gem5/src/mem/noncoherent_xbar.cc
Andreas Sandberg f16c0a4a90 sim: Decouple draining from the SimObject hierarchy
Draining is currently done by traversing the SimObject graph and
calling drain()/drainResume() on the SimObjects. This is not ideal
when non-SimObjects (e.g., ports) need draining since this means that
SimObjects owning those objects need to be aware of this.

This changeset moves the responsibility for finding objects that need
draining from SimObjects and the Python-side of the simulator to the
DrainManager. The DrainManager now maintains a set of all objects that
need draining. To reduce the overhead in classes owning non-SimObjects
that need draining, objects inheriting from Drainable now
automatically register with the DrainManager. If such an object is
destroyed, it is automatically unregistered. This means that drain()
and drainResume() should never be called directly on a Drainable
object.

While implementing the new functionality, the DrainManager has now
been made thread safe. In practice, this means that it takes a lock
whenever it manipulates the set of Drainable objects since SimObjects
in different threads may create Drainable objects
dynamically. Similarly, the drain counter is now an atomic_uint, which
ensures that it is manipulated correctly when objects signal that they
are done draining.

A nice side effect of these changes is that it makes the drain state
changes stricter, which the simulation scripts can exploit to avoid
redundant drains.
2015-07-07 09:51:05 +01:00

331 lines
12 KiB
C++

/*
* Copyright (c) 2011-2015 ARM Limited
* All rights reserved
*
* The license below extends only to copyright in the software and shall
* not be construed as granting a license to any other intellectual
* property including but not limited to intellectual property relating
* to a hardware implementation of the functionality of the software
* licensed hereunder. You may use the software subject to the license
* terms below provided that you ensure that this notice is replicated
* unmodified and in its entirety in all distributions of the software,
* modified or unmodified, in source code or in binary form.
*
* Copyright (c) 2006 The Regents of The University of Michigan
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met: redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer;
* 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;
* neither the name of the copyright holders 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 COPYRIGHT HOLDERS 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 COPYRIGHT
* OWNER 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.
*
* Authors: Ali Saidi
* Andreas Hansson
* William Wang
*/
/**
* @file
* Definition of a non-coherent crossbar object.
*/
#include "base/misc.hh"
#include "base/trace.hh"
#include "debug/NoncoherentXBar.hh"
#include "debug/XBar.hh"
#include "mem/noncoherent_xbar.hh"
NoncoherentXBar::NoncoherentXBar(const NoncoherentXBarParams *p)
: BaseXBar(p)
{
// create the ports based on the size of the master and slave
// vector ports, and the presence of the default port, the ports
// are enumerated starting from zero
for (int i = 0; i < p->port_master_connection_count; ++i) {
std::string portName = csprintf("%s.master[%d]", name(), i);
MasterPort* bp = new NoncoherentXBarMasterPort(portName, *this, i);
masterPorts.push_back(bp);
reqLayers.push_back(new ReqLayer(*bp, *this,
csprintf(".reqLayer%d", i)));
}
// see if we have a default slave device connected and if so add
// our corresponding master port
if (p->port_default_connection_count) {
defaultPortID = masterPorts.size();
std::string portName = name() + ".default";
MasterPort* bp = new NoncoherentXBarMasterPort(portName, *this,
defaultPortID);
masterPorts.push_back(bp);
reqLayers.push_back(new ReqLayer(*bp, *this, csprintf(".reqLayer%d",
defaultPortID)));
}
// create the slave ports, once again starting at zero
for (int i = 0; i < p->port_slave_connection_count; ++i) {
std::string portName = csprintf("%s.slave[%d]", name(), i);
QueuedSlavePort* bp = new NoncoherentXBarSlavePort(portName, *this, i);
slavePorts.push_back(bp);
respLayers.push_back(new RespLayer(*bp, *this,
csprintf(".respLayer%d", i)));
}
clearPortCache();
}
NoncoherentXBar::~NoncoherentXBar()
{
for (auto l: reqLayers)
delete l;
for (auto l: respLayers)
delete l;
}
bool
NoncoherentXBar::recvTimingReq(PacketPtr pkt, PortID slave_port_id)
{
// determine the source port based on the id
SlavePort *src_port = slavePorts[slave_port_id];
// we should never see express snoops on a non-coherent crossbar
assert(!pkt->isExpressSnoop());
// determine the destination based on the address
PortID master_port_id = findPort(pkt->getAddr());
// test if the layer should be considered occupied for the current
// port
if (!reqLayers[master_port_id]->tryTiming(src_port)) {
DPRINTF(NoncoherentXBar, "recvTimingReq: src %s %s 0x%x BUSY\n",
src_port->name(), pkt->cmdString(), pkt->getAddr());
return false;
}
DPRINTF(NoncoherentXBar, "recvTimingReq: src %s %s 0x%x\n",
src_port->name(), pkt->cmdString(), pkt->getAddr());
// store size and command as they might be modified when
// forwarding the packet
unsigned int pkt_size = pkt->hasData() ? pkt->getSize() : 0;
unsigned int pkt_cmd = pkt->cmdToIndex();
// store the old header delay so we can restore it if needed
Tick old_header_delay = pkt->headerDelay;
// a request sees the frontend and forward latency
Tick xbar_delay = (frontendLatency + forwardLatency) * clockPeriod();
// set the packet header and payload delay
calcPacketTiming(pkt, xbar_delay);
// determine how long to be crossbar layer is busy
Tick packetFinishTime = clockEdge(Cycles(1)) + pkt->payloadDelay;
// before forwarding the packet (and possibly altering it),
// remember if we are expecting a response
const bool expect_response = pkt->needsResponse() &&
!pkt->memInhibitAsserted();
// since it is a normal request, attempt to send the packet
bool success = masterPorts[master_port_id]->sendTimingReq(pkt);
if (!success) {
// inhibited packets should never be forced to retry
assert(!pkt->memInhibitAsserted());
DPRINTF(NoncoherentXBar, "recvTimingReq: src %s %s 0x%x RETRY\n",
src_port->name(), pkt->cmdString(), pkt->getAddr());
// restore the header delay as it is additive
pkt->headerDelay = old_header_delay;
// occupy until the header is sent
reqLayers[master_port_id]->failedTiming(src_port,
clockEdge(Cycles(1)));
return false;
}
// remember where to route the response to
if (expect_response) {
assert(routeTo.find(pkt->req) == routeTo.end());
routeTo[pkt->req] = slave_port_id;
}
reqLayers[master_port_id]->succeededTiming(packetFinishTime);
// stats updates
pktCount[slave_port_id][master_port_id]++;
pktSize[slave_port_id][master_port_id] += pkt_size;
transDist[pkt_cmd]++;
return true;
}
bool
NoncoherentXBar::recvTimingResp(PacketPtr pkt, PortID master_port_id)
{
// determine the source port based on the id
MasterPort *src_port = masterPorts[master_port_id];
// determine the destination
const auto route_lookup = routeTo.find(pkt->req);
assert(route_lookup != routeTo.end());
const PortID slave_port_id = route_lookup->second;
assert(slave_port_id != InvalidPortID);
assert(slave_port_id < respLayers.size());
// test if the layer should be considered occupied for the current
// port
if (!respLayers[slave_port_id]->tryTiming(src_port)) {
DPRINTF(NoncoherentXBar, "recvTimingResp: src %s %s 0x%x BUSY\n",
src_port->name(), pkt->cmdString(), pkt->getAddr());
return false;
}
DPRINTF(NoncoherentXBar, "recvTimingResp: src %s %s 0x%x\n",
src_port->name(), pkt->cmdString(), pkt->getAddr());
// store size and command as they might be modified when
// forwarding the packet
unsigned int pkt_size = pkt->hasData() ? pkt->getSize() : 0;
unsigned int pkt_cmd = pkt->cmdToIndex();
// a response sees the response latency
Tick xbar_delay = responseLatency * clockPeriod();
// set the packet header and payload delay
calcPacketTiming(pkt, xbar_delay);
// determine how long to be crossbar layer is busy
Tick packetFinishTime = clockEdge(Cycles(1)) + pkt->payloadDelay;
// send the packet through the destination slave port, and pay for
// any outstanding latency
Tick latency = pkt->headerDelay;
pkt->headerDelay = 0;
slavePorts[slave_port_id]->schedTimingResp(pkt, curTick() + latency);
// remove the request from the routing table
routeTo.erase(route_lookup);
respLayers[slave_port_id]->succeededTiming(packetFinishTime);
// stats updates
pktCount[slave_port_id][master_port_id]++;
pktSize[slave_port_id][master_port_id] += pkt_size;
transDist[pkt_cmd]++;
return true;
}
void
NoncoherentXBar::recvReqRetry(PortID master_port_id)
{
// responses never block on forwarding them, so the retry will
// always be coming from a port to which we tried to forward a
// request
reqLayers[master_port_id]->recvRetry();
}
Tick
NoncoherentXBar::recvAtomic(PacketPtr pkt, PortID slave_port_id)
{
DPRINTF(NoncoherentXBar, "recvAtomic: packet src %s addr 0x%x cmd %s\n",
slavePorts[slave_port_id]->name(), pkt->getAddr(),
pkt->cmdString());
unsigned int pkt_size = pkt->hasData() ? pkt->getSize() : 0;
unsigned int pkt_cmd = pkt->cmdToIndex();
// determine the destination port
PortID master_port_id = findPort(pkt->getAddr());
// stats updates for the request
pktCount[slave_port_id][master_port_id]++;
pktSize[slave_port_id][master_port_id] += pkt_size;
transDist[pkt_cmd]++;
// forward the request to the appropriate destination
Tick response_latency = masterPorts[master_port_id]->sendAtomic(pkt);
// add the response data
if (pkt->isResponse()) {
pkt_size = pkt->hasData() ? pkt->getSize() : 0;
pkt_cmd = pkt->cmdToIndex();
// stats updates
pktCount[slave_port_id][master_port_id]++;
pktSize[slave_port_id][master_port_id] += pkt_size;
transDist[pkt_cmd]++;
}
// @todo: Not setting first-word time
pkt->payloadDelay = response_latency;
return response_latency;
}
void
NoncoherentXBar::recvFunctional(PacketPtr pkt, PortID slave_port_id)
{
if (!pkt->isPrint()) {
// don't do DPRINTFs on PrintReq as it clutters up the output
DPRINTF(NoncoherentXBar,
"recvFunctional: packet src %s addr 0x%x cmd %s\n",
slavePorts[slave_port_id]->name(), pkt->getAddr(),
pkt->cmdString());
}
// since our slave ports are queued ports we need to check them as well
for (const auto& p : slavePorts) {
// if we find a response that has the data, then the
// downstream caches/memories may be out of date, so simply stop
// here
if (p->checkFunctional(pkt)) {
if (pkt->needsResponse())
pkt->makeResponse();
return;
}
}
// determine the destination port
PortID dest_id = findPort(pkt->getAddr());
// forward the request to the appropriate destination
masterPorts[dest_id]->sendFunctional(pkt);
}
NoncoherentXBar*
NoncoherentXBarParams::create()
{
return new NoncoherentXBar(this);
}
void
NoncoherentXBar::regStats()
{
// register the stats of the base class and our layers
BaseXBar::regStats();
for (auto l: reqLayers)
l->regStats();
for (auto l: respLayers)
l->regStats();
}