blob: a33102a5dcf2e745a75e091e1788c3d0557c1b01 [file] [log] [blame]
/*
* Copyright (c) 2020 Advanced Micro Devices, Inc.
* Copyright (c) 2020 Inria
* Copyright (c) 2016 Georgia Institute of Technology
* Copyright (c) 2008 Princeton University
* 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.
*/
#include "mem/ruby/network/garnet/NetworkInterface.hh"
#include <cassert>
#include <cmath>
#include "base/cast.hh"
#include "debug/RubyNetwork.hh"
#include "mem/ruby/network/MessageBuffer.hh"
#include "mem/ruby/network/garnet/Credit.hh"
#include "mem/ruby/network/garnet/flitBuffer.hh"
#include "mem/ruby/slicc_interface/Message.hh"
namespace gem5
{
namespace ruby
{
namespace garnet
{
NetworkInterface::NetworkInterface(const Params &p)
: ClockedObject(p), Consumer(this), m_id(p.id),
m_virtual_networks(p.virt_nets), m_vc_per_vnet(0),
m_vc_allocator(m_virtual_networks, 0),
m_deadlock_threshold(p.garnet_deadlock_threshold),
vc_busy_counter(m_virtual_networks, 0)
{
m_stall_count.resize(m_virtual_networks);
niOutVcs.resize(0);
}
void
NetworkInterface::addInPort(NetworkLink *in_link,
CreditLink *credit_link)
{
InputPort *newInPort = new InputPort(in_link, credit_link);
inPorts.push_back(newInPort);
DPRINTF(RubyNetwork, "Adding input port:%s with vnets %s\n",
in_link->name(), newInPort->printVnets());
in_link->setLinkConsumer(this);
credit_link->setSourceQueue(newInPort->outCreditQueue(), this);
if (m_vc_per_vnet != 0) {
in_link->setVcsPerVnet(m_vc_per_vnet);
credit_link->setVcsPerVnet(m_vc_per_vnet);
}
}
void
NetworkInterface::addOutPort(NetworkLink *out_link,
CreditLink *credit_link,
SwitchID router_id, uint32_t consumerVcs)
{
OutputPort *newOutPort = new OutputPort(out_link, credit_link, router_id);
outPorts.push_back(newOutPort);
assert(consumerVcs > 0);
// We are not allowing different physical links to have different vcs
// If it is required that the Network Interface support different VCs
// for every physical link connected to it. Then they need to change
// the logic within outport and inport.
if (niOutVcs.size() == 0) {
m_vc_per_vnet = consumerVcs;
int m_num_vcs = consumerVcs * m_virtual_networks;
niOutVcs.resize(m_num_vcs);
outVcState.reserve(m_num_vcs);
m_ni_out_vcs_enqueue_time.resize(m_num_vcs);
// instantiating the NI flit buffers
for (int i = 0; i < m_num_vcs; i++) {
m_ni_out_vcs_enqueue_time[i] = Tick(INFINITE_);
outVcState.emplace_back(i, m_net_ptr, consumerVcs);
}
// Reset VC Per VNET for input links already instantiated
for (auto &iPort: inPorts) {
NetworkLink *inNetLink = iPort->inNetLink();
inNetLink->setVcsPerVnet(m_vc_per_vnet);
credit_link->setVcsPerVnet(m_vc_per_vnet);
}
} else {
fatal_if(consumerVcs != m_vc_per_vnet,
"%s: Connected Physical links have different vc requests: %d and %d\n",
name(), consumerVcs, m_vc_per_vnet);
}
DPRINTF(RubyNetwork, "OutputPort:%s Vnet: %s\n",
out_link->name(), newOutPort->printVnets());
out_link->setSourceQueue(newOutPort->outFlitQueue(), this);
out_link->setVcsPerVnet(m_vc_per_vnet);
credit_link->setLinkConsumer(this);
credit_link->setVcsPerVnet(m_vc_per_vnet);
}
void
NetworkInterface::addNode(std::vector<MessageBuffer *>& in,
std::vector<MessageBuffer *>& out)
{
inNode_ptr = in;
outNode_ptr = out;
for (auto& it : in) {
if (it != nullptr) {
it->setConsumer(this);
}
}
}
void
NetworkInterface::dequeueCallback()
{
// An output MessageBuffer has dequeued something this cycle and there
// is now space to enqueue a stalled message. However, we cannot wake
// on the same cycle as the dequeue. Schedule a wake at the soonest
// possible time (next cycle).
scheduleEventAbsolute(clockEdge(Cycles(1)));
}
void
NetworkInterface::incrementStats(flit *t_flit)
{
int vnet = t_flit->get_vnet();
// Latency
m_net_ptr->increment_received_flits(vnet);
Tick network_delay =
t_flit->get_dequeue_time() -
t_flit->get_enqueue_time() - cyclesToTicks(Cycles(1));
Tick src_queueing_delay = t_flit->get_src_delay();
Tick dest_queueing_delay = (curTick() - t_flit->get_dequeue_time());
Tick queueing_delay = src_queueing_delay + dest_queueing_delay;
m_net_ptr->increment_flit_network_latency(network_delay, vnet);
m_net_ptr->increment_flit_queueing_latency(queueing_delay, vnet);
if (t_flit->get_type() == TAIL_ || t_flit->get_type() == HEAD_TAIL_) {
m_net_ptr->increment_received_packets(vnet);
m_net_ptr->increment_packet_network_latency(network_delay, vnet);
m_net_ptr->increment_packet_queueing_latency(queueing_delay, vnet);
}
// Hops
m_net_ptr->increment_total_hops(t_flit->get_route().hops_traversed);
}
/*
* The NI wakeup checks whether there are any ready messages in the protocol
* buffer. If yes, it picks that up, flitisizes it into a number of flits and
* puts it into an output buffer and schedules the output link. On a wakeup
* it also checks whether there are flits in the input link. If yes, it picks
* them up and if the flit is a tail, the NI inserts the corresponding message
* into the protocol buffer. It also checks for credits being sent by the
* downstream router.
*/
void
NetworkInterface::wakeup()
{
std::ostringstream oss;
for (auto &oPort: outPorts) {
oss << oPort->routerID() << "[" << oPort->printVnets() << "] ";
}
DPRINTF(RubyNetwork, "Network Interface %d connected to router:%s "
"woke up. Period: %ld\n", m_id, oss.str(), clockPeriod());
assert(curTick() == clockEdge());
MsgPtr msg_ptr;
Tick curTime = clockEdge();
// Checking for messages coming from the protocol
// can pick up a message/cycle for each virtual net
for (int vnet = 0; vnet < inNode_ptr.size(); ++vnet) {
MessageBuffer *b = inNode_ptr[vnet];
if (b == nullptr) {
continue;
}
if (b->isReady(curTime)) { // Is there a message waiting
msg_ptr = b->peekMsgPtr();
if (flitisizeMessage(msg_ptr, vnet)) {
b->dequeue(curTime);
}
}
}
scheduleOutputLink();
// Check if there are flits stalling a virtual channel. Track if a
// message is enqueued to restrict ejection to one message per cycle.
checkStallQueue();
/*********** Check the incoming flit link **********/
DPRINTF(RubyNetwork, "Number of input ports: %d\n", inPorts.size());
for (auto &iPort: inPorts) {
NetworkLink *inNetLink = iPort->inNetLink();
if (inNetLink->isReady(curTick())) {
flit *t_flit = inNetLink->consumeLink();
DPRINTF(RubyNetwork, "Recieved flit:%s\n", *t_flit);
assert(t_flit->m_width == iPort->bitWidth());
int vnet = t_flit->get_vnet();
t_flit->set_dequeue_time(curTick());
// If a tail flit is received, enqueue into the protocol buffers
// if space is available. Otherwise, exchange non-tail flits for
// credits.
if (t_flit->get_type() == TAIL_ ||
t_flit->get_type() == HEAD_TAIL_) {
if (!iPort->messageEnqueuedThisCycle &&
outNode_ptr[vnet]->areNSlotsAvailable(1, curTime)) {
// Space is available. Enqueue to protocol buffer.
outNode_ptr[vnet]->enqueue(t_flit->get_msg_ptr(), curTime,
cyclesToTicks(Cycles(1)));
// Simply send a credit back since we are not buffering
// this flit in the NI
Credit *cFlit = new Credit(t_flit->get_vc(),
true, curTick());
iPort->sendCredit(cFlit);
// Update stats and delete flit pointer
incrementStats(t_flit);
delete t_flit;
} else {
// No space available- Place tail flit in stall queue and
// set up a callback for when protocol buffer is dequeued.
// Stat update and flit pointer deletion will occur upon
// unstall.
iPort->m_stall_queue.push_back(t_flit);
m_stall_count[vnet]++;
outNode_ptr[vnet]->registerDequeueCallback([this]() {
dequeueCallback(); });
}
} else {
// Non-tail flit. Send back a credit but not VC free signal.
Credit *cFlit = new Credit(t_flit->get_vc(), false,
curTick());
// Simply send a credit back since we are not buffering
// this flit in the NI
iPort->sendCredit(cFlit);
// Update stats and delete flit pointer.
incrementStats(t_flit);
delete t_flit;
}
}
}
/****************** Check the incoming credit link *******/
for (auto &oPort: outPorts) {
CreditLink *inCreditLink = oPort->inCreditLink();
if (inCreditLink->isReady(curTick())) {
Credit *t_credit = (Credit*) inCreditLink->consumeLink();
outVcState[t_credit->get_vc()].increment_credit();
if (t_credit->is_free_signal()) {
outVcState[t_credit->get_vc()].setState(IDLE_,
curTick());
}
delete t_credit;
}
}
// It is possible to enqueue multiple outgoing credit flits if a message
// was unstalled in the same cycle as a new message arrives. In this
// case, we should schedule another wakeup to ensure the credit is sent
// back.
for (auto &iPort: inPorts) {
if (iPort->outCreditQueue()->getSize() > 0) {
DPRINTF(RubyNetwork, "Sending a credit %s via %s at %ld\n",
*(iPort->outCreditQueue()->peekTopFlit()),
iPort->outCreditLink()->name(), clockEdge(Cycles(1)));
iPort->outCreditLink()->
scheduleEventAbsolute(clockEdge(Cycles(1)));
}
}
checkReschedule();
}
void
NetworkInterface::checkStallQueue()
{
// Check all stall queues.
// There is one stall queue for each input link
for (auto &iPort: inPorts) {
iPort->messageEnqueuedThisCycle = false;
Tick curTime = clockEdge();
if (!iPort->m_stall_queue.empty()) {
for (auto stallIter = iPort->m_stall_queue.begin();
stallIter != iPort->m_stall_queue.end(); ) {
flit *stallFlit = *stallIter;
int vnet = stallFlit->get_vnet();
// If we can now eject to the protocol buffer,
// send back credits
if (outNode_ptr[vnet]->areNSlotsAvailable(1,
curTime)) {
outNode_ptr[vnet]->enqueue(stallFlit->get_msg_ptr(),
curTime, cyclesToTicks(Cycles(1)));
// Send back a credit with free signal now that the
// VC is no longer stalled.
Credit *cFlit = new Credit(stallFlit->get_vc(), true,
curTick());
iPort->sendCredit(cFlit);
// Update Stats
incrementStats(stallFlit);
// Flit can now safely be deleted and removed from stall
// queue
delete stallFlit;
iPort->m_stall_queue.erase(stallIter);
m_stall_count[vnet]--;
// If there are no more stalled messages for this vnet, the
// callback on it's MessageBuffer is not needed.
if (m_stall_count[vnet] == 0)
outNode_ptr[vnet]->unregisterDequeueCallback();
iPort->messageEnqueuedThisCycle = true;
break;
} else {
++stallIter;
}
}
}
}
}
// Embed the protocol message into flits
bool
NetworkInterface::flitisizeMessage(MsgPtr msg_ptr, int vnet)
{
Message *net_msg_ptr = msg_ptr.get();
NetDest net_msg_dest = net_msg_ptr->getDestination();
// gets all the destinations associated with this message.
std::vector<NodeID> dest_nodes = net_msg_dest.getAllDest();
// Number of flits is dependent on the link bandwidth available.
// This is expressed in terms of bytes/cycle or the flit size
OutputPort *oPort = getOutportForVnet(vnet);
assert(oPort);
int num_flits = (int)divCeil((float) m_net_ptr->MessageSizeType_to_int(
net_msg_ptr->getMessageSize()), (float)oPort->bitWidth());
DPRINTF(RubyNetwork, "Message Size:%d vnet:%d bitWidth:%d\n",
m_net_ptr->MessageSizeType_to_int(net_msg_ptr->getMessageSize()),
vnet, oPort->bitWidth());
// loop to convert all multicast messages into unicast messages
for (int ctr = 0; ctr < dest_nodes.size(); ctr++) {
// this will return a free output virtual channel
int vc = calculateVC(vnet);
if (vc == -1) {
return false ;
}
MsgPtr new_msg_ptr = msg_ptr->clone();
NodeID destID = dest_nodes[ctr];
Message *new_net_msg_ptr = new_msg_ptr.get();
if (dest_nodes.size() > 1) {
NetDest personal_dest;
for (int m = 0; m < (int) MachineType_NUM; m++) {
if ((destID >= MachineType_base_number((MachineType) m)) &&
destID < MachineType_base_number((MachineType) (m+1))) {
// calculating the NetDest associated with this destID
personal_dest.clear();
personal_dest.add((MachineID) {(MachineType) m, (destID -
MachineType_base_number((MachineType) m))});
new_net_msg_ptr->getDestination() = personal_dest;
break;
}
}
net_msg_dest.removeNetDest(personal_dest);
// removing the destination from the original message to reflect
// that a message with this particular destination has been
// flitisized and an output vc is acquired
net_msg_ptr->getDestination().removeNetDest(personal_dest);
}
// Embed Route into the flits
// NetDest format is used by the routing table
// Custom routing algorithms just need destID
RouteInfo route;
route.vnet = vnet;
route.net_dest = new_net_msg_ptr->getDestination();
route.src_ni = m_id;
route.src_router = oPort->routerID();
route.dest_ni = destID;
route.dest_router = m_net_ptr->get_router_id(destID, vnet);
// initialize hops_traversed to -1
// so that the first router increments it to 0
route.hops_traversed = -1;
m_net_ptr->increment_injected_packets(vnet);
m_net_ptr->update_traffic_distribution(route);
for (int i = 0; i < num_flits; i++) {
m_net_ptr->increment_injected_flits(vnet);
flit *fl = new flit(i, vc, vnet, route, num_flits, new_msg_ptr,
m_net_ptr->MessageSizeType_to_int(
net_msg_ptr->getMessageSize()),
oPort->bitWidth(), curTick());
fl->set_src_delay(curTick() - msg_ptr->getTime());
niOutVcs[vc].insert(fl);
}
m_ni_out_vcs_enqueue_time[vc] = curTick();
outVcState[vc].setState(ACTIVE_, curTick());
}
return true ;
}
// Looking for a free output vc
int
NetworkInterface::calculateVC(int vnet)
{
for (int i = 0; i < m_vc_per_vnet; i++) {
int delta = m_vc_allocator[vnet];
m_vc_allocator[vnet]++;
if (m_vc_allocator[vnet] == m_vc_per_vnet)
m_vc_allocator[vnet] = 0;
if (outVcState[(vnet*m_vc_per_vnet) + delta].isInState(
IDLE_, curTick())) {
vc_busy_counter[vnet] = 0;
return ((vnet*m_vc_per_vnet) + delta);
}
}
vc_busy_counter[vnet] += 1;
panic_if(vc_busy_counter[vnet] > m_deadlock_threshold,
"%s: Possible network deadlock in vnet: %d at time: %llu \n",
name(), vnet, curTick());
return -1;
}
void
NetworkInterface::scheduleOutputPort(OutputPort *oPort)
{
int vc = oPort->vcRoundRobin();
for (int i = 0; i < niOutVcs.size(); i++) {
vc++;
if (vc == niOutVcs.size())
vc = 0;
int t_vnet = get_vnet(vc);
if (oPort->isVnetSupported(t_vnet)) {
// model buffer backpressure
if (niOutVcs[vc].isReady(curTick()) &&
outVcState[vc].has_credit()) {
bool is_candidate_vc = true;
int vc_base = t_vnet * m_vc_per_vnet;
if (m_net_ptr->isVNetOrdered(t_vnet)) {
for (int vc_offset = 0; vc_offset < m_vc_per_vnet;
vc_offset++) {
int t_vc = vc_base + vc_offset;
if (niOutVcs[t_vc].isReady(curTick())) {
if (m_ni_out_vcs_enqueue_time[t_vc] <
m_ni_out_vcs_enqueue_time[vc]) {
is_candidate_vc = false;
break;
}
}
}
}
if (!is_candidate_vc)
continue;
// Update the round robin arbiter
oPort->vcRoundRobin(vc);
outVcState[vc].decrement_credit();
// Just removing the top flit
flit *t_flit = niOutVcs[vc].getTopFlit();
t_flit->set_time(clockEdge(Cycles(1)));
// Scheduling the flit
scheduleFlit(t_flit);
if (t_flit->get_type() == TAIL_ ||
t_flit->get_type() == HEAD_TAIL_) {
m_ni_out_vcs_enqueue_time[vc] = Tick(INFINITE_);
}
// Done with this port, continue to schedule
// other ports
return;
}
}
}
}
/** This function looks at the NI buffers
* if some buffer has flits which are ready to traverse the link in the next
* cycle, and the downstream output vc associated with this flit has buffers
* left, the link is scheduled for the next cycle
*/
void
NetworkInterface::scheduleOutputLink()
{
// Schedule each output link
for (auto &oPort: outPorts) {
scheduleOutputPort(oPort);
}
}
NetworkInterface::InputPort *
NetworkInterface::getInportForVnet(int vnet)
{
for (auto &iPort : inPorts) {
if (iPort->isVnetSupported(vnet)) {
return iPort;
}
}
return nullptr;
}
/*
* This function returns the outport which supports the given vnet.
* Currently, HeteroGarnet does not support multiple outports to
* support same vnet. Thus, this function returns the first-and
* only outport which supports the vnet.
*/
NetworkInterface::OutputPort *
NetworkInterface::getOutportForVnet(int vnet)
{
for (auto &oPort : outPorts) {
if (oPort->isVnetSupported(vnet)) {
return oPort;
}
}
return nullptr;
}
void
NetworkInterface::scheduleFlit(flit *t_flit)
{
OutputPort *oPort = getOutportForVnet(t_flit->get_vnet());
if (oPort) {
DPRINTF(RubyNetwork, "Scheduling at %s time:%ld flit:%s Message:%s\n",
oPort->outNetLink()->name(), clockEdge(Cycles(1)),
*t_flit, *(t_flit->get_msg_ptr()));
oPort->outFlitQueue()->insert(t_flit);
oPort->outNetLink()->scheduleEventAbsolute(clockEdge(Cycles(1)));
return;
}
panic("No output port found for vnet:%d\n", t_flit->get_vnet());
return;
}
int
NetworkInterface::get_vnet(int vc)
{
for (int i = 0; i < m_virtual_networks; i++) {
if (vc >= (i*m_vc_per_vnet) && vc < ((i+1)*m_vc_per_vnet)) {
return i;
}
}
fatal("Could not determine vc");
}
// Wakeup the NI in the next cycle if there are waiting
// messages in the protocol buffer, or waiting flits in the
// output VC buffer.
// Also check if we have to reschedule because of a clock period
// difference.
void
NetworkInterface::checkReschedule()
{
for (const auto& it : inNode_ptr) {
if (it == nullptr) {
continue;
}
while (it->isReady(clockEdge())) { // Is there a message waiting
scheduleEvent(Cycles(1));
return;
}
}
for (auto& ni_out_vc : niOutVcs) {
if (ni_out_vc.isReady(clockEdge(Cycles(1)))) {
scheduleEvent(Cycles(1));
return;
}
}
// Check if any input links have flits to be popped.
// This can happen if the links are operating at
// a higher frequency.
for (auto &iPort : inPorts) {
NetworkLink *inNetLink = iPort->inNetLink();
if (inNetLink->isReady(curTick())) {
scheduleEvent(Cycles(1));
return;
}
}
for (auto &oPort : outPorts) {
CreditLink *inCreditLink = oPort->inCreditLink();
if (inCreditLink->isReady(curTick())) {
scheduleEvent(Cycles(1));
return;
}
}
}
void
NetworkInterface::print(std::ostream& out) const
{
out << "[Network Interface]";
}
uint32_t
NetworkInterface::functionalWrite(Packet *pkt)
{
uint32_t num_functional_writes = 0;
for (auto& ni_out_vc : niOutVcs) {
num_functional_writes += ni_out_vc.functionalWrite(pkt);
}
for (auto &oPort: outPorts) {
num_functional_writes += oPort->outFlitQueue()->functionalWrite(pkt);
}
return num_functional_writes;
}
} // namespace garnet
} // namespace ruby
} // namespace gem5