Warning: Unstable document. This may change at any minute. It is unlikely to be entirely consistent.

A Memory Model for C++: Strawman Proposal

This is an attempt to outline a "memory model" for C++. It addresses the multi-threaded semantics of C and C++, particularly with respect to memory visibility. We concentrate on the question of what values a load of an object (i.e. an l-value to r-value conversion) may observe.

We somewhat informally address the issues we have thought of to date.

We use a style in which the core document is presented in the default font and color. We occasionally insert additional more detailed discussion, motivation, and status as bracketed text in a smaller green font. This additional text is not fundamental to understanding the proposal.

Rationale for the overall approach

One possible approach to a memory model would be to essentially copy the Java memory model.

The reasons we are not currently pursuing that route, in spite of the overlap among the participants are:

Hence we are currently pursuing an approach which has been less well explored, and is thus probably riskier. We concentrate on precisely defining the notion of a data race (a deficiency of the pthread definition), but then largely stick with the pthreads approach of leaving the semantics of data races undefined.

This is complicated somewhat by the desire to support "atomic operations" which are effectively synchronization operations that can legitimately participate in data races.

[ This is at least our third attempt at a description of the memory model. The first attempt was to define the semantics as sequential consistency for programs that had no data races, where a data race was defined as consecutive execution of conflicting operations in a sequentially consistent execution. That was nice and simple. But it makes it very difficult to define synchronization primitives that allow some reordering around them. These include even a pthread-like locking primitive that has only "acquire" semantics, i.e. allows preceding memory operations to be moved into the critical section. Surprisingly, at least for one of us, pthread_mutex_lock does not have that property, due to the presence of pthread_mutex_trylock. ]

Data Races and our Approach to Memory Consistency

We approach the problem of defining a memory model or multi-threaded semantics for C++ as follows:
  1. We define a memory location to be a scalar variable, data member, or array element, or a maximal sequence of contiguous, non-zero-length, bit-fields in a struct or class. (This almost corresponds to the meaning of "object" in the current standard.)
  2. We define an action performed by a thread to be a particular program point in the source program, together with the value it produced and/or stored. Each action corresponds to either a call, or a primitive operation, or an access of a memory location. (We view built-in assignments of struct or class objects as sequences of assignments to their fields.) We are primarily interested in the latter, which we classify as either load or store actions. Operations provided by the atomic operations library result in atomic or synchronization actions, as do locking primitives. Each such action may be an atomic load, atomic store, both or neither.

    [ There is an argument for introducing alternate syntax for atomic operations, e.g. by describing variables as __async volatile. (It was concluded that recycling the plain volatile keyword introduces a conflict with existing usage, a relatively small, but nonzero, fraction of which is demonstrably legitimate. Whether or not that constitutes sufficient reason to leave the current largely platform-dependent semantics of volatile in place remains controversial.) Introducing some such syntax may make it easier to code idioms like double-checked locking, and to use consistent idioms across programming languages. Without a better understanding of the form the atomic operations library will take, it is unclear whether this argument is valid. The argument against it is simplicity, and elimination of an ugly or reused keyword. ]

  3. Below we define a notion of a consistent execution. An execution will consist of a set of actions, corresponding to the steps performed by each of the threads, together with some relations that describe the order in which each thread performs those actions, and the ways in which the threads interact. For an execution to be consistent, both the actions and the associated relations have to satisfy a number of constraints, described below.

    In the absence of atomic operations, which may concurrently operate on the same data for different threads, this notion is equivalent to Lamport's definition of sequential consistency.

  4. We define when a consistent execution has a data race on a particular input. Note that the term is used to denote concurrent conflicting accesses to a location using ordinary assignments. Concurrent accesses through special "atomic operations" do not constitute a data race.
  5. Programs have undefined semantics on executions on which a data race is possible. We effectively view a data race as an error.
  6. If a program cannot encounter a data race on a given input, then it is guaranteed to behave in accordance with one of its consistent executions.
Note that although our semantics is essentially defined in terms of sequential consistency, it is much weaker than sequential consistency. We essentially forbid all programs which depend implicitly on memory ordering. Hence we continue to allow substantial amounts of memory reordering by either the compiler or hardware.

We now define two kinds of order on memory operations (loads, stores, atomic updates) performed by a single thread. The first one is intended to replace the notion of sequence point that is currently used in parts of the standard:

The is-sequenced-before relation

If a memory update or side-effect a is-sequenced-before another memory operation or side-effect b, then informally a must appear to be completely evaluated before b in the sequential execution of a single thread, e.g. all accesses and side effects of a must occur before those of b. This notion does not directly imply anything about the order in which memory updates become visible to other threads.

We will say that a subexpression A of the source program is-sequenced-before another subexpression B of the same source program to indicate that all side-effects and memory operations performed by an execution of A occur-before those performed by the corresponding execution of B, i.e. as part of the same execution of the smallest expression that includes them both.

We propose roughly that wherever the current standard states that there is a sequence point between A and B, we instead state that A is-sequenced-before B. This will constitute the precise definition of is-sequenced-before on subexpressions, and hence on memory actions and side effects.

A very similar change in the specification of intra-thread sequencing of operations is being simultaneously proposed by Clark Nelson in N1944=06-0014, which explores the issue in more detail. Our hope is that a proposal along these lines will be accepted, and can serve as the precise definition of is-sequenced-before.

[ Based on recent email discussions, at this point there appears to be some uncertainty in the interpretation of the current standard, which is complicating matters. The main issue seems to be the precise meaning of the restriction on interleaving of function calls in arguments. It appears important to resolve this even in the single-threaded case. ]

The is-inter-thread-ordered-before relation

An action A is-inter-thread-ordered-before an action B if they are both executed by the same thread, and one of them is an atomic or synchronization operation that guarantees appropriate inter-thread visibility ordering. We specify these ordering constraints with the atomic operations library.

An ordinary memory operation is never inter-thread-ordered-before another ordinary memory operation.

Most atomic operations will specify some combination of acquire and release ordering constraints, which enforce ordering with respect to subsequent and prior memory actions, respectively. These constraints are reflected in the is-inter-thread-ordered-before relation.

Lock acquisition imposes at least an acquire constraint, and lock release will normally impose a release constraint. Whenever an action A has an acquire constraint, and A is-sequenced-before B, then A is-inter-thread-ordered-before B. Whenever A has a release constraint, and B is-sequenced-before A, then B is-inter-thread-ordered-before A.

The depends-on relation

Consider a given execution of a particular thread, i.e. the sequence of actions that may be performed by a particular thread as part of the execution of the containing program. If, as a result of changing only the value read by an atomic load L, a subsequent atomic store S either can no longer occur, or must store a different value, then S depends on L.

Note that the definition of depends-on is relative to a particular execution, and always involves a dependence of an atomic store on an atomic load. Ordinary memory operation do not depend-on each other. We need the depends-on relation only to outlaw certain anomalous executions of atomic operations that informally violate causality, i.e. in which an atomic operation causes itself to be executed in a particular way.

[ In the more formal proposal Clark and Hans are working on, this is merged into the is-inter-thread-ordered-before relation in order to avoid defining so many distinct relations. That appears to be cleaner. ]

We next discuss relations on actions between threads.

The communicates-with relation

We specify interactions between threads in an execution using a communicates-with relation. Informally, an action A communicates-with another action B if B "sees" the result of A. The definition of each kind of atomic operation will specify the other operations with which it can communicate. A store to an ordinary variable communicates-with a load that retrieves the stored value.

Informally, a lock release communicates-with the next acquisition of the same lock. A barrier (in the pthread_barrier_wait sense or OpenMP sense) communicates-with all corresponding executions of the barrier in other threads. A memory fence (or barrier in the other sense) communicates-with the next execution of the fence, usually by another thread. An atomic store communicates-with all atomic loads that read the value saved by the store, i.e. for this purpose they behave like ordinary loads and stores.

[ In the more formal proposal, we rename communicates-with back to synchronizes-with and limit it to synchronization operations. The down side is that it becomes harder to express the constraint that happens-before as defined below must be acyclic. That is restated in the more formal proposal. ]

We now have enough machinery to describe the ordering we really use to describe memory visibility.

The happens-before relation

[ Note that this definition of happens-before is a bit different from that used in Lamport's original 1978 paper ("Time, Clocks, and the Ordering of Events in Distributed Systems", CACM 21,7), and eventually in the Java model, but it is essentially just an adaptation of Lamport's definition to a system in which actions within a thread are also not totally ordered. The detailed style of definition grew out of a discussion with Bill Pugh, though we're not yet sure he approves of the result. ]

Given is-sequenced-before, is-inter-thread-ordered-before, depends-on and communicates-with relations on a set of actions, we define the happens-before relation to be the smallest transitively closed relation satisfying the following constraints:

Consistent executions

A program execution is a quintuple consisting of
  1. set of thread actions, and corresponding
  2. is-sequenced-before,
  3. is-inter-thread-ordered-before,
  4. depends-on, and
  5. communicates-with relations.
These give rise to a corresponding happens-before relation. We say that an execution is consistent if:
  1. The actions of any particular thread (excluding values read from potentially shared locations), and the corresponding is-sequenced-before relation, is-inter-thread-ordered-before relation, and depends-on relation, are all consistent with the normal sequential semantics as given in the rest of the standard.
  2. The communicates-with relation is such that for every ordinary load L which sees a value stored by another thread, a store S communicates-with L, such that S stores the value seen by L
  3. The communicates-with relation is consistent with the constraints imposed by the definitions of the synchronization primitives. For example, if S is an atomic store which communicates-with an atomic load L, then the loaded and stored values must be the same.
  4. (intra-thread visibility) If a load L sees a store S from the same thread, then L must not be sequenced-before S, and there must be no intervening store S' such that S is-sequenced-before S' and S' is-sequenced-before L.
  5. (inter-thread visibility) Each load L of any shared variable (including synchronization variables) sees a store S, such that L does not happen-before S and such that there is no intervening store S' such that S happens-before S' and S' happens-before L.
  6. The happens-before relation is "acyclic", i.e. no action happens-before itself.

    [ This means we view the relation as normally irreflexive. If we normally want it to be reflexive, we can tweak this slightly. ]

Note that if no thread performs any synchronization actions then the happens-before relation requires that the actions of a given thread effectively occur in "is-sequenced-before" order, which is as close as C++ gets to purely sequential execution. This in this case an execution is consistent iff it is sequentially consistent.

If lock/unlock enforce only acquire/release ordering, and there is no other form of synchronization, then it is less apparent that our definition is equivalent to sequential consistency. However, this can be proven if there is no trylock primitive.

[ At least some of us believe that the most plausible interpretation of the existing pthread semantics can be closely approximated by defining the various lock() primitives such that they have both acquire and release semantics. This still leaves issues related to failing pthread calls, etc. We believe that these introduce no fundamental technical challenges, but the details are not currently completely clear. ]

[ The fact that we insist that we require much stronger ordering for ordinary memory accesses than for atomic accesses initially seems out of place here. But, as Bill Pugh points out, simple Java-like happens-before consistency is otherwise insufficient. (For an example, see below.) And the ordering constraints on ordinary memory actions really only affect the definition of a data race; the meaning of data-race-free program is not affected, since this ordering is invisible. ]

Data races

We define a memory location to be a variable, (non-bitfield) data member, array element, or contiguous sequence of bitfields. We define two actions to be conflicting if they access the same memory location, and at least one of them is a store access.

We define an execution to contain an intra-thread race if a thread performs two conflicting actions, and neither is-sequenced-before the other.

We define an execution to contain an inter-thread race if two threads perform two conflicting actions, neither happens-before the other, and at least one is not a synchronization action.

[ I'm not sure this is quite the best way to state this, since the "communicates-with" relation on ordinary memory accesses contributes to happens-before. Thus "conflicting" actions may "communicate-with" each other, and thus "happen-before" each other, and thus no longer conflict. I think this technically doesn't matter because non-conflicting actions on ordinary memory normally only imply happens-before relationships that must exist anyway. And if there is an execution with an initial conflict that is eliminated by the "communicates-with" edge generated by the conflict, then there is an alternate consistent execution in which that edge doesn't exist, and there is a real conflict. But I don't like the fact that a subtle argument is required to demonstrate that this definition is sane.

We do need to include the extra "communicates-with" edges in the requirement that happens-before is acyclic. Otherwise we get nothing like sequential consistency in the data-race-free case.

I the more formal proposal, this is changed, but as a result we do need to introduce an extended-happens-before relation to check that it is acyclic. ]

If, for a given program and input, there are consistent executions containing either kind of race, then the program has undefined semantics on that input. Otherwise the program behaves according to one of its consistent executions.

[ Bill Pugh points out that that the notion of "input" here isn't well defined for a program that interacts with its environment. And we don't want to give undefined semantics to a program just because there is some other sequence of interactions with the environment that results in a data race. We probably want something more along the lines of stating that every program behavior either

  1. corresponds to a consistent execution in which loads see stores that happen-before them, or
  2. there is a consistent execution with a data race, such that calls to library IO functions before the data race are consistent with observed behavior.

I think the notion of "before" in the second clause is easily definable, since we can insist that IO operations be included in the effectively total order of ordinary variable accesses.

It is unclear to me whether this is something that needs to be addressed with great precision, since the current standard doesn't appear to, and I think the danger of confusion is minimal. ]

For purposes of the above definitions, object destruction is viewed as a store to all its sub-objects, as is assignment to an object through a char * pointer if the target object was constructed as a different type. Assignment to a component of a particular union member is treated as a store into all components of the other union members. Different threads may not concurrently access different union members.

Discussion and Examples

Unlike the Java memory model, we do not insist on a total order between synchronization actions. Instead we insist only on a communicates-with relation, which must be an irreflexive partial order. (This follows from the fact that it is a subset of happens-before, which is irreflexive and transitive.) This means that synchronization actions such as atomic operations are themselves not guaranteed to be seen in a consistent order by all other threads, and may become visible at other threads in an order different from the intra-thread is-sequenced-before order.

Simple Locks

In the case of simple locks, this is possible only in that a "later" (in is-sequenced-before order) lock acquisition may become visible before an "earlier" unlock action on a different thread. Thus (with hypothetical syntax and lock and unlock primitives that take references and which have only acquire and release semantics, respectively):
lock(a); x = 1; unlock(a); lock(b); y = 1; unlock(b);
may appear to another thread to be executed as
lock(a); lock(b); y = 1; x = 1; unlock(a); unlock(b);
Unlike in Java, a hypothetical observer thread might see the assignments occur out of order. However, so long as x and y are ordinary variables, and the assignments do not reflect atomic operations, this is not observable, since such an observer thread would introduce a race.

Note that although our semantics allows the above reordering in a particular execution, compilers may not in general perform such rearrangements, since that might introduce deadlocks.

We claim that for data-race-free programs using only simple locks and no atomic operations, our memory model is identical to the Java one.

In the case of simple locks, we effectively insist on a total order among synchronization operations, happens-before is an irreflexive partial order, and everything behaves as in the Java memory model.

Simple Atomic Operations

For the remainder of this section, we assume that all variables are initialized to zero, variables whose names begin with r are local, and all other variables are shared, and that all operations are considered to be synchronization operations, and hence may be safely involved in data races, even if they are written using simple assignment syntax. Acquire and release operations will be explicitly specified, however.

Next consider load_acquire and store_release primitives, where there must be an action that communicates-with every load_acquire: either an initialization or a store_release on the same variable. But there are no other restrictions on communicates-with.

Consider the following example:

Thread1 Thread2
x = 1;
store_release(&flag, 1);
r1 = load_acquire(&flag);
r2 = x;

Due to the acquire and release ordering constraints on the references to flag, the individual pairs of assignments in each thread are ordered by is-inter-thread-ordered-before.

Only two possible actions can communicate-with the load_acquire in this example: The initialization of the flag variable, or the store_release action. It follows that if we get r1 = 1, then the store_release must have communicated-with the load-acquire. Given the ordering constraints,this implies that the assignment to x happens-before the assignment to r2. Hence r1 = 1 and r2 = 0 is an impossible outcome, as desired.

Next consider the following example, under the same ground rules:

Thread1 Thread2
store_release(&y, 1);
r1 = load_acquire(&x);
store_release(&x, 1);
r2 = load_acquire(&y);

There is no is-inter-thread-ordered-before ordering between the statements in each thread. Initialization operations may synchronize with both load_acquire operations. Hence we can get r1 = r2 = 0, as expected.

We can model memory fences as synchronization operations with both acquire and release semantics, which must be totally ordered, and the communicates-with relation must respect the total order, i.e. each fence instance communicates-with the next fence instance in the total order. Consider:

Thread1 Thread2
x = 1;
r1 = z;
y = 1;
r2 = w;
w = 1;
r3 = y;
z = 1;
r4 = x;

In any given execution either the thread 1 fence communicates-with the thread 2 fence, i.e. the thread 1 fence executes first (a), or vice-versa (b). If r3 = 1, then we must be in the first case. (If thread 2's fence came first, the load of y happens-before the store, and hence cannot see the store.) Hence r4 must also be one. Similarly r1 = 1 implies r2 = 1.

Next consider the following suggestive example, where initially self_destruct_left and self_destruct_right are false:
Thread1 Thread2
while (!self_destruct_left);
self_destruct_right = true;
while (!self_destruct_right);
self_destruct_left = true;
We would like to avoid the situation in which each while-loop condition sees the store from the other thread, and the ship spontaneously self-destructs without intervention of a third thread.

Assume there is no other thread which sets either of the self_destruct... variables. Assume the thread 1 loop nonetheless terminates. This is only possible if it saw the store in thread 2. For such an execution to be intra-thread consistent, thread 2's loop must have seen the store from thread1. Thus in each case, the store depends-on the load in the while loop in the same thread, and the store communicates-with the loop in the other thread. The happens-before relation must be consistent with all of these and transitively closed. Hence all actions mentioned are in a cycle, i.e. happen-before themselves. Thus such an execution is not allowed.

If we use ordinary memory operations instead of atomic operations in the above example, then the loads in the while loops are sequenced-before the stores in the same thread. We need the same communicates-with relationships as before, thus happens-before will again be cyclic, and this version of the program is also safe.

[ Unlike earlier versions of this proposal, the presence of the depends-on relation seems to introduce enough of a causality requirement here to prevent the anomalous outcome if we use unordered atomic operations. With ordinary memory operations there is nothing surprising. By adding communicates-with relationships for all matching cross-thread store-load pairs, and insisting that the result not contain cycles, we are essentially insisting on sequential consistency. Needs more examples. ]

Member and Bitfield Assignments

As we stated above, struct and class members are normally considered to be separate memory locations, and thus assignments to distinct fields do not conflict. The only exception to this is that an assignment to a bit-field conflicts with accesses to any other bit-field in the same sequence of contiguous bit-fields. For example, consider the declaration:

struct s {
  char a;
  int  b:9;
  int  c:5;
  char d;
  int  e:1;

An assignment to the bit-field b conflicts with an access to c, but no accesses to a different pair of fields conflicts.

Note that in some existing ABIs, fields a,b, c, and d are allocated to the same 32-bit word. With such ABIs, compilers may not implement an assignment to b as a 32-bit load, followed by an in-register bit replacement, followed by a 32-bit store. Such implementations do not produce the correct result if a or d are updated concurrently with b.

[ Note that the above example illustrates the most controversial aspect of this rule that we have found. It will take work to make existing compilers conform. For example, gcc on X86 currently does not. However, the resulting slowdown in object code appears to be very minor. To our knowledge, cases like the above exist, but are rare, in production code. And the necessary overhead amounts to some additional in-register computation, and a second store instruction to the same cache line as the first.

The alternatives would add complexity to the programming task, which some of us believe we want to avoid unless the benefits are much larger than this. Consider the slight variant of the above example:

struct s {
  something_declared_third_party_header a;
  int  b:9;
  int  c:5;
  char d;
  int  e:1;

Whether or not accesses to a and b conflict is now hard to predict without implementation knowledge of a's type, even if we understand the ABI conventions.

Probably the only plausible alternative is to allow bit-field accesses to conflict with accesses to any other data member in the same struct or class, and to thus encourage bit-fields to only appear in a nested struct. But this would add a rather obscure rule to the already complicated set that must be understood by threads programmers. ]

As far as the programmer is concerned, padding bytes are not written as part of a bit-field or data member update, i.e. such writes do not need to be considered in determining whether there is a data race. But we are not aware of cases in which a compiler-generated store that includes a padding byte would adversely impact correctness.


The preceding rules have some significant non-obvious consequences. Here we list some of these rather informally. Proofs would clearly be useful, where they are missing. Here we list some more detailed implications of the last statement:


We make no specific guarantees that initialization performed by a constructor will be seen to have been completed when another thread accesses the object. A correct program must use some sort of synchronization to make a newly created object visible to another thread. Otherwise the store of the object point and the load by the other thread constitute a data race, and the program has undefined semantics.

If proper synchronization is used, there is no need for additional guarantees, since the synchronization will ensure visibility as needed.

This is consistent with current practice, which does not guarantee even vtable visibility in the absence of synchronization, and allows insufficiently synchronized programs to crash, jump to the wrong member function, etc.

Function local statics

If the declaration of a function local static is preceded by the keyword protected, then the access to the implicit flag used to track whether a function local static has been constructed is a synchronization operation. Otherwise it is not, and it is the programmer's responsibility to ensure that neither construction of the object not reference to the implicit flag variable introduces a data race.

[There appeared to be consensus among those attending the Lillehammer C++ standards meeting that both options should be provided to the programmer. Subsequent discussion pointed out that it is more reasonable than some of us had thought to always require thread-safety. In particular, there seem to be no practical cases in which a compiler decision to implement an initialization statically breaks ordering guarantees that would reasonably be expected. The down side is that this imposes some overhead on uses that do not require synchronization. On X86, this overhead can be significant for the initialization, but probably not for later uses. On some other architectures significant overhead may be introduced even for later references. Currently I think this issue is unresolved.

The protected keyword was chosen arbitrarily, and should be considered more carefully. ]

Volatile variables and data members

Accessed to regular volatile variables are not viewed as synchronization operations. Volatile implies only safety in the presence of implicit or unpredictable actions by the same thread.

If the atomic operations library turns out to be insufficiently convenient to provide for lock-free inter-thread communication, we propose that accesses to __async volatile variables and data members are viewed as synchronization operations.

Loads of such variables would have an acquire ordering constraint, and stores would have a release constraint.

[ It seems to make sense to put this on hold until we have a better handle on the atomic operations library, so that we can tell whether that would be a major inconvenience.

The possible reasons to retain this are (1) possibly improved convenience, and (2) possibly better consistency in programming idioms across languages (in this case Java and C#). The argument for discarding it is simplicity.

If we want to retain it, we now have to ask whether there is a total order among volatile accesses.

Current implementations of volatile generally use weaker semantics, which do not prevent hardware reordering of volatiles. This appears to have no use in portable code for threads, since such code cannot take advantage of the fact that operations are reordered "only by the hardware". It is occasionally useful for variables that are either modified after a setjmp, that may be accessible through multiple memory mappings, or the like. ]

There are no atomicity guarantees for accesses to volatile variables. Accesses to __async volatile variables of pointer types, integral types (other than long long variants), bool type, enumeration types, and type bool are atomic. The same applies to the individual data member accesses in e.g. struct assignments, but not to the assignment as a whole. There is no defined order between these individual atomic operations.

[We can't talk about async-signal-safety here. We might suggest that __async volatile int and __async volatile pointers be async-signal-safe where that's possible and meaningful. My concern here is with uniprocessor embedded platforms, which might have to use restartable critical sections to implement atomicity, and might misalign things. }

Thread-local variables and stack locations

This issue is addressed more thoroughly in Lawrence Crowl's proposal (N1874=05-0134). We defer to that discussion.

Library Changes and Clarifications

We need the following kinds of changes to the library specification:

Clarify thread safety

The library specification needs to be clear as to which pieces of the library are thread-safe, and in what sense, and how various calls interact with the memory model. We propose the following basic approach, consistent with the approach used by the SGI STL implementation: We expect that some effort will be required to pin down exactly which operations "logically update" shared state.

[ Paragraph 21.3(5), which deals with basic_string copy-on-write support, will be difficult to support here in any reasonable fashion. I've long advocated stripping it to prohibit copy-on-write since I'm not convinced it makes sense without threads either. Unfortunately, removing it will drastically change the performance characteristics of existing implementations, often for the better, but occasionally for the worse. ]

Add thread-specific library components

At least two kinds of additions will be needed: In the long term, a library containing basic lock-free and scalable data structures is also highly desirable.

All of these are discussed elsewhere.

Exceptions, Signals, and Thread Cancellation

[ It is unclear to what extent this needs to or should be addressed here. I think there is agreement that thread cancellation (though not PTHREAD_CANCEL_ASYNCHRONOUS-style) and exceptions should be unified. But the details are controversial, and that seems to be more of a threads API issue.

Nick Maclaren argues that we need to say something about the state that is visible to an exception handler that was thrown to reflect a synchronous error, such as an arithmetic overflow. Since we are effectively respecifying intra-thread memory visibility, there are strong interactions with threads issues, and the presence of synchronization primitives gives us an opportunity for a meaningful specification that is at least somewhat useful to a programmer, I'm inclined to agree. What follows is an approximate restatement of one of the options he proposed.

This essentially requires that compilers treat operations that may generate exceptions as memory operations, and not move them out of critical sections etc. I would be surprised if existing implementations did so.

This may need further work, even if we go with substantially this statement. In particular, the handler kind of needs to be modelled as a new thread replacing the old one, since it can have an inconsistent view of updates performed by the original thread. But on the other hand, it potentially has access to local variables whose address was not taken, and hence can see otherwise private state of the original thread. ]

If an action A throws an intra-thread out-of-band exception, then all actions that happen-before a synchronization action that happens-before A are visible to the exception handler. Conversely, if A happens-before another synchronization action B, then no action C such that B happens-before C is visible to the exception handler.

For this purpose, there are implicit synchronization actions with both acquire and release semantics (effectively memory fences) at the beginning and end of each thread execution.

[ I'm not sure whether the preceding paragraph really buys us anything. ]