Prev: Language Manual | Next: Command Line Interface

---

CIVL Libraries

In progress.

bundle

Header: bundle.cvh. Uses: op.

A bundle is a sequence of objects of the same type, wrapped into an atomic package. A bundle is created using a function that specifies a region of memory. One can create a bundle from an array of integers, and another bundle from an array of reals. Both bundles have the same type, $bundle. They can therefore be entered into an array of $bundle, for example. Hence bundles are useful for mixing objects of different (even statically unknown) types into a single data structure. Later, the contents of a bundle can be extracted with another function that specifies a region of memory into which to unpack the bundle; if that memory does not have the right type to receive the contents of the bundle, a runtime error occurs.

Types

• $bundle: the bundle type

Functions

• $system int $bundle_size($bundle b)
  • Returns the size of the given bundle, i.e., the number of bytes consumed by all elements of the bundle. This is the product of the number of elements and the size of one element.
• $system $bundle $bundle_pack(void *ptr, int size)
  • This function creates a bundle from the memory region specified by ptr and size (the total number of bytes to copy), copying the data into the new bundle. It returns the new bundle.
• $system void $bundle_unpack($bundle b, void *ptr)
  • This function copies the data from the given bundle into the memory region starting at ptr. The memory region extends for $bundle_size(b) bytes.
• $atomic_f void $bundle_unpack_apply($bundle data, void *buf, $operation op, int count, void *result)
  • This function unpacks the bundle while applying the specified numeric operation. Parameter op specifies a binary operation. For each i in 0..count-1, the operation is applied to the i-th element of buf and the i-th element of the bundle, and the result is stored in the i-th position of result. Parameter count is the number of elements in the bundle; buf and result should each point to a region of memory capable of holding at least count elements.

Example

#include <bundle.cvh>
int main() {
  int n = 10;
  int a[n], b[n], c[n];
  for (int i=0; i<n; i++) a[i] = i;
  $bundle bun = $bundle_pack(a, n*sizeof(int));
  $assert($bundle_size(bun) == n*sizeof(int));
  $bundle_unpack(bun, b);
  $assert($forall (int i: 0..n-1) b[i]==a[i]);
  $bundle_unpack_apply(bun, b, _SUM, n, c);
  $assert($forall (int i: 0..n-1) c[i]==2*a[i]);
}

comm

Header: comm.cvh. Uses: bundle.

The comm library supports message-passing-style communication. It provides a number of basic data structures which can be used to implement specific message-passing interfaces, such as MPI.

Types

• $message: A message consists of data and meta-data. The data is a finite sequence of objects, all of the same type. The meta-data consists of 3 integers: the source, destination, and tag. The source identifies the process that sent the message. The destination identifies the process to which the message is sent. Both source and destination are "places", which are nonnegative integer IDs that abstract places from which messages are sent or received. Typically, a single place is associated to each process in a message-passing system, but this is not required. The tag is an additional nonnegative integer that can be used to represent anything the user wants; the receiving process may select a message based on the tag.
• $gcomm: This is an opaque handle to an object that stores the state of the message-passing system. In particular, the buffered messages (those that have been sent, but not yet received) are stored in the gcomm.
• $comm: Each place has its own comm, an opaque handle local to that place that is used to access all message-passing functions (e.g., sends and receives) at that place. It contains a reference to the gcomm.

Constants

• $COMM_ANY_SOURCE: a constant that can be used as the source argument in a dequeue operation to indicate that a message may be taken from any source
• $COMM_ANY_TAG: a constant that can be used as the tag argument in a dequeue operation to indicate that a message with any tag may be selected
• $comm_null: a constant representing the absence of $comm

Functions

• $atomic_f $message $message_pack(int source, int dest, int tag, const void *data, int size)
  • Creates a new message with the given meta-data, and the data taken from the memory region that begins at data and extends for size bytes. The data is copied into the new message object. The source, destination, and tag must be nonnegative. The size may be 0.
• $atomic_f int $message_source($message message)
  • Gets the source of the message.
• $atomic_f int $message_tag($message message)
  • Gets the tag of the message.
• $atomic_f int $message_dest($message message)
  • Gets the destination of the message.
• $atomic_f int $message_size($message message)
  • Gets the size of the message data (in bytes).
• $atomic_f void $message_unpack($message message, void *buf, int size)
  • Extracts the data from a message, storing it in the buffer that starts at buf. The argument size must be greater than or equal to the size of message, else an error occurs.
• $atomic_f $gcomm $gcomm_create($scope scope, int size)
  • Allocates a new gcomm object that will be stored in the heap of the specified scope. The size is the number of "places" that will be associated to this gcomm. The place IDs range from 0 to size-1 (inclusive).
• $atomic_f int $gcomm_destroy($gcomm gcomm, void * junkMsgs)
  • Destroys the gcomm, deallocating memory that was allocated for it. If junkMsgs is not NULL it shall be a pointer to an array of indeterminate length of $message. This function will populate that array with all messages remaining the gcomm's buffers and return the number of such messages.
• $atomic_f void $gcomm_dup($comm comm, $comm newcomm)
  • Duplicates the communicator comm. The newcomm is initialized to have the same gcomm and place as the original comm.
• $atomic_f $comm $comm_create($scope scope, $gcomm gcomm, int place)
  • Creates a new comm associated with the specified place in gcomm. If place is less than 0, or greater than the size of gcomm, or another process has already occupied place in gcomm, an error occurs. The new comm object is allocated in the heap of the given scope.
• $atomic_f void $comm_destroy($comm comm)
  • Deallocates the comm.
• $atomic_f int $comm_size($comm comm)
  • Gets the size of the gcomm that owns comm.
• $atomic_f int $comm_place($comm comm)
  • Gets the place of the given comm in its gcomm object.
• $system void $comm_enqueue($comm comm, $message message)
  • Enqueues a message in a FIFO channel in the gcomm. The message should have its source be the place of comm. The message will enqueued on channel (i,j) in the gcomm, where i is the place of the source, and j is the place of the destination of the message.
• $system $state_f _Bool $comm_probe($comm comm, int source, int tag)
  • Determines whether there exists a message with the given source and tag, and with destination the place of comm, in the gcomm associated to comm. source may be $COMM_ANY_SOURCE and tag may be $COMM_ANY_TAG.
• $system $message $comm_seek($comm comm, int source, int tag)
  • Returns the oldest message in the gcomm associated to comm that matches source and tag without modifying gcomm. source may be $COMM_ANY_SOURCE and tag may be $COMM_ANY_TAG. If no such message found, returns a message with empty data and all meta-data being -1.
• $system $message $comm_dequeue($comm comm, int source, int tag)
  • Dequeues a message in a FIFO channel in the gcomm. The message should have its destination be the place of comm. The message will be dequeued from channel (i,j) in the gcomm, where i is the place of the source, and j is the place of the destination of the message. A call to this function will not return until a matching message can be dequeued.
• $system $state_f _Bool $comm_empty_in($comm )
  • deprecated
• $system $state_f _Bool $comm_empty_out($comm )
  • deprecated

concurrency

Header: concurrency.cvh. Uses: bundle.

Types

• $gbarrier
  • A data type representing a handle to a global barrier object.

• $barrier
  • A data type representing a process-local handle used to operate a global barrier

Functions

• $atomic_f $gbarrier $gbarrier_create($scope scope, int size)
  • Creates a new global barrier object and returns a handle to it. The barrier has the specified size. The new object will be allocated in the given scope.
• $atomic_f void $gbarrier_destroy($gbarrier gbarrier)
  • Destroys the global barrier object referred by the handle gbarrier
• int $get_nprocs($gbarrier gbarrier)
  • Returns the size of the global barrier object
• $atomic_f $barrier $barrier_create($scope scope, $gbarrier gbarrier, int place)
  • Creates a process-local handle for operating the global barrier referred by gbarrier. The place must be in [0,get_nprocs(gbarrier)-1]. It specifies the place in the global barrier that will be occupied by the local handle.
• $atomic_f void $barrier_destroy($barrier barrier)
  • Destroys the local handle to a global barrier
• void $barrier_call($barrier barrier)
  • Calls the barrier referred by barrier
• void $barrier_call_yield($barrier barrier)
  • Calls the barrier from within an atomic section: this will yield at the point where process is waiting to exit barrier.
• void $barrier_call_subset($barrier barrier, int nprocs)
  • under construction
• void $barrier_call_execute($barrier barrier, void foo(void))
  • under construction

domain

Header: domain.cvh. Uses: (nothing).

Types

• $domain_strategy
  • An enumeration type used to specify how domains are decomposed. There are three options: ALL, RANDOM, and ROUND_ROBIN.
• $domain_decomposition
  • A data structure representing the decomposition result of a domain.

Functions

• $system $domain_decomposition $domain_partition($domain domain, $domain_strategy strategy, int n)
  • Takes a domain and some n>0 and returns some partition of the domain into n sub-domains, according to the decomposition strategy specified.
• $system $range $range_of_rectangular_domain($domain dom, int index)
  • Returns the index-th range of the given domain dom. Requires index to be between 1 and the dimension of dom minus 1.
• $system int $high_of_regular_range($range range)
  • Returns the upper bound of a regular range.
• $system int $low_of_regular_range($range range)
  • Returns the lower bound of a regular range.
• $system int $step_of_regular_range($range range)
  • Returns the step of a regular range
• $system _Bool $is_rectangular_domain($domain domain)
  • Returns true if this domain is a rectangular domain.
• $system _Bool $is_regular_range($range range)
  • Returns true if this range is a regular range, i.e., it is composed of a lower bound, a higher bound and a step.
• $system int $dimension_of($domain domain)
  • Returns the dimension of the given domain.

mem

Header: mem.cvh. Uses: (nothing).

The mem library provides functions to manipulate objects of the CIVL-C built-in type---$mem. A $mem object represents a set of memory locations.

Functions

• $system void $write_set_push()
  • Pushes a new entry (i.e. an empty $mem) onto the write stack of the calling process. Each process maintains a write stack. A stack entry is a $mem object. A write to a memory location will cause the memory location being added to the top entry of the write stack, if the stack is non-empty.
• $system $mem $write_set_pop()
  • Pops and returns the top entry of the write stack of the calling process.
• $state_f $system $mem $write_set_peek()
  • Returns the top entry of the write stack of the calling process.
• $system void $read_set_push()
  • Pushes a new entry (i.e. an empty $mem) onto the read stack of the calling process. Each process maintains a read stack. A stack entry is a $mem object. Reading a memory location will cause the memory location being added to the top entry of the read stack, if the stack is non-empty.
• $system $mem $read_set_pop()
  • Pops and returns the top entry of the read stack of the calling process.
• $state_f $system $mem $read_set_peek()
  • Returns the top entry of the read stack of the calling process.
• $atomic_f $system $mem $mem_empty()
  • Creates a new empty $mem object.
• $atomic_f $system _Bool $mem_equals($mem m0, $mem m1)
  • Returns true iff m0 and m1 are the identical set of memory locations.
• $atomic_f $system _Bool $mem_contains($mem super, $mem sub)
  • Returns true iff super is a super-set of sub.
• $atomic_f $system _Bool $mem_no_intersect($mem m0, $mem m1, $mem *out0, $mem *out1)
  • Returns true iff m0 and m1 are two disjoint sets. If m0 and m1 overlap, out0 and out1 will be assigned two overlapping memory location sets, which are as small as the function can figure out from m0 and m1, respectively.
• $atomic_f $system $mem $mem_union($mem mem0, $mem mem1)
  • Returns the union of the two sets of memory locations mem0 and mem1.
• $atomic_f $system $mem $mem_union_widening($mem, $mem)
  • Similar to the $mem_union function but may return an over-approximated super-set of the union set. The over-approximation is done by the model checker in order to make state space finite.
• $atomic_f $system void $mem_havoc($mem m)
  • Assigns arbitrary values to memory locations in m
• $atomic_f $system $mem $mem_unary_widening($mem m)
  • Returns an over-approximated super-set of m. The over-approximation is done by the model checker in order to make state space finite.
• $atomic_f $system void $mem_assign_from($state s, $mem m)
  • deprecated

op

Header: op.h. Uses: (nothing).

Defines an enumerated type $operation with the following values:

Constant	Description
_NO_OP	no operation
_MAX	maximum
_MIN	minimun
_SUM	sum
_PROD	product
_LAND	logical and
_BAND	bit-wise and
_LOR	logical or
_BOR	bit-wise or
_LXOR	logical exclusive or
_BXOR	bit-wise exclusive or
_MINLOC	min value and location
_MAXLOC	max value and location
_EQ	equal to
_NEQ	not Equal to

pointer

Header: pointer.cvh. Uses: op.

scope

Header: scope.cvh. Uses: (nothing).

seq

Header: seq.cvh. Uses: (nothing).

C Standard library

stdlib

stdio

string

---

Prev: Language Manual | Next: Command Line Interface