generic-library/libzmq
1
0
Fork 0
mirror of https://github.com/zeromq/libzmq.git synced 2024-12-12 18:40:27 +01:00
Code Issues Projects Releases Wiki Activity

signaler transports commands per se rather than one-bit signals

Browse Source
This commit is contained in:
Martin Sustrik 2010-05-04 10:22:16 +02:00
parent 8b9bd05726
commit 235ed3a3dc
10 changed files with 177 additions and 332 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
src/app_thread.cpp
View File

@ -82,9 +82,12 @@ zmq::signaler_t *zmq::app_thread_t::get_signaler ()
bool zmq::app_thread_t::process_commands (bool block_, bool throttle_)
{
uint32_t signal;
if (block_)
signal = signaler.poll ();
bool received;
command_t cmd;
if (block_) {
received = signaler.recv (&cmd, true);
zmq_assert (received);
}
else {
#if defined ZMQ_DELAY_COMMANDS
@ -117,15 +120,12 @@ bool zmq::app_thread_t::process_commands (bool block_, bool throttle_)
#endif
// Check whether there are any commands pending for this thread.
signal = signaler.check ();
received = signaler.recv (&cmd, false);
}
// Process all the commands from the signaling source if there is one.
if (signal != signaler_t::no_signal) {
command_t cmd;
while (get_dispatcher ()->read (signal, get_thread_slot (), &cmd))
cmd.destination->process_command (cmd);
}
// Process the command, if any.
if (received)
cmd.destination->process_command (cmd);
return !terminated;
}

5
src/config.hpp
View File

@ -32,11 +32,6 @@ namespace zmq
// memory allocation by approximately 99.6%
message_pipe_granularity = 256,
// Number of new commands in command pipe needed to trigger new memory
// allocation. The number should be kept low to decrease the memory
// footprint of dispatcher.
command_pipe_granularity = 4,
// Number of signals that can be read by the signaler
// using a single system call.
signal_buffer_size = 8,

71
src/dispatcher.cpp
View File

@ -18,6 +18,7 @@
*/
#include <new>
#include <string.h>
#include "../include/zmq.h"
@ -48,36 +49,31 @@ zmq::dispatcher_t::dispatcher_t (uint32_t app_threads_, uint32_t io_threads_) :
HIBYTE (wsa_data.wVersion) == 2);
#endif
// Initialise the array of signalers.
signalers_count = app_threads_ + io_threads_;
signalers = (signaler_t**) malloc (sizeof (signaler_t*) * signalers_count);
zmq_assert (signalers);
memset (signalers, 0, sizeof (signaler_t*) * signalers_count);
// Create I/O thread objects.
for (uint32_t i = 0; i != io_threads_; i++) {
io_thread_t *io_thread = new (std::nothrow) io_thread_t (this, i);
zmq_assert (io_thread);
io_threads.push_back (io_thread);
signalers [i] = io_thread->get_signaler ();
}
// Create application thread proxies.
for (uint32_t i = 0; i != app_threads_; i++) {
app_thread_info_t info;
info.associated = false;
info.app_thread = new (std::nothrow) app_thread_t (this, i);
info.app_thread = new (std::nothrow) app_thread_t (this,
i + io_threads_);
zmq_assert (info.app_thread);
app_threads.push_back (info);
signalers.push_back (info.app_thread->get_signaler ());
signalers [i + io_threads_] = info.app_thread->get_signaler ();
}
// Create I/O thread objects.
for (uint32_t i = 0; i != io_threads_; i++) {
io_thread_t *io_thread = new (std::nothrow) io_thread_t (this,
i + app_threads_);
zmq_assert (io_thread);
io_threads.push_back (io_thread);
signalers.push_back (io_thread->get_signaler ());
}
// Create the administrative thread. Nothing special is needed. NULL
// is used instead of signaler given that as for now, administrative
// thread doesn't receive any commands. The only thing it is used for
// is sending 'stop' command to I/O threads on shutdown.
signalers.push_back (NULL);
// Create command pipe matrix.
command_pipes = new (std::nothrow) command_pipe_t [signalers.size () *
signalers.size ()];
zmq_assert (command_pipes);
// Launch I/O threads.
for (uint32_t i = 0; i != io_threads_; i++)
io_threads [i]->start ();
@ -123,12 +119,11 @@ zmq::dispatcher_t::~dispatcher_t ()
while (!pipes.empty ())
delete *pipes.begin ();
// TODO: Deallocate any commands still in the pipes. Keep in mind that
// simple reading from a pipe and deallocating commands won't do as
// command pipe has template parameter D set to true, meaning that
// read may return false even if there are still commands in the pipe.
delete [] command_pipes;
// Deallocate the array of pointers to signalers. No special work is
// needed as signalers themselves were deallocated with their
// corresponding (app_/io_) thread objects.
free (signalers);
#ifdef ZMQ_HAVE_WINDOWS
// On Windows, uninitialise socket layer.
int rc = WSACleanup ();
@ -136,11 +131,6 @@ zmq::dispatcher_t::~dispatcher_t ()
#endif
}
uint32_t zmq::dispatcher_t::thread_slot_count ()
{
return (uint32_t) signalers.size ();
}
zmq::socket_base_t *zmq::dispatcher_t::create_socket (int type_)
{
app_threads_sync.lock ();
@ -213,21 +203,16 @@ void zmq::dispatcher_t::no_sockets (app_thread_t *thread_)
app_threads_sync.unlock ();
}
void zmq::dispatcher_t::write (uint32_t source_, uint32_t destination_,
void zmq::dispatcher_t::send_command (uint32_t destination_,
const command_t &command_)
{
command_pipe_t &pipe =
command_pipes [source_ * signalers.size () + destination_];
pipe.write (command_);
if (!pipe.flush ())
signalers [destination_]->signal (source_);
signalers [destination_]->send (command_);
}
bool zmq::dispatcher_t::read (uint32_t source_, uint32_t destination_,
command_t *command_)
bool zmq::dispatcher_t::recv_command (uint32_t thread_slot_,
command_t *command_, bool block_)
{
return command_pipes [source_ * signalers.size () +
destination_].read (command_);
return signalers [thread_slot_]->recv (command_, block_);
}
zmq::io_thread_t *zmq::dispatcher_t::choose_io_thread (uint64_t affinity_)

30
src/dispatcher.hpp
View File

@ -27,7 +27,6 @@
#include "signaler.hpp"
#include "ypipe.hpp"
#include "command.hpp"
#include "config.hpp"
#include "mutex.hpp"
#include "stdint.hpp"
@ -69,19 +68,12 @@ namespace zmq
// should disassociate the object from the current OS thread.
void no_sockets (class app_thread_t *thread_);
// Returns number of thread slots in the dispatcher. To be used by
// individual threads to find out how many distinct signals can be
// received.
uint32_t thread_slot_count ();
// Send command to the destination thread.
void send_command (uint32_t destination_, const command_t &command_);
// Send command from the source to the destination.
void write (uint32_t source_, uint32_t destination_,
const command_t &command_);
// Receive command from the source. Returns false if there is no
// command available.
bool read (uint32_t source_, uint32_t destination_,
command_t *command_);
// Receive command from another thread.
bool recv_command (uint32_t thread_slot_, command_t *command_,
bool block_);
// Returns the I/O thread that is the least busy at the moment.
// Taskset specifies which I/O threads are eligible (0 = all).
@ -126,15 +118,9 @@ namespace zmq
typedef std::vector <class io_thread_t*> io_threads_t;
io_threads_t io_threads;
// Signalers for both application and I/O threads.
std::vector <signaler_t*> signalers;
// Pipe to hold the commands.
typedef ypipe_t <command_t, true,
command_pipe_granularity> command_pipe_t;
// NxN matrix of command pipes.
command_pipe_t *command_pipes;
// Array of pointers to signalers for both application and I/O threads.
int signalers_count;
signaler_t **signalers;
// As pipes may reside in orphaned state in particular moments
// of the pipe shutdown process, i.e. neither pipe reader nor

17
src/io_thread.cpp
View File

@ -22,10 +22,8 @@
#include "../include/zmq.h"
#include "io_thread.hpp"
#include "command.hpp"
#include "platform.hpp"
#include "err.hpp"
#include "command.hpp"
#include "dispatcher.hpp"
zmq::io_thread_t::io_thread_t (dispatcher_t *dispatcher_,
@ -67,17 +65,18 @@ int zmq::io_thread_t::get_load ()
void zmq::io_thread_t::in_event ()
{
// TODO: Do we want to limit number of commands I/O thread can
// process in a single go?
while (true) {
// Get the next signal.
uint32_t signal = signaler.check ();
if (signal == signaler_t::no_signal)
// Get the next command. If there is none, exit.
command_t cmd;
if (!signaler.recv (&cmd, false))
break;
// Process all the commands from the thread that sent the signal.
command_t cmd;
while (get_dispatcher ()->read (signal, get_thread_slot (), &cmd))
cmd.destination->process_command (cmd);
// Process the command.
cmd.destination->process_command (cmd);
}
}

6
src/object.cpp
View File

@ -157,11 +157,10 @@ void zmq::object_t::send_stop ()
{
// 'stop' command goes always from administrative thread to
// the current object.
uint32_t admin_thread_id = dispatcher->thread_slot_count () - 1;
command_t cmd;
cmd.destination = this;
cmd.type = command_t::stop;
dispatcher->write (admin_thread_id, thread_slot, cmd);
dispatcher->send_command (thread_slot, cmd);
}
void zmq::object_t::send_plug (owned_t *destination_, bool inc_seqnum_)
@ -370,7 +369,6 @@ void zmq::object_t::process_seqnum ()
void zmq::object_t::send_command (command_t &cmd_)
{
uint32_t destination_thread_slot = cmd_.destination->get_thread_slot ();
dispatcher->write (thread_slot, destination_thread_slot, cmd_);
dispatcher->send_command (cmd_.destination->get_thread_slot (), cmd_);
}

2
src/pipe.hpp
View File

@ -145,7 +145,7 @@ namespace zmq
};
// Message pipe.
class pipe_t : public ypipe_t <zmq_msg_t, false, message_pipe_granularity>
class pipe_t : public ypipe_t <zmq_msg_t, message_pipe_granularity>
{
public:

234
src/signaler.cpp
View File

@ -30,52 +30,9 @@
#else
#include <unistd.h>
#include <fcntl.h>
#include <limits.h>
#endif
const uint32_t zmq::signaler_t::no_signal = 0xffffffff;
uint32_t zmq::signaler_t::poll ()
{
// Return next signal.
if (current != count) {
uint32_t result = buffer [current];
current++;
return result;
}
// If there is no signal buffered, poll for new signals.
xpoll ();
// Return first signal.
zmq_assert (current != count);
uint32_t result = buffer [current];
current++;
return result;
}
uint32_t zmq::signaler_t::check ()
{
// Return next signal.
if (current != count) {
uint32_t result = buffer [current];
current++;
return result;
}
// If there is no signal buffered, check whether more signals
// can be obtained.
xcheck ();
// Return first signal if any.
if (current != count) {
uint32_t result = buffer [current];
current++;
return result;
}
return no_signal;
}
zmq::fd_t zmq::signaler_t::get_fd ()
{
return r;
@ -84,8 +41,6 @@ zmq::fd_t zmq::signaler_t::get_fd ()
#if defined ZMQ_HAVE_WINDOWS
zmq::signaler_t::signaler_t () :
current (0),
count (0)
{
// Windows have no 'socketpair' function. CreatePipe is no good as pipe
// handles cannot be polled on. Here we create the socketpair by hand.
@ -146,51 +101,49 @@ zmq::signaler_t::~signaler_t ()
wsa_assert (rc != SOCKET_ERROR);
}
void zmq::signaler_t::signal (uint32_t signal_)
void zmq::signaler_t::send (const command_t &cmd_)
{
// TODO: Note that send is a blocking operation.
// How should we behave if the signal cannot be written to the signaler?
int rc = send (w, (char*) &signal_, sizeof (signal_), 0);
// Even worse: What if half of a command is written?
int rc = send (w, (char*) &cmd_, sizeof (command_t), 0);
win_assert (rc != SOCKET_ERROR);
zmq_assert (rc == sizeof (signal_));
zmq_assert (rc == sizeof (command_t));
}
void zmq::signaler_t::xpoll ()
bool zmq::signaler_t::recv (command_t *cmd_, bool block_)
{
// Switch to blocking mode.
unsigned long argp = 0;
int rc = ioctlsocket (r, FIONBIO, &argp);
wsa_assert (rc != SOCKET_ERROR);
if (block_) {
// Get the signals. Given that we are in the blocking mode now,
// there should be at least a single signal returned.
xcheck ();
zmq_assert (current != count);
// Switch back to non-blocking mode.
argp = 1;
rc = ioctlsocket (r, FIONBIO, &argp);
wsa_assert (rc != SOCKET_ERROR);
}
void zmq::signaler_t::xcheck ()
{
int nbytes = recv (r, (char*) buffer, sizeof (buffer), 0);
// No signals are available.
if (nbytes == -1 && WSAGetLastError () == WSAEWOULDBLOCK) {
current = 0;
count = 0;
return;
// Switch to blocking mode.
unsigned long argp = 0;
int rc = ioctlsocket (r, FIONBIO, &argp);
wsa_assert (rc != SOCKET_ERROR);
}
wsa_assert (nbytes != -1);
bool result;
int nbytes = recv (r, (char*) cmd_, sizeof (command_t), 0);
if (nbytes == -1 && WSAGetLastError () == WSAEWOULDBLOCK) {
result = false;
}
else {
wsa_assert (nbytes != -1);
// Check whether we haven't got half of a signal.
zmq_assert (nbytes % sizeof (uint32_t) == 0);
// Check whether we haven't got half of a signal.
zmq_assert (nbytes % sizeof (uint32_t) == 0);
current = 0;
count = nbytes / sizeof (uint32_t);
result = true;
}
if (block_) {
// Switch back to non-blocking mode.
unsigned long argp = 1;
int rc = ioctlsocket (r, FIONBIO, &argp);
wsa_assert (rc != SOCKET_ERROR);
}
return result;
}
#elif defined ZMQ_HAVE_HPUX || defined ZMQ_HAVE_AIX
@ -198,9 +151,7 @@ void zmq::signaler_t::xcheck ()
#include <sys/types.h>
#include <sys/socket.h>
zmq::signaler_t::signaler_t () :
current (0),
count (0)
zmq::signaler_t::signaler_t ()
{
int sv [2];
int rc = socketpair (AF_UNIX, SOCK_STREAM, 0, sv);
@ -222,49 +173,50 @@ zmq::signaler_t::~signaler_t ()
close (r);
}
void zmq::signaler_t::signal (uint32_t signal_)
void zmq::signaler_t::send (const command_t &cmd_)
{
ssize_t nbytes = send (w, &signal_, sizeof (signal_), 0);
ssize_t nbytes = send (w, &cmd_, sizeof (command_t), 0);
errno_assert (nbytes != -1);
zmq_assert (nbytes == sizeof (signal_);
zmq_assert (nbytes == sizeof (command_t));
}
void zmq::signaler_t::xpoll ()
bool zmq::signaler_t::recv (command_t &cmd_, bool block_)
{
// Set the reader to blocking mode.
int flags = fcntl (r, F_GETFL, 0);
if (flags == -1)
flags = 0;
int rc = fcntl (r, F_SETFL, flags & ~O_NONBLOCK);
errno_assert (rc != -1);
if (block_) {
// Poll for events.
xcheck ();
zmq_assert (current != count);
// Set the reader to non-blocking mode.
flags = fcntl (r, F_GETFL, 0);
if (flags == -1)
flags = 0;
rc = fcntl (r, F_SETFL, flags | O_NONBLOCK);
errno_assert (rc != -1);
}
void zmq::signaler_t::xcheck ()
{
ssize_t nbytes = recv (r, buffer, sizeof (buffer), 0);
if (nbytes == -1 && errno == EAGAIN) {
current = 0;
count = 0;
return;
// Set the reader to blocking mode.
int flags = fcntl (r, F_GETFL, 0);
if (flags == -1)
flags = 0;
int rc = fcntl (r, F_SETFL, flags & ~O_NONBLOCK);
errno_assert (rc != -1);
}
zmq_assert (nbytes != -1);
// Check whether we haven't got half of a signal.
zmq_assert (nbytes % sizeof (uint32_t) == 0);
bool result;
ssize_t nbytes = recv (r, buffer, sizeof (command_t), 0);
if (nbytes == -1 && errno == EAGAIN) {
result = false;
}
else {
zmq_assert (nbytes != -1);
current = 0;
count = nbytes / sizeof (uint32_t);
// Check whether we haven't got half of command.
zmq_assert (nbytes == sizeof (command_t));
result = true;
}
if (block_)
// Set the reader to non-blocking mode.
int flags = fcntl (r, F_GETFL, 0);
if (flags == -1)
flags = 0;
int rc = fcntl (r, F_SETFL, flags | O_NONBLOCK);
errno_assert (rc != -1);
}
return result;
}
#else
@ -272,10 +224,13 @@ void zmq::signaler_t::xcheck ()
#include <sys/types.h>
#include <sys/socket.h>
zmq::signaler_t::signaler_t () :
current (0),
count (0)
zmq::signaler_t::signaler_t ()
{
// Make sure that command can be written to the socket in atomic fashion.
// If this wasn't guaranteed, commands from different threads would be
// interleaved.
zmq_assert (sizeof (command_t) <= PIPE_BUF);
int sv [2];
int rc = socketpair (AF_UNIX, SOCK_STREAM, 0, sv);
errno_assert (rc == 0);
@ -289,42 +244,33 @@ zmq::signaler_t::~signaler_t ()
close (r);
}
void zmq::signaler_t::signal (uint32_t signal_)
void zmq::signaler_t::send (const command_t &cmd_)
{
// TODO: Note that send is a blocking operation.
// How should we behave if the signal cannot be written to the signaler?
ssize_t nbytes = send (w, &signal_, sizeof (signal_), 0);
// How should we behave if the command cannot be written to the signaler?
ssize_t nbytes = ::send (w, &cmd_, sizeof (command_t), 0);
errno_assert (nbytes != -1);
zmq_assert (nbytes == sizeof (signal_));
// This should never happen as we've already checked that command size is
// less than PIPE_BUF.
zmq_assert (nbytes == sizeof (command_t));
}
void zmq::signaler_t::xpoll ()
bool zmq::signaler_t::recv (command_t *cmd_, bool block_)
{
ssize_t nbytes = recv (r, buffer, sizeof (buffer), 0);
errno_assert (nbytes != -1);
// Check whether we haven't got half of a signal.
zmq_assert (nbytes % sizeof (uint32_t) == 0);
ssize_t nbytes = ::recv (r, cmd_, sizeof (command_t),
block_ ? 0 : MSG_DONTWAIT);
current = 0;
count = nbytes / sizeof (uint32_t);
}
// If there's no signal available return false.
if (nbytes == -1 && errno == EAGAIN)
return false;
void zmq::signaler_t::xcheck ()
{
ssize_t nbytes = recv (r, buffer, 64, MSG_DONTWAIT);
if (nbytes == -1 && errno == EAGAIN) {
current = 0;
count = 0;
return;
}
errno_assert (nbytes != -1);
// Check whether we haven't got half of a signal.
zmq_assert (nbytes % sizeof (uint32_t) == 0);
// Check whether we haven't got half of command.
zmq_assert (nbytes == sizeof (command_t));
current = 0;
count = nbytes / sizeof (uint32_t);
return true;
}
#endif

45
src/signaler.hpp
View File

@ -26,15 +26,11 @@
#include "fd.hpp"
#include "stdint.hpp"
#include "config.hpp"
#include "command.hpp"
namespace zmq
{
// This object can be used to send individual signals from one thread to
// another. The specific of this pipe is that it has associated file
// descriptor and so it can be polled on. Same signal cannot be sent twice
// unless signals are retrieved by the reader side in the meantime.
class signaler_t
{
public:
@ -42,44 +38,29 @@ namespace zmq
signaler_t ();
~signaler_t ();
static const uint32_t no_signal;
void signal (uint32_t signal_);
uint32_t poll ();
uint32_t check ();
fd_t get_fd ();
void send (const command_t &cmd_);
bool recv (command_t *cmd_, bool block_);
private:
void xpoll ();
void xcheck ();
#if defined ZMQ_HAVE_OPENVMS
// Whilst OpenVMS supports socketpair - it maps to AF_INET only.
// Further, it does not set the socket options TCP_NODELAY and
// TCP_NODELACK which can lead to performance problems. We'll
// overload the socketpair function for this class.
//
// The bug will be fixed in V5.6 ECO4 and beyond. In the
// meantime, we'll create the socket pair manually.
static int socketpair (int domain_, int type_, int protocol_,
int sv_ [2]);
// Whilst OpenVMS supports socketpair - it maps to AF_INET only.
// Further, it does not set the socket options TCP_NODELAY and
// TCP_NODELACK which can lead to performance problems. We'll
// overload the socketpair function for this class.
//
// The bug will be fixed in V5.6 ECO4 and beyond. In the
// meantime, we'll create the socket pair manually.
static int socketpair (int domain_, int type_, int protocol_,
int sv_ [2]);
#endif
// Write & read end of the socketpair.
fd_t w;
fd_t r;
// Signal buffer.
uint32_t buffer [signal_buffer_size];
// Position of the next signal in the buffer to return to the user.
size_t current;
// Number of signals in the signal buffer.
size_t count;
// Disable copying of fd_signeler object.
signaler_t (const signaler_t&);
void operator = (const signaler_t&);

79
src/ypipe.hpp
View File

@ -30,31 +30,24 @@ namespace zmq
// Lock-free queue implementation.
// Only a single thread can read from the pipe at any specific moment.
// Only a single thread can write to the pipe at any specific moment.
//
// T is the type of the object in the queue.
// If the template parameter D is set to true, it is quaranteed that
// the pipe will die in a finite time (so that you can swich to some
// other task). If D is set to false, reading from the pipe may result
// in an infinite cycle (if the pipe is continuosly fed by new elements).
// N is granularity of the pipe (how many elements have to be inserted
// till actual memory allocation is required).
// N is granularity of the pipe, i.e. how many messages are needed to
// perform next memory allocation.
template <typename T, bool D, int N> class ypipe_t
template <typename T, int N> class ypipe_t
{
public:
// Initialises the pipe. In D scenario it is created in dead state.
// Otherwise it's alive.
inline ypipe_t () :
stop (false)
// Initialises the pipe.
inline ypipe_t ()
{
// Insert terminator element into the queue.
queue.push ();
// Let all the pointers to point to the terminator
// Let all the pointers to point to the terminator.
// (unless pipe is dead, in which case c is set to NULL).
r = w = &queue.back ();
c.set (D ? NULL : &queue.back ());
c.set (&queue.back ());
}
// Following function (write) deliberately copies uninitialised data
@ -125,50 +118,17 @@ namespace zmq
return true;
// There's no prefetched value, so let us prefetch more values.
// (Note that D is a template parameter. Becaue of that one of
// the following branches will be completely optimised away
// by the compiler.)
if (D) {
// Prefetching is to simply retrieve the
// pointer from c in atomic fashion. If there are no
// items to prefetch, set c to NULL (using compare-and-swap).
r = c.cas (&queue.front (), NULL);
// If one prefetch was already done since last sleeping,
// don't do a new one, rather ask caller to go asleep.
if (stop) {
stop = false;
return false;
}
// Get new items. Perform the operation in atomic fashion.
r = c.xchg (NULL);
// If there are no elements prefetched, exit and go asleep.
// During pipe's lifetime r should never be NULL, however,
// during pipe shutdown when retrieving messages from it
// to deallocate them, this can happen.
if (&queue.front () == r || !r) {
stop = false;
return false;
}
// We want to do only a single prefetch in D scenario
// before going asleep. Thus, we set stop variable to true
// so that we can return false next time the prefetch is
// attempted.
stop = true;
}
else {
// Prefetching in non-D scenario is to simply retrieve the
// pointer from c in atomic fashion. If there are no
// items to prefetch, set c to NULL (using compare-and-swap).
r = c.cas (&queue.front (), NULL);
// If there are no elements prefetched, exit.
// During pipe's lifetime r should never be NULL, however,
// it can happen during pipe shutdown when messages
// are being deallocated.
if (&queue.front () == r || !r)
return false;
}
// If there are no elements prefetched, exit.
// During pipe's lifetime r should never be NULL, however,
// it can happen during pipe shutdown when messages
// are being deallocated.
if (&queue.front () == r || !r)
return false;
// There was at least one value prefetched.
return true;
@ -211,11 +171,6 @@ namespace zmq
// atomic operations.
atomic_ptr_t <T> c;
// Used only if 'D' template parameter is set to true. If true,
// prefetch was already done since last sleeping and the reader
// should go asleep instead of prefetching once more.
bool stop;
// Disable copying of ypipe object.
ypipe_t (const ypipe_t&);
void operator = (const ypipe_t&);