From 3b6a143ec577b74a3bc65b84d3fe0416a094b2d0 Mon Sep 17 00:00:00 2001 From: Andreas Hansson Date: Fri, 21 Sep 2012 11:48:13 -0400 Subject: DRAM: Introduce SimpleDRAM to capture a high-level controller This patch introduces a high-level model of a DRAM controller, with a basic read/write buffer structure, a selectable and customisable arbiter, a few address mapping options, and the basic DRAM timing constraints. The parameters make it possible to turn this model into any desired DDRx/LPDDRx/WideIOx memory controller. The intention is not to be cycle accurate or capture every aspect of a DDR DRAM interface, but rather to enable exploring of the high-level knobs with a good simulation speed. Thus, contrary to e.g. DRAMSim this module emphasizes simulation speed with a good-enough accuracy. This module is merely a starting point, and there are plenty additions and improvements to come. A notable addition is the support for address-striping in the bus to enable a multi-channel DRAM controller. Also note that there are still a few "todo's" in the code base that will be addressed as we go along. A follow-up patch will add basic performance regressions that use the traffic generator to exercise a few well-defined corner cases. --- src/mem/SConscript | 4 + src/mem/SimpleDRAM.py | 130 +++++ src/mem/simple_dram.cc | 1264 ++++++++++++++++++++++++++++++++++++++++++++++++ src/mem/simple_dram.hh | 478 ++++++++++++++++++ 4 files changed, 1876 insertions(+) create mode 100644 src/mem/SimpleDRAM.py create mode 100644 src/mem/simple_dram.cc create mode 100644 src/mem/simple_dram.hh (limited to 'src/mem') diff --git a/src/mem/SConscript b/src/mem/SConscript index 7aba4f0e7..1c43975df 100644 --- a/src/mem/SConscript +++ b/src/mem/SConscript @@ -53,10 +53,12 @@ Source('se_translating_port_proxy.cc') if env['TARGET_ISA'] != 'no': SimObject('AbstractMemory.py') SimObject('SimpleMemory.py') + SimObject('SimpleDRAM.py') Source('abstract_mem.cc') Source('simple_mem.cc') Source('page_table.cc') Source('physical.cc') + Source('simple_dram.cc') DebugFlag('BaseBus') DebugFlag('BusAddrRanges') @@ -67,6 +69,8 @@ CompoundFlag('Bus', ['BaseBus', 'BusAddrRanges', 'CoherentBus', DebugFlag('Bridge') DebugFlag('CommMonitor') +DebugFlag('DRAM') +DebugFlag('DRAMWR') DebugFlag('LLSC') DebugFlag('MMU') DebugFlag('MemoryAccess') diff --git a/src/mem/SimpleDRAM.py b/src/mem/SimpleDRAM.py new file mode 100644 index 000000000..c87425610 --- /dev/null +++ b/src/mem/SimpleDRAM.py @@ -0,0 +1,130 @@ +# Copyright (c) 2012 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. +# +# 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: Andreas Hansson +# Ani Udipi + +from m5.params import * +from AbstractMemory import * + +# Enum for memory scheduling algorithms, currently First-Come +# First-Served and a First-Row Hit then First-Come First-Served +class MemSched(Enum): vals = ['fcfs', 'frfcfs'] + +# Enum for the address mapping, currently corresponding to either +# optimising for sequential accesses hitting in the open row, or +# striping across banks. +class AddrMap(Enum): vals = ['openmap', 'closemap'] + +# Enum for the page policy, either open or close. +class PageManage(Enum): vals = ['open', 'close'] + +# SimpleDRAM is a single-channel single-ported DRAM controller model +# that aims to model the most important system-level performance +# effects of a DRAM without getting into too much detail of the DRAM +# itself. +class SimpleDRAM(AbstractMemory): + type = 'SimpleDRAM' + + # single-ported on the system interface side, instantiate with a + # bus in front of the controller for multiple ports + port = SlavePort("Slave port") + + # the physical organisation of the DRAM + lines_per_rowbuffer = Param.Unsigned(64, "Row buffer size in cache lines") + ranks_per_channel = Param.Unsigned(2, "Number of ranks per channel") + banks_per_rank = Param.Unsigned(8, "Number of banks per rank") + + # the basic configuration of the controller architecture + write_buffer_size = Param.Unsigned(32, "Number of read queue entries") + read_buffer_size = Param.Unsigned(32, "Number of write queue entries") + + # threshold in percent for when to trigger writes and start + # emptying the write buffer as it starts to get full + write_thresh_perc = Param.Percent(70, "Threshold to trigger writes") + + # scheduler, address map and page policy + mem_sched_policy = Param.MemSched('fcfs', "Memory scheduling policy") + addr_mapping = Param.AddrMap('openmap', "Address mapping policy") + page_policy = Param.PageManage('open', "Page closure management policy") + + # timing behaviour and constraints - all in nanoseconds + + # the amount of time in nanoseconds from issuing an activate command + # to the data being available in the row buffer for a read/write + tRCD = Param.Latency("14ns", "RAS to CAS delay") + + # the time from issuing a read/write command to seeing the actual data + tCL = Param.Latency("14ns", "CAS latency") + + # minimum time between a precharge and subsequent activate + tRP = Param.Latency("14ns", "Row precharge time") + + # time to complete a burst transfer, typically the burst length + # divided by two due to the DDR bus, but by making it a parameter + # it is easier to also evaluate SDR memories like WideIO. + # This parameter has to account for bus width and burst length. + # Adjustment also necessary if cache line size is greater than + # data size read/written by one full burst. + tBURST = Param.Latency("4ns", + "Burst duration (for DDR burst length / 2 cycles)") + + # time taken to complete one refresh cycle (N rows in all banks) + tRFC = Param.Latency("300ns", "Refresh cycle time") + + # refresh command interval, how often a "ref" command needs + # to be sent. It is 7.8 us for a 64ms refresh requirement + tREFI = Param.Latency("7.8us", "Refresh command interval") + + # write-to-read turn around penalty, assumed same as read-to-write + tWTR = Param.Latency("1ns", "Write to read switching time") + + # Currently unimplemented, unused, deduced or rolled into other params + ###################################################################### + + # the minimum amount of time between a row being activated, and + # precharged (de-activated) + # tRAS - assumed to be 3 * tRP + + # tRC - assumed to be 4 * tRP + + # burst length for an access derived from peerBlockSize + + # @todo: Implement tFAW in the model + # minimum time window in which a maximum of four activates are + # allowed to take place + # tFAW = Param.Latency("30ns", "Four activation window") + + diff --git a/src/mem/simple_dram.cc b/src/mem/simple_dram.cc new file mode 100644 index 000000000..5ee8643e1 --- /dev/null +++ b/src/mem/simple_dram.cc @@ -0,0 +1,1264 @@ +/* + * Copyright (c) 2010-2012 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. + * + * 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: Andreas Hansson + * Ani Udipi + */ + +#include "debug/DRAM.hh" +#include "debug/DRAMWR.hh" +#include "mem/simple_dram.hh" +#include "sim/stat_control.hh" + +using namespace std; + +SimpleDRAM::SimpleDRAM(const SimpleDRAMParams* p) : + AbstractMemory(p), + port(name() + ".port", *this), + retryRdReq(false), retryWrReq(false), + rowHitFlag(false), stopReads(false), + writeEvent(this), respondEvent(this), + refreshEvent(this), nextReqEvent(this), drainEvent(NULL), + bytesPerCacheLine(0), + linesPerRowBuffer(p->lines_per_rowbuffer), + ranksPerChannel(p->ranks_per_channel), + banksPerRank(p->banks_per_rank), rowsPerBank(0), + readBufferSize(p->read_buffer_size), + writeBufferSize(p->write_buffer_size), + writeThresholdPerc(p->write_thresh_perc), + tWTR(p->tWTR), tBURST(p->tBURST), + tRCD(p->tRCD), tCL(p->tCL), tRP(p->tRP), + tRFC(p->tRFC), tREFI(p->tREFI), + memSchedPolicy(p->mem_sched_policy), addrMapping(p->addr_mapping), + pageMgmt(p->page_policy), + busBusyUntil(0), prevdramaccess(0), writeStartTime(0), + prevArrival(0), numReqs(0) +{ + // create the bank states based on the dimensions of the ranks and + // banks + banks.resize(ranksPerChannel); + for (size_t c = 0; c < ranksPerChannel; ++c) { + banks[c].resize(banksPerRank); + } + + // round the write threshold percent to a whole number of entries + // in the buffer + writeThreshold = writeBufferSize * writeThresholdPerc / 100.0; +} + +void +SimpleDRAM::init() +{ + if (!port.isConnected()) { + fatal("SimpleDRAM %s is unconnected!\n", name()); + } else { + port.sendRangeChange(); + } + + // get the cache line size from the connected port + bytesPerCacheLine = port.peerBlockSize(); + + // we could deal with plenty options here, but for now do a quick + // sanity check + if (bytesPerCacheLine != 64 && bytesPerCacheLine != 32) + panic("Unexpected cache line size %d", bytesPerCacheLine); + + // determine the rows per bank by looking at the total capacity + uint64_t capacity = AbstractMemory::size(); + uint64_t i = 1; + while (i < 64 && capacity > ((1 << i))) { + ++i; + } + + // rounded up to nearest power of two + DPRINTF(DRAM, "i is %lld\n", i); + capacity = 1 << i; + + DPRINTF(DRAM, "Memory capacity %lld (%lld) bytes\n", capacity, + AbstractMemory::size()); + rowsPerBank = capacity / (bytesPerCacheLine * linesPerRowBuffer * + banksPerRank * ranksPerChannel); + +} + +void +SimpleDRAM::startup() +{ + // print the configuration of the controller + printParams(); + + // kick off the refresh + schedule(&refreshEvent, curTick() + tREFI); +} + + +Tick +SimpleDRAM::recvAtomic(PacketPtr pkt) +{ + DPRINTF(DRAM, "recvAtomic: %s 0x%x\n", pkt->cmdString(), pkt->getAddr()); + + // do the actual memory access and turn the packet into a response + access(pkt); + + Tick latency = 0; + if (!pkt->memInhibitAsserted() && pkt->hasData()) { + // this value is not supposed to be accurate, just enough to + // keep things going, mimic a closed page + latency = tRP + tRCD + tCL; + } + return latency; +} + +bool +SimpleDRAM::readQueueFull() const +{ + DPRINTF(DRAM, "Read queue limit %d current size %d\n", + readBufferSize, dramReadQueue.size() + dramRespQueue.size()); + + return (dramReadQueue.size() + dramRespQueue.size()) == readBufferSize; +} + +bool +SimpleDRAM::writeQueueFull() const +{ + DPRINTF(DRAM, "Write queue limit %d current size %d\n", + writeBufferSize, dramWriteQueue.size()); + return dramWriteQueue.size() == writeBufferSize; +} + + +SimpleDRAM::DRAMPacket* +SimpleDRAM::decodeAddr(PacketPtr pkt) +{ + uint8_t rank; + uint16_t bank; + uint16_t row; + + Addr addr = pkt->getAddr(); + Addr temp = addr; + + // truncate the address to the access granularity + addr = addr / bytesPerCacheLine; + + if (addrMapping == Enums::openmap) { + addr = addr / linesPerRowBuffer; + + bank = addr % banksPerRank; + addr = addr / banksPerRank; + + rank = addr % ranksPerChannel; + addr = addr / ranksPerChannel; + + row = addr % rowsPerBank; + addr = addr / rowsPerBank; + } else if (addrMapping == Enums::closemap) { + bank = addr % banksPerRank; + addr = addr / banksPerRank; + + rank = addr % ranksPerChannel; + addr = addr / ranksPerChannel; + + addr = addr / linesPerRowBuffer; + + row = addr % rowsPerBank; + addr = addr / rowsPerBank; + } else + panic("Unknown address mapping policy chosen!"); + + assert(rank < ranksPerChannel); + assert(bank < banksPerRank); + assert(row < rowsPerBank); + + DPRINTF(DRAM, "Address: %lld Rank %d Bank %d Row %d\n", + temp, rank, bank, row); + + // create the corresponding DRAM packet with the entry time and + // ready time set to the current tick, they will be updated later + DRAMPacket* dram_pkt = new DRAMPacket(pkt, rank, bank, row, temp, + banks[rank][bank]); + + return dram_pkt; +} + +void +SimpleDRAM::addToReadQueue(PacketPtr pkt) +{ + // only add to the read queue here. whenever the request is + // eventually done, set the readyTime, and call schedule() + assert(!pkt->isWrite()); + + // First check write buffer to see if the data is already at + // the controller + std::list::const_iterator i; + Addr addr = pkt->getAddr(); + + // @todo: add size check + for (i = dramWriteQueue.begin(); i != dramWriteQueue.end(); ++i) { + if ((*i)->addr == addr){ + servicedByWrQ++; + DPRINTF(DRAM,"Serviced by write Q\n"); + bytesRead += bytesPerCacheLine; + bytesConsumedRd += pkt->getSize(); + accessAndRespond(pkt); + return; + } + } + + DRAMPacket* dram_pkt = decodeAddr(pkt); + + assert(dramReadQueue.size() + dramRespQueue.size() < readBufferSize); + rdQLenPdf[dramReadQueue.size() + dramRespQueue.size()]++; + + DPRINTF(DRAM, "Adding to read queue\n"); + + dramReadQueue.push_back(dram_pkt); + + // Update stats + uint32_t bank_id = banksPerRank * dram_pkt->rank + dram_pkt->bank; + assert(bank_id < ranksPerChannel * banksPerRank); + perBankRdReqs[bank_id]++; + + avgRdQLen = dramReadQueue.size() + dramRespQueue.size(); + + // Special case where no arbitration is required between requests + if (!nextReqEvent.scheduled() && !stopReads) { + DPRINTF(DRAM, "Request %lld - need to schedule immediately"); + schedule(&nextReqEvent, curTick() + 1); + } +} + +void +SimpleDRAM::processWriteEvent() +{ + assert(!dramWriteQueue.empty()); + uint32_t numWritesThisTime = 0; + + DPRINTF(DRAMWR, "Beginning DRAM Writes\n"); + Tick temp1 M5_VAR_USED = std::max(curTick(), busBusyUntil); + Tick temp2 M5_VAR_USED = std::max(curTick(), maxBankFreeAt()); + + // @todo: are there any dangers with the untimed while loop? + while (!dramWriteQueue.empty()) { + if (numWritesThisTime > writeThreshold) + break; + + chooseNextWrite(); + DRAMPacket* dram_pkt = dramWriteQueue.front(); + // What's the earlier the request can be put on the bus + Tick schedTime = std::max(curTick(), busBusyUntil); + + DPRINTF(DRAMWR, "Asking for latency estimate at %lld\n", + schedTime + tBURST); + + pair lat = estimateLatency(dram_pkt, schedTime + tBURST); + Tick accessLat = lat.second; + + // look at the rowHitFlag set by estimateLatency + + // @todo: Race condition here where another packet gives rise + // to another call to estimateLatency in the meanwhile? + if (rowHitFlag) + writeRowHits++; + + Bank& bank = dram_pkt->bank_ref; + + if (pageMgmt == Enums::open) { + bank.openRow = dram_pkt->row; + bank.freeAt = schedTime + tBURST + accessLat; + + if (!rowHitFlag) + bank.tRASDoneAt = bank.freeAt + tRP; + + } else if (pageMgmt == Enums::close) { + bank.freeAt = schedTime + tBURST + accessLat + tRP + tRP; + DPRINTF(DRAMWR, "processWriteEvent::bank.freeAt for " + "banks_id %d is %lld\n", + dram_pkt->rank * banksPerRank + dram_pkt->bank, + bank.freeAt); + } else + panic("Unknown page management policy chosen\n"); + + // @todo: As of now, write goes on the databus asap, maybe + // be held up at bank. May want to change it to delay the + // schedTime itself. + busBusyUntil = schedTime + tBURST; + DPRINTF(DRAMWR,"Done writing to address %lld\n",dram_pkt->addr); + + + DPRINTF(DRAMWR,"schedtime is %lld, tBURST is %lld, " + "busbusyuntil is %lld\n", + schedTime, tBURST, busBusyUntil); + + dramWriteQueue.pop_front(); + delete dram_pkt; + + numWritesThisTime++; + } + + DPRINTF(DRAMWR, "Completed %d writes, bus busy for %lld ticks,"\ + "banks busy for %lld ticks\n", numWritesThisTime, + busBusyUntil - temp1, maxBankFreeAt() - temp2); + + // Update stats + avgWrQLen = dramWriteQueue.size(); + + // turn the bus back around for reads again + busBusyUntil += tWTR; + stopReads = false; + + if (retryWrReq) { + retryWrReq = false; + port.sendRetry(); + } + + // if there is nothing left in any queue, signal a drain + if (dramWriteQueue.empty() && dramReadQueue.empty() && + dramRespQueue.empty () && drainEvent) { + drainEvent->process(); + drainEvent = NULL; + } + + // Once you're done emptying the write queue, check if there's + // anything in the read queue, and call schedule if required + schedule(&nextReqEvent, busBusyUntil); +} + +void +SimpleDRAM::triggerWrites() +{ + DPRINTF(DRAM, "Writes triggered at %lld\n", curTick()); + // Flag variable to stop any more read scheduling + stopReads = true; + + writeStartTime = std::max(busBusyUntil, curTick()) + tWTR; + + DPRINTF(DRAM, "Writes scheduled at %lld\n", writeStartTime); + + assert(writeStartTime >= curTick()); + assert(!writeEvent.scheduled()); + schedule(&writeEvent, writeStartTime); +} + +void +SimpleDRAM::addToWriteQueue(PacketPtr pkt) +{ + // only add to the write queue here. whenever the request is + // eventually done, set the readyTime, and call schedule() + assert(pkt->isWrite()); + + DRAMPacket* dram_pkt = decodeAddr(pkt); + + assert(dramWriteQueue.size() < writeBufferSize); + wrQLenPdf[dramWriteQueue.size()]++; + + DPRINTF(DRAM, "Adding to write queue\n"); + + dramWriteQueue.push_back(dram_pkt); + + // Update stats + uint32_t bank_id = banksPerRank * dram_pkt->rank + dram_pkt->bank; + assert(bank_id < ranksPerChannel * banksPerRank); + perBankWrReqs[bank_id]++; + + avgWrQLen = dramWriteQueue.size(); + + // we do not wait for the writes to be send to the actual memory, + // but instead take responsibility for the consistency here and + // snoop the write queue for any upcoming reads + + bytesConsumedWr += pkt->getSize(); + bytesWritten += bytesPerCacheLine; + accessAndRespond(pkt); + + // If your write buffer is starting to fill up, drain it! + if (dramWriteQueue.size() > writeThreshold && !stopReads){ + triggerWrites(); + } +} + +void +SimpleDRAM::printParams() const +{ + // Sanity check print of important parameters + DPRINTF(DRAM, + "Memory controller %s physical organization\n" \ + "Bytes per cacheline %d\n" \ + "Lines per row buffer %d\n" \ + "Rows per bank %d\n" \ + "Banks per rank %d\n" \ + "Ranks per channel %d\n" \ + "Total mem capacity %u\n", + name(), bytesPerCacheLine ,linesPerRowBuffer, rowsPerBank, + banksPerRank, ranksPerChannel, bytesPerCacheLine * + linesPerRowBuffer * rowsPerBank * banksPerRank * ranksPerChannel); + + string scheduler = memSchedPolicy == Enums::fcfs ? "FCFS" : "FR-FCFS"; + string address_mapping = addrMapping == Enums::openmap ? "OPENMAP" : + "CLOSEMAP"; + string page_policy = pageMgmt == Enums::open ? "OPEN" : "CLOSE"; + + DPRINTF(DRAM, + "Memory controller %s characteristics\n" \ + "Read buffer size %d\n" \ + "Write buffer size %d\n" \ + "Write buffer thresh %d\n" \ + "Scheduler %s\n" \ + "Address mapping %s\n" \ + "Page policy %s\n", + name(), readBufferSize, writeBufferSize, writeThreshold, + scheduler, address_mapping, page_policy); + + DPRINTF(DRAM, "Memory controller %s timing specs\n" \ + "tRCD %d ticks\n" \ + "tCL %d ticks\n" \ + "tRP %d ticks\n" \ + "tBURST %d ticks\n" \ + "tRFC %d ticks\n" \ + "tREFI %d ticks\n" \ + "tWTR %d ticks\n", + name(), tRCD, tCL, tRP, tBURST, tRFC, tREFI, tWTR); +} + +void +SimpleDRAM::printQs() const { + + list::const_iterator i; + + DPRINTF(DRAM, "===READ QUEUE===\n\n"); + for (i = dramReadQueue.begin() ; i != dramReadQueue.end() ; ++i) { + DPRINTF(DRAM, "Read %lu\n", (*i)->addr); + } + DPRINTF(DRAM, "\n===RESP QUEUE===\n\n"); + for (i = dramRespQueue.begin() ; i != dramRespQueue.end() ; ++i) { + DPRINTF(DRAM, "Response %lu\n", (*i)->addr); + } + DPRINTF(DRAM, "\n===WRITE QUEUE===\n\n"); + for (i = dramWriteQueue.begin() ; i != dramWriteQueue.end() ; ++i) { + DPRINTF(DRAM, "Write %lu\n", (*i)->addr); + } +} + +bool +SimpleDRAM::recvTimingReq(PacketPtr pkt) +{ + // This is where we enter from the outside world + DPRINTF(DRAM, "Inside recvTimingReq: request %s addr %lld size %d\n", + pkt->cmdString(),pkt->getAddr(), pkt->getSize()); + + int index; + + if (pkt->getSize() == bytesPerCacheLine) + cpuReqs++; + + if (numReqs % 1000000 == 0) + printQs(); + + // Calc avg gap between requests + if (prevArrival != 0) { + totGap += curTick() - prevArrival; + } + prevArrival = curTick(); + + // simply drop inhibited packets for now + if (pkt->memInhibitAsserted()) { + DPRINTF(DRAM,"Inhibited packet -- Dropping it now\n"); + delete pkt; + return true; + } + + unsigned size = pkt->getSize(); + if (size > bytesPerCacheLine) + panic("Request size %d is greater than cache line size %d", + size, bytesPerCacheLine); + + if (size == 0) + index = log2(bytesPerCacheLine) + 1; + else + index = log2(size); + + if (size != 0 && (1 << index) != size) + index = log2(bytesPerCacheLine) + 2; + + // @todo: Do we really want to do all this before the packet is + // actually accepted? + + /* Index 0 - Size 1 byte + Index 1 - Size 2 bytes + Index 2 - Size 4 bytes + . + . + Index 6 - Size 64 bytes + Index 7 - Size 0 bytes + Index 8 - Non-power-of-2 size */ + + if (pkt->isRead()) + readPktSize[index]++; + else if (pkt->isWrite()) + writePktSize[index]++; + else + neitherPktSize[index]++; + + // check local buffers and do not accept if full + if (pkt->isRead()) { + if (readQueueFull()) { + DPRINTF(DRAM,"Read queue full, not accepting\n"); + // remember that we have to retry this port + retryRdReq = true; + numRdRetry++; + return false; + } else { + addToReadQueue(pkt); + readReqs++; + numReqs++; + } + } else if (pkt->isWrite()) { + if (writeQueueFull()) { + DPRINTF(DRAM,"Write queue full, not accepting\n"); + // remember that we have to retry this port + retryWrReq = true; + numWrRetry++; + return false; + } else { + addToWriteQueue(pkt); + writeReqs++; + numReqs++; + } + } else { + DPRINTF(DRAM,"Neither read nor write, ignore timing\n"); + neitherReadNorWrite++; + accessAndRespond(pkt); + } + + + retryRdReq = false; + retryWrReq = false; + return true; +} + +void +SimpleDRAM::processRespondEvent() +{ + DPRINTF(DRAM, + "processRespondEvent(): Some req has reached its readyTime\n"); + + PacketPtr pkt = dramRespQueue.front()->pkt; + + // Actually responds to the requestor + bytesConsumedRd += pkt->getSize(); + bytesRead += bytesPerCacheLine; + accessAndRespond(pkt); + + DRAMPacket* dram_pkt = dramRespQueue.front(); + dramRespQueue.pop_front(); + delete dram_pkt; + + // Update stats + avgRdQLen = dramReadQueue.size() + dramRespQueue.size(); + + if (!dramRespQueue.empty()){ + assert(dramRespQueue.front()->readyTime >= curTick()); + assert(!respondEvent.scheduled()); + schedule(&respondEvent, dramRespQueue.front()->readyTime); + } else { + // if there is nothing left in any queue, signal a drain + if (dramWriteQueue.empty() && dramReadQueue.empty() && + drainEvent) { + drainEvent->process(); + drainEvent = NULL; + } + } +} + +void +SimpleDRAM::chooseNextWrite() +{ + // This method does the arbitration between requests. The chosen + // packet is simply moved to the head of the queue. The other + // methods know that this is the place to look. For example, with + // FCFS, this method does nothing + assert(!dramWriteQueue.empty()); + + if (dramWriteQueue.size() == 1) { + DPRINTF(DRAMWR, "chooseNextWrite(): Single element, nothing to do\n"); + return; + } + + if (memSchedPolicy == Enums::fcfs) { + + // Do nothing, since the correct request is already head + + } else if (memSchedPolicy == Enums::frfcfs) { + + list::iterator i = dramWriteQueue.begin(); + bool foundRowHit = false; + while (!foundRowHit && i != dramWriteQueue.end()) { + DRAMPacket* dram_pkt = *i; + const Bank& bank = dram_pkt->bank_ref; + if (bank.openRow == dram_pkt->row) { //FR part + DPRINTF(DRAMWR,"Row buffer hit\n"); + dramWriteQueue.erase(i); + dramWriteQueue.push_front(dram_pkt); + foundRowHit = true; + } else { //FCFS part + ; + } + ++i; + } + + } else + panic("No scheduling policy chosen\n"); + + DPRINTF(DRAMWR, "chooseNextWrite(): Something chosen\n"); +} + +bool +SimpleDRAM::chooseNextReq() +{ + // This method does the arbitration between requests. + // The chosen packet is simply moved to the head of the + // queue. The other methods know that this is the place + // to look. For example, with FCFS, this method does nothing + list::iterator i; + DRAMPacket* dram_pkt; + + if (dramReadQueue.empty()){ + DPRINTF(DRAM, "chooseNextReq(): Returning False\n"); + return false; + } + + if (dramReadQueue.size() == 1) + return true; + + if (memSchedPolicy == Enums::fcfs) { + + // Do nothing, since the correct request is already head + + } else if (memSchedPolicy == Enums::frfcfs) { + + for (i = dramReadQueue.begin() ; i != dramReadQueue.end() ; ++i) { + dram_pkt = *i; + const Bank& bank = dram_pkt->bank_ref; + if (bank.openRow == dram_pkt->row) { //FR part + DPRINTF(DRAM, "Row buffer hit\n"); + dramReadQueue.erase(i); + dramReadQueue.push_front(dram_pkt); + break; + } else { //FCFS part + ; + } + + } + + } else + panic("No scheduling policy chosen!\n"); + + + DPRINTF(DRAM,"chooseNextReq(): Chosen something, returning True\n"); + return true; +} + +void +SimpleDRAM::accessAndRespond(PacketPtr pkt) +{ + DPRINTF(DRAM, "Responding to Address %lld.. ",pkt->getAddr()); + + bool needsResponse = pkt->needsResponse(); + // do the actual memory access which also turns the packet into a + // response + access(pkt); + + // turn packet around to go back to requester if response expected + if (needsResponse) { + // access already turned the packet into a response + assert(pkt->isResponse()); + + // queue the packet in the response queue to be sent out the + // next tick + port.schedTimingResp(pkt, curTick() + 1); + } else { + } + + DPRINTF(DRAM, "Done\n"); + + return; +} + +pair +SimpleDRAM::estimateLatency(DRAMPacket* dram_pkt, Tick inTime) +{ + // If a request reaches a bank at tick 'inTime', how much time + // *after* that does it take to finish the request, depending + // on bank status and page open policy. Note that this method + // considers only the time taken for the actual read or write + // to complete, NOT any additional time thereafter for tRAS or + // tRP. + Tick accLat = 0; + Tick bankLat = 0; + rowHitFlag = false; + + const Bank& bank = dram_pkt->bank_ref; + if (pageMgmt == Enums::open) { // open-page policy + if (bank.openRow == dram_pkt->row) { + // When we have a row-buffer hit, + // we don't care about tRAS having expired or not, + // but do care about bank being free for access + rowHitFlag = true; + + if (bank.freeAt < inTime) { + // CAS latency only + accLat += tCL; + bankLat += tCL; + } else { + accLat += 0; + bankLat += 0; + } + + } else { + // Row-buffer miss, need to close existing row + // once tRAS has expired, then open the new one, + // then add cas latency. + Tick freeTime = std::max(bank.tRASDoneAt, bank.freeAt); + + if (freeTime > inTime) + accLat += freeTime - inTime; + + accLat += tRP + tRCD + tCL; + bankLat += tRP + tRCD + tCL; + } + } else if (pageMgmt == Enums::close) { + + // With a close page policy, no notion of + // bank.tRASDoneAt + if (bank.freeAt > inTime) + accLat += bank.freeAt - inTime; + + // page already closed, simply open the row, and + // add cas latency + accLat += tRCD + tCL; + bankLat += tRCD + tCL; + } else + panic("No page management policy chosen\n"); + + DPRINTF(DRAM, "Returning %lld from estimateLatency()\n",accLat); + + return make_pair(bankLat, accLat); +} + +void +SimpleDRAM::processNextReqEvent() +{ + scheduleNextReq(); +} + +void +SimpleDRAM::doDRAMAccess(DRAMPacket* dram_pkt) +{ + + DPRINTF(DRAM, "Timing access to addr %lld, rank/bank/row %d %d %d\n", + dram_pkt->addr, dram_pkt->rank, dram_pkt->bank, dram_pkt->row); + + assert(curTick() >= prevdramaccess); + prevdramaccess = curTick(); + + // estimate the bank and access latency + pair lat = estimateLatency(dram_pkt, curTick()); + Tick bankLat = lat.first; + Tick accessLat = lat.second; + + // This request was woken up at this time based on a prior call + // to estimateLatency(). However, between then and now, both the + // accessLatency and/or busBusyUntil may have changed. We need + // to correct for that. + + Tick addDelay = (curTick() + accessLat < busBusyUntil) ? + busBusyUntil - (curTick() + accessLat) : 0; + + Bank& bank = dram_pkt->bank_ref; + + // Update bank state + if (pageMgmt == Enums::open) { + bank.openRow = dram_pkt->row; + bank.freeAt = curTick() + addDelay + accessLat; + // If you activated a new row do to this access, the next access + // will have to respect tRAS for this bank. Assume tRAS ~= 3 * tRP + if (!rowHitFlag) + bank.tRASDoneAt = bank.freeAt + tRP; + + } else if (pageMgmt == Enums::close) { // accounting for tRAS also + // assuming that tRAS ~= 3 * tRP, and tRAS ~= 4 * tRP, as is common + // (refer Jacob/Ng/Wang and Micron datasheets) + bank.freeAt = curTick() + addDelay + accessLat + tRP + tRP; + DPRINTF(DRAM,"doDRAMAccess::bank.freeAt is %lld\n",bank.freeAt); + } else + panic("No page management policy chosen\n"); + + // Update request parameters + dram_pkt->readyTime = curTick() + addDelay + accessLat + tBURST; + + + DPRINTF(DRAM, "Req %lld: curtick is %lld accessLat is %d " \ + "readytime is %lld busbusyuntil is %lld. " \ + "Scheduling at readyTime\n", dram_pkt->addr, + curTick(), accessLat, dram_pkt->readyTime, busBusyUntil); + + // Make sure requests are not overlapping on the databus + assert (dram_pkt->readyTime - busBusyUntil >= tBURST); + + // Update bus state + busBusyUntil = dram_pkt->readyTime; + + DPRINTF(DRAM,"Access time is %lld\n", + dram_pkt->readyTime - dram_pkt->entryTime); + + // Update stats + totMemAccLat += dram_pkt->readyTime - dram_pkt->entryTime; + totBankLat += bankLat; + totBusLat += tBURST; + totQLat += dram_pkt->readyTime - dram_pkt->entryTime - bankLat - tBURST; + + if (rowHitFlag) + readRowHits++; + + // At this point we're done dealing with the request + // It will be moved to a separate response queue with a + // correct readyTime, and eventually be sent back at that + //time + moveToRespQ(); + + // The absolute soonest you have to start thinking about the + // next request is the longest access time that can occur before + // busBusyUntil. Assuming you need to meet tRAS, then precharge, + // open a new row, and access, it is ~4*tRCD. + + + Tick newTime = (busBusyUntil > 4 * tRCD) ? + std::max(busBusyUntil - 4 * tRCD, curTick()) : + curTick(); + + if (!nextReqEvent.scheduled() && !stopReads){ + schedule(&nextReqEvent, newTime); + } else { + if (newTime < nextReqEvent.when()) + reschedule(&nextReqEvent, newTime); + } + + +} + +void +SimpleDRAM::moveToRespQ() +{ + // Remove from read queue + DRAMPacket* dram_pkt = dramReadQueue.front(); + dramReadQueue.pop_front(); + + // Insert into response queue sorted by readyTime + // It will be sent back to the requestor at its + // readyTime + if (dramRespQueue.empty()) { + dramRespQueue.push_front(dram_pkt); + assert(!respondEvent.scheduled()); + assert(dram_pkt->readyTime >= curTick()); + schedule(&respondEvent, dram_pkt->readyTime); + } else { + bool done = false; + std::list::iterator i = dramRespQueue.begin(); + while (!done && i != dramRespQueue.end()) { + if ((*i)->readyTime > dram_pkt->readyTime) { + dramRespQueue.insert(i, dram_pkt); + done = true; + } + ++i; + } + + if (!done) + dramRespQueue.push_back(dram_pkt); + + assert(respondEvent.scheduled()); + + if (dramRespQueue.front()->readyTime < respondEvent.when()) { + assert(dramRespQueue.front()->readyTime >= curTick()); + reschedule(&respondEvent, dramRespQueue.front()->readyTime); + } + } + + if (retryRdReq) { + retryRdReq = false; + port.sendRetry(); + } +} + +void +SimpleDRAM::scheduleNextReq() +{ + DPRINTF(DRAM, "Reached scheduleNextReq()\n"); + + // Figure out which request goes next, and move it to front() + if (!chooseNextReq()) + return; + + doDRAMAccess(dramReadQueue.front()); +} + + + + +Tick +SimpleDRAM::maxBankFreeAt() const +{ + Tick banksFree = 0; + + for(int i = 0; i < ranksPerChannel; i++) + for(int j = 0; j < banksPerRank; j++) + banksFree = std::max(banks[i][j].freeAt, banksFree); + + return banksFree; +} + +void +SimpleDRAM::processRefreshEvent() +{ + DPRINTF(DRAM, "Refreshing at tick %ld\n", curTick()); + + Tick banksFree = std::max(curTick(), maxBankFreeAt()) + tRFC; + + for(int i = 0; i < ranksPerChannel; i++) + for(int j = 0; j < banksPerRank; j++) + banks[i][j].freeAt = banksFree; + + schedule(&refreshEvent, curTick() + tREFI); +} + +void +SimpleDRAM::regStats() +{ + using namespace Stats; + + AbstractMemory::regStats(); + + readReqs + .name(name() + ".readReqs") + .desc("Total number of read requests seen"); + + writeReqs + .name(name() + ".writeReqs") + .desc("Total number of write requests seen"); + + servicedByWrQ + .name(name() + ".servicedByWrQ") + .desc("Number of read reqs serviced by write Q"); + + cpuReqs + .name(name() + ".cpureqs") + .desc("Reqs generatd by CPU via cache - shady"); + + neitherReadNorWrite + .name(name() + ".neitherReadNorWrite") + .desc("Reqs where no action is needed"); + + perBankRdReqs + .init(banksPerRank * ranksPerChannel) + .name(name() + ".perBankRdReqs") + .desc("Track reads on a per bank basis"); + + perBankWrReqs + .init(banksPerRank * ranksPerChannel) + .name(name() + ".perBankWrReqs") + .desc("Track writes on a per bank basis"); + + avgRdQLen + .name(name() + ".avgRdQLen") + .desc("Average read queue length over time") + .precision(2); + + avgWrQLen + .name(name() + ".avgWrQLen") + .desc("Average write queue length over time") + .precision(2); + + totQLat + .name(name() + ".totQLat") + .desc("Total cycles spent in queuing delays"); + + totBankLat + .name(name() + ".totBankLat") + .desc("Total cycles spent in bank access"); + + totBusLat + .name(name() + ".totBusLat") + .desc("Total cycles spent in databus access"); + + totMemAccLat + .name(name() + ".totMemAccLat") + .desc("Sum of mem lat for all requests"); + + avgQLat + .name(name() + ".avgQLat") + .desc("Average queueing delay per request") + .precision(2); + + avgQLat = totQLat / (readReqs - servicedByWrQ); + + avgBankLat + .name(name() + ".avgBankLat") + .desc("Average bank access latency per request") + .precision(2); + + avgBankLat = totBankLat / (readReqs - servicedByWrQ); + + avgBusLat + .name(name() + ".avgBusLat") + .desc("Average bus latency per request") + .precision(2); + + avgBusLat = totBusLat / (readReqs - servicedByWrQ); + + avgMemAccLat + .name(name() + ".avgMemAccLat") + .desc("Average memory access latency") + .precision(2); + + avgMemAccLat = totMemAccLat / (readReqs - servicedByWrQ); + + numRdRetry + .name(name() + ".numRdRetry") + .desc("Number of times rd buffer was full causing retry"); + + numWrRetry + .name(name() + ".numWrRetry") + .desc("Number of times wr buffer was full causing retry"); + + readRowHits + .name(name() + ".readRowHits") + .desc("Number of row buffer hits during reads"); + + writeRowHits + .name(name() + ".writeRowHits") + .desc("Number of row buffer hits during writes"); + + readRowHitRate + .name(name() + ".readRowHitRate") + .desc("Row buffer hit rate for reads") + .precision(2); + + readRowHitRate = (readRowHits / (readReqs - servicedByWrQ)) * 100; + + writeRowHitRate + .name(name() + ".writeRowHitRate") + .desc("Row buffer hit rate for writes") + .precision(2); + + writeRowHitRate = (writeRowHits / writeReqs) * 100; + + readPktSize + .init(log2(bytesPerCacheLine)+3) + .name(name() + ".readPktSize") + .desc("Categorize read packet sizes"); + + writePktSize + .init(log2(bytesPerCacheLine)+3) + .name(name() + ".writePktSize") + .desc("categorize write packet sizes"); + + neitherPktSize + .init(log2(bytesPerCacheLine)+3) + .name(name() + ".neitherpktsize") + .desc("categorize neither packet sizes"); + + rdQLenPdf + .init(readBufferSize + 1) + .name(name() + ".rdQLenPdf") + .desc("What read queue length does an incoming req see"); + + wrQLenPdf + .init(writeBufferSize + 1) + .name(name() + ".wrQLenPdf") + .desc("What write queue length does an incoming req see"); + + + bytesRead + .name(name() + ".bytesRead") + .desc("Total number of bytes read from memory"); + + bytesWritten + .name(name() + ".bytesWritten") + .desc("Total number of bytes written to memory"); + + bytesConsumedRd + .name(name() + ".bytesConsumedRd") + .desc("bytesRead derated as per pkt->getSize()"); + + bytesConsumedWr + .name(name() + ".bytesConsumedWr") + .desc("bytesWritten derated as per pkt->getSize()"); + + avgRdBW + .name(name() + ".avgRdBW") + .desc("Average achieved read bandwidth in MB/s") + .precision(2); + + avgRdBW = (bytesRead / 1000000) / simSeconds; + + avgWrBW + .name(name() + ".avgWrBW") + .desc("Average achieved write bandwidth in MB/s") + .precision(2); + + avgWrBW = (bytesWritten / 1000000) / simSeconds; + + avgConsumedRdBW + .name(name() + ".avgConsumedRdBW") + .desc("Average consumed read bandwidth in MB/s") + .precision(2); + + avgConsumedRdBW = (bytesConsumedRd / 1000000) / simSeconds; + + avgConsumedWrBW + .name(name() + ".avgConsumedWrBW") + .desc("Average consumed write bandwidth in MB/s") + .precision(2); + + avgConsumedWrBW = (bytesConsumedWr / 1000000) / simSeconds; + + peakBW + .name(name() + ".peakBW") + .desc("Theoretical peak bandwidth in MB/s") + .precision(2); + + peakBW = (SimClock::Frequency / tBURST) * bytesPerCacheLine / 1000000; + + busUtil + .name(name() + ".busUtil") + .desc("Data bus utilization in percentage") + .precision(2); + + busUtil = (avgRdBW + avgWrBW) / peakBW * 100; + + totGap + .name(name() + ".totGap") + .desc("Total gap between requests"); + + avgGap + .name(name() + ".avgGap") + .desc("Average gap between requests") + .precision(2); + + avgGap = totGap / (readReqs + writeReqs); +} + +void +SimpleDRAM::recvFunctional(PacketPtr pkt) +{ + // rely on the abstract memory + functionalAccess(pkt); +} + +SlavePort& +SimpleDRAM::getSlavePort(const string &if_name, int idx) +{ + if (if_name != "port") { + return MemObject::getSlavePort(if_name, idx); + } else { + return port; + } +} + +unsigned int +SimpleDRAM::drain(Event *de) +{ + unsigned int count = port.drain(de); + + // if there is anything in any of our internal queues, keep track + // of that as well + if (!(dramWriteQueue.empty() && dramReadQueue.empty() && + dramRespQueue.empty())) { + ++count; + drainEvent = de; + } + + if (count) + changeState(Draining); + else + changeState(Drained); + return count; +} + +SimpleDRAM::MemoryPort::MemoryPort(const std::string& name, SimpleDRAM& _memory) + : QueuedSlavePort(name, &_memory, queue), queue(_memory, *this), + memory(_memory) +{ } + +AddrRangeList +SimpleDRAM::MemoryPort::getAddrRanges() const +{ + AddrRangeList ranges; + ranges.push_back(memory.getAddrRange()); + return ranges; +} + +void +SimpleDRAM::MemoryPort::recvFunctional(PacketPtr pkt) +{ + pkt->pushLabel(memory.name()); + + if (!queue.checkFunctional(pkt)) { + // Default implementation of SimpleTimingPort::recvFunctional() + // calls recvAtomic() and throws away the latency; we can save a + // little here by just not calculating the latency. + memory.recvFunctional(pkt); + } + + pkt->popLabel(); +} + +Tick +SimpleDRAM::MemoryPort::recvAtomic(PacketPtr pkt) +{ + return memory.recvAtomic(pkt); +} + +bool +SimpleDRAM::MemoryPort::recvTimingReq(PacketPtr pkt) +{ + // pass it to the memory controller + return memory.recvTimingReq(pkt); +} + +SimpleDRAM* +SimpleDRAMParams::create() +{ + return new SimpleDRAM(this); +} diff --git a/src/mem/simple_dram.hh b/src/mem/simple_dram.hh new file mode 100644 index 000000000..c8e37ee18 --- /dev/null +++ b/src/mem/simple_dram.hh @@ -0,0 +1,478 @@ +/* + * Copyright (c) 2012 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. + * + * 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: Andreas Hansson + * Ani Udipi + */ + +/** + * @file + * SimpleDRAM declaration + */ + +#ifndef __MEM_SIMPLE_DRAM_HH__ +#define __MEM_SIMPLE_DRAM_HH__ + +#include "base/statistics.hh" +#include "enums/AddrMap.hh" +#include "enums/MemSched.hh" +#include "enums/PageManage.hh" +#include "mem/abstract_mem.hh" +#include "mem/qport.hh" +#include "params/SimpleDRAM.hh" +#include "sim/eventq.hh" + +/** + * The simple DRAM is a basic single-channel memory controller aiming + * to mimic a high-level DRAM controller and the most important timing + * constraints associated with the DRAM. The focus is really on + * modelling the impact on the system rather than the DRAM itself, + * hence the focus is on the controller model and not on the + * memory. By adhering to the correct timing constraints, ultimately + * there is no need for a memory model in addition to the controller + * model. + * + * As a basic design principle, this controller is not cycle callable, + * but instead uses events to decide when new decisions can be made, + * when resources become available, when things are to be considered + * done, and when to send things back. Through these simple + * principles, we achieve a performant model that is not + * cycle-accurate, but enables us to evaluate the system impact of a + * wide range of memory technologies, and also collect statistics + * about the use of the memory. + */ +class SimpleDRAM : public AbstractMemory +{ + + private: + + // For now, make use of a queued slave port to avoid dealing with + // flow control for the responses being sent back + class MemoryPort : public QueuedSlavePort + { + + SlavePacketQueue queue; + SimpleDRAM& memory; + + public: + + MemoryPort(const std::string& name, SimpleDRAM& _memory); + + protected: + + Tick recvAtomic(PacketPtr pkt); + + void recvFunctional(PacketPtr pkt); + + bool recvTimingReq(PacketPtr); + + virtual AddrRangeList getAddrRanges() const; + + }; + + /** + * Our incoming port, for a multi-ported controller add a crossbar + * in front of it + */ + MemoryPort port; + + /** + * Remember if we have to retry a request when available. + */ + bool retryRdReq; + bool retryWrReq; + + /** + * Remember that a row buffer hit occured + */ + bool rowHitFlag; + + /** + * Use this flag to shutoff reads, i.e. do not schedule any reads + * beyond those already done so that we can turn the bus around + * and do a few writes, or refresh, or whatever + */ + bool stopReads; + + /** + * A basic class to track the bank state indirectly via + * times "freeAt" and "tRASDoneAt" and what page is currently open + */ + class Bank + { + + public: + + static const uint32_t INVALID_ROW = -1; + + uint32_t openRow; + + Tick freeAt; + Tick tRASDoneAt; + + Bank() : openRow(INVALID_ROW), freeAt(0), tRASDoneAt(0) + { } + }; + + /** + * A DRAM packet stores packets along with the timestamp of when + * the packet entered the queue, and also the decoded address. + */ + class DRAMPacket { + + public: + + /** When did request enter the controller */ + const Tick entryTime; + + /** When will request leave the controller */ + Tick readyTime; + + /** This comes from the outside world */ + const PacketPtr pkt; + + /** Will be populated by address decoder */ + const uint8_t rank; + const uint16_t bank; + const uint16_t row; + const Addr addr; + Bank& bank_ref; + + DRAMPacket(PacketPtr _pkt, uint8_t _rank, + uint16_t _bank, uint16_t _row, Addr _addr, Bank& _bank_ref) + : entryTime(curTick()), readyTime(curTick()), + pkt(_pkt), rank(_rank), bank(_bank), row(_row), addr(_addr), + bank_ref(_bank_ref) + { } + + }; + + /** + * Bunch of things requires to setup "events" in gem5 + * When event "writeEvent" occurs for example, the method + * processWriteEvent is called; no parameters are allowed + * in these methods + */ + void processWriteEvent(); + EventWrapper writeEvent; + + void processRespondEvent(); + EventWrapper respondEvent; + + void processRefreshEvent(); + EventWrapper refreshEvent; + + void processNextReqEvent(); + EventWrapper nextReqEvent; + + + /** + * Check if the read queue has room for more entries + * + * @return true if read queue is full, false otherwise + */ + bool readQueueFull() const; + + /** + * Check if the write queue has room for more entries + * + * @return true if write queue is full, false otherwise + */ + bool writeQueueFull() const; + + /** + * When a new read comes in, first check if the write q has a + * pending request to the same address.\ If not, decode the + * address to populate rank/bank/row, create a "dram_pkt", and + * push it to the back of the read queue.\ If this is the only + * read request in the system, schedule an event to start + * servicing it. + * + * @param pkt The request packet from the outside world + */ + void addToReadQueue(PacketPtr pkt); + + /** + * Decode the incoming pkt, create a dram_pkt and push to the + * back of the write queue. \If the write q length is more than + * the threshold specified by the user, ie the queue is beginning + * to get full, stop reads, and start draining writes. + * + * @param pkt The request packet from the outside world + */ + void addToWriteQueue(PacketPtr pkt); + + /** + * Actually do the DRAM access - figure out the latency it + * will take to service the req based on bank state, channel state etc + * and then update those states to account for this request.\ Based + * on this, update the packet's "readyTime" and move it to the + * response q from where it will eventually go back to the outside + * world. + * + * @param pkt The DRAM packet created from the outside world pkt + */ + void doDRAMAccess(DRAMPacket* dram_pkt); + + /** + * Check when the channel is free to turnaround, add turnaround + * delay and schedule a whole bunch of writes. + */ + void triggerWrites(); + + /** + * When a packet reaches its "readyTime" in the response Q, + * use the "access()" method in AbstractMemory to actually + * create the response packet, and send it back to the outside + * world requestor. + * + * @param pkt The packet from the outside world + */ + void accessAndRespond(PacketPtr pkt); + + /** + * Address decoder to figure out physical mapping onto ranks, + * banks, and rows. + * + * @param pkt The packet from the outside world + * @return A DRAMPacket pointer with the decoded information + */ + DRAMPacket* decodeAddr(PacketPtr pkt); + + /** + * The memory schduler/arbiter - picks which request needs to + * go next, based on the specified policy such as fcfs or frfcfs + * and moves it to the head of the read queue + * + * @return True if a request was chosen, False if Q is empty + */ + bool chooseNextReq(); + + /** + * Calls chooseNextReq() to pick the right request, then calls + * doDRAMAccess on that request in order to actually service + * that request + */ + void scheduleNextReq(); + + /** + *Looks at the state of the banks, channels, row buffer hits etc + * to estimate how long a request will take to complete. + * + * @param dram_pkt The request for which we want to estimate latency + * @param inTime The tick at which you want to probe the memory + * + * @return A pair of ticks, one indicating how many ticks *after* + * inTime the request require, and the other indicating how + * much of that was just the bank access time, ignoring the + * ticks spent simply waiting for resources to become free + */ + std::pair estimateLatency(DRAMPacket* dram_pkt, Tick inTime); + + /** + * Move the request at the head of the read queue to the response + * queue, sorting by readyTime.\ If it is the only packet in the + * response queue, schedule a respond event to send it back to the + * outside world + */ + void moveToRespQ(); + + /** + * Scheduling policy within the write Q + */ + void chooseNextWrite(); + + /** + * Looking at all banks, determine the moment in time when they + * are all free. + * + * @return The tick when all banks are free + */ + Tick maxBankFreeAt() const; + + void printParams() const; + void printQs() const; + + /** + * The controller's main read and write queues + */ + std::list dramReadQueue; + std::list dramWriteQueue; + + /** + * Response queue where read packets wait after we're done working + * with them, but it's not time to send the response yet.\ It is + * seperate mostly to keep the code clean and help with gem5 events, + * but for all logical purposes such as sizing the read queue, this + * and the main read queue need to be added together. + */ + std::list dramRespQueue; + + /** If we need to drain, keep the drain event around until we're done + * here. + */ + Event *drainEvent; + + /** + * Multi-dimensional vector of banks, first dimension is ranks, + * second is bank + */ + std::vector > banks; + + /** + * The following are basic design parameters of the memory + * controller, and are initialized based on parameter values. The + * bytesPerCacheLine is based on the neighbouring port and thus + * determined outside the constructor. Similarly, the rowsPerBank + * is determined based on the capacity, number of ranks and banks, + * the cache line size, and the row buffer size. + */ + uint32_t bytesPerCacheLine; + const uint32_t linesPerRowBuffer; + const uint32_t ranksPerChannel; + const uint32_t banksPerRank; + uint32_t rowsPerBank; + const uint32_t readBufferSize; + const uint32_t writeBufferSize; + const double writeThresholdPerc; + uint32_t writeThreshold; + + /** + * Basic memory timing parameters initialized based on parameter + * values. + */ + const Tick tWTR; + const Tick tBURST; + const Tick tRCD; + const Tick tCL; + const Tick tRP; + const Tick tRFC; + const Tick tREFI; + + /** + * Memory controller configuration initialized based on parameter + * values. + */ + Enums::MemSched memSchedPolicy; + Enums::AddrMap addrMapping; + Enums::PageManage pageMgmt; + + /** + * Till when has the main data bus been spoken for already? + */ + Tick busBusyUntil; + + Tick prevdramaccess; + Tick writeStartTime; + Tick prevArrival; + int numReqs; + + // All statistics that the model needs to capture + Stats::Scalar readReqs; + Stats::Scalar writeReqs; + Stats::Scalar cpuReqs; + Stats::Scalar bytesRead; + Stats::Scalar bytesWritten; + Stats::Scalar bytesConsumedRd; + Stats::Scalar bytesConsumedWr; + Stats::Scalar servicedByWrQ; + Stats::Scalar neitherReadNorWrite; + Stats::Vector perBankRdReqs; + Stats::Vector perBankWrReqs; + Stats::Scalar numRdRetry; + Stats::Scalar numWrRetry; + Stats::Scalar totGap; + Stats::Vector readPktSize; + Stats::Vector writePktSize; + Stats::Vector neitherPktSize; + Stats::Vector rdQLenPdf; + Stats::Vector wrQLenPdf; + + + // Latencies summed over all requests + Stats::Scalar totQLat; + Stats::Scalar totMemAccLat; + Stats::Scalar totBusLat; + Stats::Scalar totBankLat; + + // Average latencies per request + Stats::Formula avgQLat; + Stats::Formula avgBankLat; + Stats::Formula avgBusLat; + Stats::Formula avgMemAccLat; + + // Average bandwidth + Stats::Formula avgRdBW; + Stats::Formula avgWrBW; + Stats::Formula avgConsumedRdBW; + Stats::Formula avgConsumedWrBW; + Stats::Formula peakBW; + Stats::Formula busUtil; + + // Average queue lengths + Stats::Average avgRdQLen; + Stats::Average avgWrQLen; + + // Row hit count and rate + Stats::Scalar readRowHits; + Stats::Scalar writeRowHits; + Stats::Formula readRowHitRate; + Stats::Formula writeRowHitRate; + Stats::Formula avgGap; + + public: + + void regStats(); + + SimpleDRAM(const SimpleDRAMParams* p); + + unsigned int drain(Event* de); + + virtual SlavePort& getSlavePort(const std::string& if_name, + int idx = InvalidPortID); + + virtual void init(); + virtual void startup(); + + protected: + + Tick recvAtomic(PacketPtr pkt); + void recvFunctional(PacketPtr pkt); + bool recvTimingReq(PacketPtr pkt); + +}; + +#endif //__MEM_SIMPLE_DRAM_HH__ -- cgit v1.2.3