This patch changes the cache-related latencies from an absolute time expressed in Ticks, to a number of cycles that can be scaled with the clock period of the caches. Ultimately this patch serves to enable future work that involves dynamic frequency scaling. As an immediate benefit it also makes it more convenient to specify cache performance without implicitly assuming a specific CPU core operating frequency. The stat blocked_cycles that actually counter in ticks is now updated to count in cycles. As the timing is now rounded to the clock edges of the cache, there are some regressions that change. Plenty of them have very minor changes, whereas some regressions with a short run-time are perturbed quite significantly. A follow-on patch updates all the statistics for the regressions.
284 lines
8.7 KiB
C++
284 lines
8.7 KiB
C++
/*
|
|
* Copyright (c) 2005 The Regents of The University of Michigan
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions are
|
|
* met: redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer;
|
|
* redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution;
|
|
* neither the name of the copyright holders nor the names of its
|
|
* contributors may be used to endorse or promote products derived from
|
|
* this software without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
|
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
|
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
|
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
|
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
|
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
|
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*
|
|
* Authors: Ron Dreslinski
|
|
*/
|
|
|
|
/**
|
|
* @file
|
|
* Hardware Prefetcher Definition.
|
|
*/
|
|
|
|
#include <list>
|
|
|
|
#include "arch/isa_traits.hh"
|
|
#include "base/trace.hh"
|
|
#include "config/the_isa.hh"
|
|
#include "debug/HWPrefetch.hh"
|
|
#include "mem/cache/prefetch/base.hh"
|
|
#include "mem/cache/base.hh"
|
|
#include "mem/request.hh"
|
|
#include "sim/system.hh"
|
|
|
|
BasePrefetcher::BasePrefetcher(const Params *p)
|
|
: ClockedObject(p), size(p->size), latency(p->latency), degree(p->degree),
|
|
useMasterId(p->use_master_id), pageStop(!p->cross_pages),
|
|
serialSquash(p->serial_squash), onlyData(p->data_accesses_only),
|
|
system(p->sys), masterId(system->getMasterId(name()))
|
|
{
|
|
}
|
|
|
|
void
|
|
BasePrefetcher::setCache(BaseCache *_cache)
|
|
{
|
|
cache = _cache;
|
|
blkSize = cache->getBlockSize();
|
|
}
|
|
|
|
void
|
|
BasePrefetcher::regStats()
|
|
{
|
|
pfIdentified
|
|
.name(name() + ".prefetcher.num_hwpf_identified")
|
|
.desc("number of hwpf identified")
|
|
;
|
|
|
|
pfMSHRHit
|
|
.name(name() + ".prefetcher.num_hwpf_already_in_mshr")
|
|
.desc("number of hwpf that were already in mshr")
|
|
;
|
|
|
|
pfCacheHit
|
|
.name(name() + ".prefetcher.num_hwpf_already_in_cache")
|
|
.desc("number of hwpf that were already in the cache")
|
|
;
|
|
|
|
pfBufferHit
|
|
.name(name() + ".prefetcher.num_hwpf_already_in_prefetcher")
|
|
.desc("number of hwpf that were already in the prefetch queue")
|
|
;
|
|
|
|
pfRemovedFull
|
|
.name(name() + ".prefetcher.num_hwpf_evicted")
|
|
.desc("number of hwpf removed due to no buffer left")
|
|
;
|
|
|
|
pfRemovedMSHR
|
|
.name(name() + ".prefetcher.num_hwpf_removed_MSHR_hit")
|
|
.desc("number of hwpf removed because MSHR allocated")
|
|
;
|
|
|
|
pfIssued
|
|
.name(name() + ".prefetcher.num_hwpf_issued")
|
|
.desc("number of hwpf issued")
|
|
;
|
|
|
|
pfSpanPage
|
|
.name(name() + ".prefetcher.num_hwpf_span_page")
|
|
.desc("number of hwpf spanning a virtual page")
|
|
;
|
|
|
|
pfSquashed
|
|
.name(name() + ".prefetcher.num_hwpf_squashed_from_miss")
|
|
.desc("number of hwpf that got squashed due to a miss "
|
|
"aborting calculation time")
|
|
;
|
|
}
|
|
|
|
inline bool
|
|
BasePrefetcher::inCache(Addr addr)
|
|
{
|
|
if (cache->inCache(addr)) {
|
|
pfCacheHit++;
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
inline bool
|
|
BasePrefetcher::inMissQueue(Addr addr)
|
|
{
|
|
if (cache->inMissQueue(addr)) {
|
|
pfMSHRHit++;
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
PacketPtr
|
|
BasePrefetcher::getPacket()
|
|
{
|
|
DPRINTF(HWPrefetch, "Requesting a hw_pf to issue\n");
|
|
|
|
if (pf.empty()) {
|
|
DPRINTF(HWPrefetch, "No HW_PF found\n");
|
|
return NULL;
|
|
}
|
|
|
|
PacketPtr pkt = *pf.begin();
|
|
while (!pf.empty()) {
|
|
pkt = *pf.begin();
|
|
pf.pop_front();
|
|
|
|
Addr blk_addr = pkt->getAddr() & ~(Addr)(blkSize-1);
|
|
|
|
if (!inCache(blk_addr) && !inMissQueue(blk_addr))
|
|
// we found a prefetch, return it
|
|
break;
|
|
|
|
DPRINTF(HWPrefetch, "addr 0x%x in cache, skipping\n", pkt->getAddr());
|
|
delete pkt->req;
|
|
delete pkt;
|
|
|
|
if (pf.empty()) {
|
|
cache->deassertMemSideBusRequest(BaseCache::Request_PF);
|
|
return NULL; // None left, all were in cache
|
|
}
|
|
}
|
|
|
|
pfIssued++;
|
|
assert(pkt != NULL);
|
|
DPRINTF(HWPrefetch, "returning 0x%x\n", pkt->getAddr());
|
|
return pkt;
|
|
}
|
|
|
|
|
|
Tick
|
|
BasePrefetcher::notify(PacketPtr &pkt, Tick time)
|
|
{
|
|
if (!pkt->req->isUncacheable() && !(pkt->req->isInstFetch() && onlyData)) {
|
|
// Calculate the blk address
|
|
Addr blk_addr = pkt->getAddr() & ~(Addr)(blkSize-1);
|
|
|
|
// Check if miss is in pfq, if so remove it
|
|
std::list<PacketPtr>::iterator iter = inPrefetch(blk_addr);
|
|
if (iter != pf.end()) {
|
|
DPRINTF(HWPrefetch, "Saw a miss to a queued prefetch addr: "
|
|
"0x%x, removing it\n", blk_addr);
|
|
pfRemovedMSHR++;
|
|
delete (*iter)->req;
|
|
delete (*iter);
|
|
iter = pf.erase(iter);
|
|
if (pf.empty())
|
|
cache->deassertMemSideBusRequest(BaseCache::Request_PF);
|
|
}
|
|
|
|
// Remove anything in queue with delay older than time
|
|
// since everything is inserted in time order, start from end
|
|
// and work until pf.empty() or time is earlier
|
|
// This is done to emulate Aborting the previous work on a new miss
|
|
// Needed for serial calculators like GHB
|
|
if (serialSquash) {
|
|
iter = pf.end();
|
|
if (iter != pf.begin())
|
|
iter--;
|
|
while (!pf.empty() && ((*iter)->time >= time)) {
|
|
pfSquashed++;
|
|
DPRINTF(HWPrefetch, "Squashing old prefetch addr: 0x%x\n",
|
|
(*iter)->getAddr());
|
|
delete (*iter)->req;
|
|
delete (*iter);
|
|
iter = pf.erase(iter);
|
|
if (iter != pf.begin())
|
|
iter--;
|
|
}
|
|
if (pf.empty())
|
|
cache->deassertMemSideBusRequest(BaseCache::Request_PF);
|
|
}
|
|
|
|
|
|
std::list<Addr> addresses;
|
|
std::list<Cycles> delays;
|
|
calculatePrefetch(pkt, addresses, delays);
|
|
|
|
std::list<Addr>::iterator addrIter = addresses.begin();
|
|
std::list<Cycles>::iterator delayIter = delays.begin();
|
|
for (; addrIter != addresses.end(); ++addrIter, ++delayIter) {
|
|
Addr addr = *addrIter;
|
|
|
|
pfIdentified++;
|
|
|
|
DPRINTF(HWPrefetch, "Found a pf candidate addr: 0x%x, "
|
|
"inserting into prefetch queue with delay %d time %d\n",
|
|
addr, *delayIter, time);
|
|
|
|
// Check if it is already in the pf buffer
|
|
if (inPrefetch(addr) != pf.end()) {
|
|
pfBufferHit++;
|
|
DPRINTF(HWPrefetch, "Prefetch addr already in pf buffer\n");
|
|
continue;
|
|
}
|
|
|
|
// create a prefetch memreq
|
|
Request *prefetchReq = new Request(*addrIter, blkSize, 0, masterId);
|
|
PacketPtr prefetch =
|
|
new Packet(prefetchReq, MemCmd::HardPFReq);
|
|
prefetch->allocate();
|
|
prefetch->req->setThreadContext(pkt->req->contextId(),
|
|
pkt->req->threadId());
|
|
|
|
prefetch->time = time + clock * *delayIter;
|
|
|
|
// We just remove the head if we are full
|
|
if (pf.size() == size) {
|
|
pfRemovedFull++;
|
|
PacketPtr old_pkt = *pf.begin();
|
|
DPRINTF(HWPrefetch, "Prefetch queue full, "
|
|
"removing oldest 0x%x\n", old_pkt->getAddr());
|
|
delete old_pkt->req;
|
|
delete old_pkt;
|
|
pf.pop_front();
|
|
}
|
|
|
|
pf.push_back(prefetch);
|
|
}
|
|
}
|
|
|
|
return pf.empty() ? 0 : pf.front()->time;
|
|
}
|
|
|
|
std::list<PacketPtr>::iterator
|
|
BasePrefetcher::inPrefetch(Addr address)
|
|
{
|
|
// Guaranteed to only be one match, we always check before inserting
|
|
std::list<PacketPtr>::iterator iter;
|
|
for (iter = pf.begin(); iter != pf.end(); iter++) {
|
|
if (((*iter)->getAddr() & ~(Addr)(blkSize-1)) == address) {
|
|
return iter;
|
|
}
|
|
}
|
|
return pf.end();
|
|
}
|
|
|
|
bool
|
|
BasePrefetcher::samePage(Addr a, Addr b)
|
|
{
|
|
return roundDown(a, TheISA::VMPageSize) == roundDown(b, TheISA::VMPageSize);
|
|
}
|
|
|
|
|