementing lock-free data structures, or * talking to hardware devices. * * There is a lot of FUD about memory barriers circulating on the * internet. Please ask your course staff if you have questions or * concerns. */ /* Inlining support - for making sure an out-of-line copy gets built */ #ifndef MEMBAR_INLINE #define MEMBAR_INLINE INLINE #endif MEMBAR_INLINE void membar_load_load(void); MEMBAR_INLINE void membar_store_store(void); MEMBAR_INLINE void membar_store_any(void); MEMBAR_INLINE void membar_any_store(void); MEMBAR_INLINE void membar_any_any(void); /* Get the implementation. */ #include #endif /* _MEMBAR_H_ */