4f7b3ed827
There are several parts to this PR to work towards #1349 . (1) Make RubySystem::getBlockSizeBytes non-static by providing ways to access the block size or passing the block size explicitly to classes. The main changes are: - DataBlocks must be explicitly allocated. A default ctor still exists to avoid needing to heavily modify SLICC. The size can be set using a realloc function, operator=, or copy ctor. This is handled completely transparently meaning no protocol or config changes are required. - WriteMask now requires block size to be set. This is also handled transparently by modifying the SLICC parser to identify WriteMask types and call setBlockSize(). - AbstractCacheEntry and TBE classes now require block size to be set. This is handled transparently by modifying the SLICC parser to identify these classes and call initBlockSize() which calls setBlockSize() for any DataBlock or WriteMask. - All AbstractControllers now have a pointer to RubySystem. This is assigned in SLICC generated code and requires no changes to protocol or configs. - The Ruby Message class now requires block size in all constructors. This is added to the argument list automatically by the SLICC parser. (2) Relax dependence on common functions in src/mem/ruby/common/Address.hh so that RubySystem::getBlockSizeBits is no longer static. Many classes already have a way to get block size from the previous commit, so they simply multiple by 8 to get the number of bits. For handling SLICC and reducing the number of changes, define makeCacheLine, getOffset, etc. in RubyPort and AbstractController. The only protocol changes required are to change any "RubySystem::foo()" calls with "m_ruby_system->foo()". For classes which do not have a way to get access to block size but still used makeLineAddress, getOffset, etc., the block size must be passed to that class. This requires some changes to the SimObject interface for two commonly used classes: DirectoryMemory and RubyPrefecther, resulting in user-facing API changes User-facing API changes: - DirectoryMemory and RubyPrefetcher now require the cache line size as a non-optional argument. - RubySequencer SimObjects now require RubySystem as a non-optional argument. - TesterThread in the GPU ruby tester now requires the cache line size as a non-optional argument. (3) Removes static member variables in RubySystem which control randomization, cooldown, and warmup. These are mostly used by the Ruby Network. The network classes are modified to take these former static variables as parameters which are passed to the corresponding method (e.g., enqueue, delayHead, etc.) rather than needing a RubySystem object at all. Change-Id: Ia63c2ad5cf0bf9d1cbdffba5d3a679bb4d3b1220 (4) There are two major SLICC generated static methods: getNumControllers() on each cache controller which returns the number of controllers created by the configs at run time and the functions which access this method, which are MachineType_base_count and MachineType_base_number. These need to be removed to create multiple RubySystem objects otherwise NetDest, version value, and other objects are incorrect. To remove the static requirement, MachineType_base_count and MachineType_base_number are moved to RubySystem. Any class which needs to call these methods must now have a pointer to a RubySystem. To enable that, several changes are made: - RubyRequest and Message now require a RubySystem pointer in the constructor. The pointer is passed to fields in the Message class which require a RubySystem pointer (e.g., NetDest). SLICC is modified to do this automatically. - SLICC structures may now optionally take an "implicit constructor" which can be used to call a non-default constructor for locally defined variables (e.g., temporary variables within SLICC actions). A statement such as "NetDest bcast_dest;" in SLICC will implicitly append a call to the NetDest constructor taking RubySystem, for example. - RubySystem gets passed to Ruby network objects (Network, Topology).
397 lines
14 KiB
C++
397 lines
14 KiB
C++
/*
|
|
* Copyright (c) 2021 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) 1999-2008 Mark D. Hill and David A. Wood
|
|
* 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/simple/Throttle.hh"
|
|
|
|
#include <cassert>
|
|
|
|
#include "base/cast.hh"
|
|
#include "base/cprintf.hh"
|
|
#include "debug/RubyNetwork.hh"
|
|
#include "mem/ruby/network/MessageBuffer.hh"
|
|
#include "mem/ruby/network/Network.hh"
|
|
#include "mem/ruby/network/simple/Switch.hh"
|
|
#include "mem/ruby/slicc_interface/Message.hh"
|
|
#include "mem/ruby/system/RubySystem.hh"
|
|
#include "sim/stats.hh"
|
|
|
|
namespace gem5
|
|
{
|
|
|
|
namespace ruby
|
|
{
|
|
|
|
const int MESSAGE_SIZE_MULTIPLIER = 1000;
|
|
//const int BROADCAST_SCALING = 4; // Have a 16p system act like a 64p systems
|
|
const int BROADCAST_SCALING = 1;
|
|
const int PRIORITY_SWITCH_LIMIT = 128;
|
|
|
|
static int network_message_to_size(Message* net_msg_ptr);
|
|
|
|
Throttle::Throttle(int sID, RubySystem *rs, NodeID node, Cycles link_latency,
|
|
int endpoint_bandwidth, Switch *em)
|
|
: Consumer(em, Switch::THROTTLE_EV_PRI),
|
|
m_switch_id(sID), m_switch(em), m_node(node),
|
|
m_physical_vnets(false), m_ruby_system(rs),
|
|
throttleStats(em, node)
|
|
{
|
|
m_vnets = 0;
|
|
|
|
m_link_latency = link_latency;
|
|
m_endpoint_bandwidth = endpoint_bandwidth;
|
|
|
|
m_wakeups_wo_switch = 0;
|
|
}
|
|
|
|
Throttle::Throttle(int sID, RubySystem *rs, NodeID node, Cycles link_latency,
|
|
int link_bandwidth_multiplier, int endpoint_bandwidth,
|
|
Switch *em)
|
|
: Throttle(sID, rs, node, link_latency, endpoint_bandwidth, em)
|
|
{
|
|
gem5_assert(link_bandwidth_multiplier > 0);
|
|
m_link_bandwidth_multiplier.push_back(link_bandwidth_multiplier);
|
|
}
|
|
|
|
Throttle::Throttle(int sID, RubySystem *rs, NodeID node, Cycles link_latency,
|
|
const std::vector<int> &vnet_channels,
|
|
const std::vector<int> &vnet_bandwidth_multiplier,
|
|
int endpoint_bandwidth, Switch *em)
|
|
: Throttle(sID, rs, node, link_latency, endpoint_bandwidth, em)
|
|
{
|
|
m_physical_vnets = true;
|
|
for (auto link_bandwidth_multiplier : vnet_bandwidth_multiplier){
|
|
gem5_assert(link_bandwidth_multiplier > 0);
|
|
m_link_bandwidth_multiplier.push_back(link_bandwidth_multiplier);
|
|
}
|
|
for (auto channels : vnet_channels){
|
|
gem5_assert(channels > 0);
|
|
m_vnet_channels.push_back(channels);
|
|
}
|
|
gem5_assert(m_link_bandwidth_multiplier.size() == m_vnet_channels.size());
|
|
}
|
|
|
|
void
|
|
Throttle::addLinks(const std::vector<MessageBuffer*>& in_vec,
|
|
const std::vector<MessageBuffer*>& out_vec)
|
|
{
|
|
assert(in_vec.size() == out_vec.size());
|
|
|
|
for (int vnet = 0; vnet < in_vec.size(); ++vnet) {
|
|
MessageBuffer *in_ptr = in_vec[vnet];
|
|
MessageBuffer *out_ptr = out_vec[vnet];
|
|
|
|
m_units_remaining.emplace_back(getChannelCnt(vnet),0);
|
|
m_in.push_back(in_ptr);
|
|
m_out.push_back(out_ptr);
|
|
|
|
// Set consumer and description
|
|
in_ptr->setConsumer(this);
|
|
std::string desc = "[Queue to Throttle " +
|
|
std::to_string(m_switch_id) + " " + std::to_string(m_node) + "]";
|
|
}
|
|
|
|
m_vnets = in_vec.size();
|
|
|
|
gem5_assert(m_physical_vnets ?
|
|
(m_link_bandwidth_multiplier.size() == m_vnets) :
|
|
(m_link_bandwidth_multiplier.size() == 1));
|
|
}
|
|
|
|
int
|
|
Throttle::getLinkBandwidth(int vnet) const
|
|
{
|
|
int bw = m_physical_vnets ?
|
|
m_link_bandwidth_multiplier[vnet] :
|
|
m_link_bandwidth_multiplier[0];
|
|
gem5_assert(bw > 0);
|
|
return m_endpoint_bandwidth * bw;
|
|
}
|
|
|
|
int
|
|
Throttle::getTotalLinkBandwidth() const
|
|
{
|
|
int sum = getLinkBandwidth(0) * getChannelCnt(0);
|
|
if (m_physical_vnets) {
|
|
for (unsigned i = 1; i < m_vnets; ++i)
|
|
sum += getLinkBandwidth(i) * getChannelCnt(i);
|
|
}
|
|
return sum;
|
|
}
|
|
|
|
int
|
|
Throttle::getChannelCnt(int vnet) const
|
|
{
|
|
return m_physical_vnets ? m_vnet_channels[vnet] : 1;
|
|
}
|
|
|
|
void
|
|
Throttle::operateVnet(int vnet, int channel, int &total_bw_remaining,
|
|
bool &bw_saturated, bool &output_blocked,
|
|
MessageBuffer *in, MessageBuffer *out)
|
|
{
|
|
if (out == nullptr || in == nullptr) {
|
|
return;
|
|
}
|
|
|
|
int &units_remaining = m_units_remaining[vnet][channel];
|
|
|
|
gem5_assert(units_remaining >= 0);
|
|
Tick current_time = m_switch->clockEdge();
|
|
|
|
int bw_remaining = m_physical_vnets ?
|
|
getLinkBandwidth(vnet) : total_bw_remaining;
|
|
|
|
auto hasPendingWork = [&]{ return in->isReady(current_time) ||
|
|
units_remaining > 0; };
|
|
while ((bw_remaining > 0) && hasPendingWork() &&
|
|
out->areNSlotsAvailable(1, current_time)) {
|
|
// See if we are done transferring the previous message on
|
|
// this virtual network
|
|
if (units_remaining == 0 && in->isReady(current_time)) {
|
|
// Find the size of the message we are moving
|
|
MsgPtr msg_ptr = in->peekMsgPtr();
|
|
Message *net_msg_ptr = msg_ptr.get();
|
|
Tick msg_enqueue_time = msg_ptr->getLastEnqueueTime();
|
|
units_remaining = network_message_to_size(net_msg_ptr);
|
|
|
|
DPRINTF(RubyNetwork, "throttle: %d my bw %d bw spent "
|
|
"enqueueing net msg %d time: %lld.\n",
|
|
m_node, getLinkBandwidth(vnet), units_remaining,
|
|
m_ruby_system->curCycle());
|
|
|
|
// Move the message
|
|
in->dequeue(current_time);
|
|
out->enqueue(msg_ptr, current_time,
|
|
m_switch->cyclesToTicks(m_link_latency),
|
|
m_ruby_system->getRandomization(),
|
|
m_ruby_system->getWarmupEnabled());
|
|
|
|
// Count the message
|
|
(*(throttleStats.
|
|
msg_counts[net_msg_ptr->getMessageSize()]))[vnet]++;
|
|
throttleStats.total_msg_count += 1;
|
|
uint32_t total_size =
|
|
Network::MessageSizeType_to_int(net_msg_ptr->getMessageSize());
|
|
throttleStats.total_msg_bytes += total_size;
|
|
total_size -=
|
|
Network::MessageSizeType_to_int(MessageSizeType_Control);
|
|
throttleStats.total_data_msg_bytes += total_size;
|
|
throttleStats.total_msg_wait_time +=
|
|
current_time - msg_enqueue_time;
|
|
DPRINTF(RubyNetwork, "%s\n", *out);
|
|
}
|
|
|
|
// Calculate the amount of bandwidth we spent on this message
|
|
int spent = std::min(units_remaining, bw_remaining);
|
|
units_remaining -= spent;
|
|
bw_remaining -= spent;
|
|
total_bw_remaining -= spent;
|
|
}
|
|
|
|
gem5_assert(units_remaining >= 0);
|
|
gem5_assert(bw_remaining >= 0);
|
|
gem5_assert(total_bw_remaining >= 0);
|
|
|
|
// Notify caller if
|
|
// - we ran out of bandwith and still have stuff to do
|
|
// - we had something to do but output queue was unavailable
|
|
if (hasPendingWork()) {
|
|
gem5_assert((bw_remaining == 0) ||
|
|
!out->areNSlotsAvailable(1, current_time));
|
|
bw_saturated = bw_saturated || (bw_remaining == 0);
|
|
output_blocked = output_blocked ||
|
|
!out->areNSlotsAvailable(1, current_time);
|
|
}
|
|
}
|
|
|
|
void
|
|
Throttle::wakeup()
|
|
{
|
|
// Limits the number of message sent to a limited number of bytes/cycle.
|
|
assert(getTotalLinkBandwidth() > 0);
|
|
int bw_remaining = getTotalLinkBandwidth();
|
|
|
|
m_wakeups_wo_switch++;
|
|
bool bw_saturated = false;
|
|
bool output_blocked = false;
|
|
|
|
// variable for deciding the direction in which to iterate
|
|
bool iteration_direction = false;
|
|
|
|
|
|
// invert priorities to avoid starvation seen in the component network
|
|
if (m_wakeups_wo_switch > PRIORITY_SWITCH_LIMIT) {
|
|
m_wakeups_wo_switch = 0;
|
|
iteration_direction = true;
|
|
}
|
|
|
|
if (iteration_direction) {
|
|
for (int vnet = 0; vnet < m_vnets; ++vnet) {
|
|
for (int channel = 0; channel < getChannelCnt(vnet); ++channel) {
|
|
operateVnet(vnet, channel, bw_remaining,
|
|
bw_saturated, output_blocked,
|
|
m_in[vnet], m_out[vnet]);
|
|
}
|
|
}
|
|
} else {
|
|
for (int vnet = m_vnets-1; vnet >= 0; --vnet) {
|
|
for (int channel = 0; channel < getChannelCnt(vnet); ++channel) {
|
|
operateVnet(vnet, channel, bw_remaining,
|
|
bw_saturated, output_blocked,
|
|
m_in[vnet], m_out[vnet]);
|
|
}
|
|
}
|
|
}
|
|
|
|
// We should only wake up when we use the bandwidth
|
|
// This is only mostly true
|
|
// assert(bw_remaining != getLinkBandwidth());
|
|
|
|
// Record that we used some or all of the link bandwidth this cycle
|
|
double ratio = 1.0 - (double(bw_remaining) /
|
|
double(getTotalLinkBandwidth()));
|
|
|
|
// If ratio = 0, we used no bandwidth, if ratio = 1, we used all
|
|
throttleStats.acc_link_utilization += ratio;
|
|
|
|
if (bw_saturated) throttleStats.total_bw_sat_cy += 1;
|
|
if (output_blocked) throttleStats.total_stall_cy += 1;
|
|
|
|
if (bw_saturated || output_blocked) {
|
|
// We are out of bandwidth for this cycle, so wakeup next
|
|
// cycle and continue
|
|
DPRINTF(RubyNetwork, "%s scheduled again\n", *this);
|
|
scheduleEvent(Cycles(1));
|
|
}
|
|
}
|
|
|
|
void
|
|
Throttle::print(std::ostream& out) const
|
|
{
|
|
ccprintf(out, "[%i bw:", m_node);
|
|
if (m_physical_vnets) {
|
|
for (unsigned i = 0; i < m_vnets; ++i)
|
|
ccprintf(out, " vnet%d=%i", i, getLinkBandwidth(i));
|
|
} else {
|
|
ccprintf(out, " %i", getTotalLinkBandwidth());
|
|
}
|
|
ccprintf(out, "]");
|
|
}
|
|
|
|
int
|
|
network_message_to_size(Message *net_msg_ptr)
|
|
{
|
|
assert(net_msg_ptr != NULL);
|
|
|
|
int size = Network::MessageSizeType_to_int(net_msg_ptr->getMessageSize());
|
|
size *= MESSAGE_SIZE_MULTIPLIER;
|
|
|
|
// Artificially increase the size of broadcast messages
|
|
if (BROADCAST_SCALING > 1 && net_msg_ptr->getDestination().isBroadcast())
|
|
size *= BROADCAST_SCALING;
|
|
|
|
return size;
|
|
}
|
|
|
|
Throttle::
|
|
ThrottleStats::ThrottleStats(Switch *parent, const NodeID &nodeID)
|
|
: statistics::Group(parent, csprintf("throttle%02i", nodeID).c_str()),
|
|
ADD_STAT(acc_link_utilization, statistics::units::Count::get(),
|
|
"Accumulated link utilization"),
|
|
ADD_STAT(link_utilization, statistics::units::Ratio::get(),
|
|
"Average link utilization"),
|
|
ADD_STAT(total_msg_count, statistics::units::Count::get(),
|
|
"Total number of messages forwarded by this switch"),
|
|
ADD_STAT(total_msg_bytes, statistics::units::Byte::get(),
|
|
"Total number of bytes forwarded by this switch"),
|
|
ADD_STAT(total_data_msg_bytes, statistics::units::Byte::get(),
|
|
"Total number of data bytes forwarded by this switch"),
|
|
ADD_STAT(total_msg_wait_time, statistics::units::Tick::get(),
|
|
"Total time spend forwarding messages"),
|
|
ADD_STAT(total_stall_cy, statistics::units::Cycle::get(),
|
|
"Total time spent blocked on any output link"),
|
|
ADD_STAT(total_bw_sat_cy, statistics::units::Cycle::get(),
|
|
"Total time bandwidth was saturated on any output link"),
|
|
ADD_STAT(avg_msg_wait_time, statistics::units::Ratio::get(),
|
|
"Average time a message took to be forwarded"),
|
|
ADD_STAT(avg_bandwidth, statistics::units::Ratio::get(),
|
|
"Average bandwidth (GB/s)"),
|
|
ADD_STAT(avg_useful_bandwidth, statistics::units::Ratio::get(),
|
|
"Average usefull (only data) bandwidth (GB/s)")
|
|
{
|
|
link_utilization = 100 * acc_link_utilization /
|
|
(simTicks / parent->clockPeriod());
|
|
|
|
avg_msg_wait_time = total_msg_wait_time / total_msg_count;
|
|
|
|
avg_bandwidth.precision(2);
|
|
avg_bandwidth = (total_msg_bytes / simSeconds) /
|
|
statistics::constant(1024*1024*1024);
|
|
|
|
avg_useful_bandwidth.precision(2);
|
|
avg_useful_bandwidth = (total_data_msg_bytes / simSeconds) /
|
|
statistics::constant(1024*1024*1024);
|
|
|
|
for (MessageSizeType type = MessageSizeType_FIRST;
|
|
type < MessageSizeType_NUM; ++type) {
|
|
msg_counts[(unsigned int)type] =
|
|
new statistics::Vector(this,
|
|
csprintf("msg_count.%s", MessageSizeType_to_string(type)).c_str());
|
|
msg_counts[(unsigned int)type]
|
|
->init(Network::getNumberOfVirtualNetworks())
|
|
.flags(statistics::nozero)
|
|
;
|
|
|
|
msg_bytes[(unsigned int) type] =
|
|
new statistics::Formula(this,
|
|
csprintf("msg_bytes.%s", MessageSizeType_to_string(type)).c_str());
|
|
msg_bytes[(unsigned int) type]
|
|
->flags(statistics::nozero)
|
|
;
|
|
|
|
*(msg_bytes[(unsigned int) type]) =
|
|
*(msg_counts[type]) * statistics::constant(
|
|
Network::MessageSizeType_to_int(type));
|
|
}
|
|
}
|
|
|
|
} // namespace ruby
|
|
} // namespace gem5
|