sanchayanmaity/gem5
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Mem: Use cycles to express cache-related latencies

Browse source 
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.
This commit is contained in:
Andreas Hansson 2012-10-15 08:10:54 -04:00
parent d17f5084ed
commit 88554790c3
48 changed files with 330 additions and 259 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
configs/common/Caches.py
View file

@ -1,3 +1,15 @@
# 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.
#
# Copyright (c) 2006-2007 The Regents of The University of Michigan
# All rights reserved.
#
@ -31,8 +43,8 @@ from m5.objects import *
class L1Cache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
tgts_per_mshr = 20
is_top_level = True
@ -40,16 +52,16 @@ class L1Cache(BaseCache):
class L2Cache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 20
tgts_per_mshr = 12
class PageTableWalkerCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
size = '1kB'
tgts_per_mshr = 12
@ -58,8 +70,8 @@ class PageTableWalkerCache(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12

21
configs/common/O3_ARM_v7a.py
View file

@ -145,10 +145,9 @@ class O3_ARM_v7a_3(DerivO3CPU):
defer_registration= False
# Instruction Cache
# All latencys assume a 1GHz clock rate, with a faster clock they would be faster
class O3_ARM_v7a_ICache(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 1
response_latency = 1
block_size = 64
mshrs = 2
tgts_per_mshr = 8
@ -157,10 +156,9 @@ class O3_ARM_v7a_ICache(BaseCache):
is_top_level = 'true'
# Data Cache
# All latencys assume a 1GHz clock rate, with a faster clock they would be faster
class O3_ARM_v7a_DCache(BaseCache):
hit_latency = '2ns'
response_latency = '2ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 6
tgts_per_mshr = 8
@ -172,8 +170,8 @@ class O3_ARM_v7a_DCache(BaseCache):
# TLB Cache
# Use a cache as a L2 TLB
class O3_ARM_v7aWalkCache(BaseCache):
hit_latency = '4ns'
response_latency = '4ns'
hit_latency = 4
response_latency = 4
block_size = 64
mshrs = 6
tgts_per_mshr = 8
@ -184,10 +182,9 @@ class O3_ARM_v7aWalkCache(BaseCache):
# L2 Cache
# All latencys assume a 1GHz clock rate, with a faster clock they would be faster
class O3_ARM_v7aL2(BaseCache):
hit_latency = '12ns'
response_latency = '12ns'
hit_latency = 12
response_latency = 12
block_size = 64
mshrs = 16
tgts_per_mshr = 8
@ -196,5 +193,5 @@ class O3_ARM_v7aL2(BaseCache):
write_buffers = 8
prefetch_on_access = 'true'
# Simple stride prefetcher
prefetcher = StridePrefetcher(degree=8, latency='1.0ns')
prefetcher = StridePrefetcher(degree=8, latency = 1)

3
configs/example/fs.py
View file

@ -122,7 +122,8 @@ if bm[0]:
else:
mem_size = SysConfig().mem()
if options.caches or options.l2cache:
test_sys.iocache = IOCache(addr_ranges=[test_sys.physmem.range])
test_sys.iocache = IOCache(clock = '1GHz',
addr_ranges=[test_sys.physmem.range])
test_sys.iocache.cpu_side = test_sys.iobus.master
test_sys.iocache.mem_side = test_sys.membus.slave
else:

16
src/mem/cache/BaseCache.py vendored
View file

@ -1,3 +1,15 @@
# 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.
#
# Copyright (c) 2005-2007 The Regents of The University of Michigan
# All rights reserved.
#
@ -36,8 +48,8 @@ class BaseCache(MemObject):
type = 'BaseCache'
assoc = Param.Int("associativity")
block_size = Param.Int("block size in bytes")
hit_latency = Param.Latency("The hit latency for this cache")
response_latency = Param.Latency(
hit_latency = Param.Cycles("The hit latency for this cache")
response_latency = Param.Cycles(
"Additional cache latency for the return path to core on a miss");
hash_delay = Param.Cycles(1, "time in cycles of hash access")
max_miss_count = Param.Counter(0,

4
src/mem/cache/base.cc vendored
View file

@ -81,6 +81,10 @@ BaseCache::BaseCache(const Params *p)
addrRanges(p->addr_ranges.begin(), p->addr_ranges.end()),
system(p->system)
{
// ensure the clock is not running at an unreasonable clock speed
if (clock == 1)
panic("Cache %s has a cycle time of 1 tick. Specify a clock.\n",
name());
}
void

10
src/mem/cache/base.hh vendored
View file

@ -229,7 +229,7 @@ class BaseCache : public MemObject
/**
* The latency of a hit in this device.
*/
const Tick hitLatency;
const Cycles hitLatency;
/**
* The latency of sending reponse to its upper level cache/core on a
@ -237,7 +237,7 @@ class BaseCache : public MemObject
* miss is much quicker that the hit latency. The responseLatency parameter
* tries to capture this latency.
*/
const Tick responseLatency;
const Cycles responseLatency;
/** The number of targets for each MSHR. */
const int numTarget;
@ -260,7 +260,7 @@ class BaseCache : public MemObject
uint64_t order;
/** Stores time the cache blocked for statistics. */
Tick blockedCycle;
Cycles blockedCycle;
/** Pointer to the MSHR that has no targets. */
MSHR *noTargetMSHR;
@ -492,7 +492,7 @@ class BaseCache : public MemObject
uint8_t flag = 1 << cause;
if (blocked == 0) {
blocked_causes[cause]++;
blockedCycle = curTick();
blockedCycle = curCycle();
cpuSidePort->setBlocked();
}
blocked |= flag;
@ -512,7 +512,7 @@ class BaseCache : public MemObject
blocked &= ~flag;
DPRINTF(Cache,"Unblocking for cause %d, mask=%d\n", cause, blocked);
if (blocked == 0) {
blocked_cycles[cause] += curTick() - blockedCycle;
blocked_cycles[cause] += curCycle() - blockedCycle;
cpuSidePort->clearBlocked();
}
}

8
src/mem/cache/cache.hh vendored
View file

@ -206,7 +206,7 @@ class Cache : public BaseCache
* @return Boolean indicating whether the request was satisfied.
*/
bool access(PacketPtr pkt, BlkType *&blk,
int &lat, PacketList &writebacks);
Cycles &lat, PacketList &writebacks);
/**
*Handle doing the Compare and Swap function for SPARC.
@ -272,7 +272,7 @@ class Cache : public BaseCache
/**
* Performs the access specified by the request.
* @param pkt The request to perform.
* @return The result of the access.
* @return The number of ticks required for the access.
*/
Tick atomicAccess(PacketPtr pkt);
@ -299,9 +299,9 @@ class Cache : public BaseCache
* Snoop for the provided request in the cache and return the estimated
* time of completion.
* @param pkt The memory request to snoop
* @return The estimated completion time.
* @return The number of cycles required for the snoop.
*/
Tick snoopAtomic(PacketPtr pkt);
Cycles snoopAtomic(PacketPtr pkt);
/**
* Squash all requests associated with specified thread.

26
src/mem/cache/cache_impl.hh vendored
View file

@ -275,7 +275,7 @@ Cache<TagStore>::squash(int threadNum)
template<class TagStore>
bool
Cache<TagStore>::access(PacketPtr pkt, BlkType *&blk,
int &lat, PacketList &writebacks)
Cycles &lat, PacketList &writebacks)
{
if (pkt->req->isUncacheable()) {
if (pkt->req->isClearLL()) {
@ -392,7 +392,7 @@ Cache<TagStore>::timingAccess(PacketPtr pkt)
pendingDelete.clear();
// we charge hitLatency for doing just about anything here
Tick time = curTick() + hitLatency;
Tick time = clockEdge(hitLatency);
if (pkt->isResponse()) {
// must be cache-to-cache response from upper to lower level
@ -463,7 +463,7 @@ Cache<TagStore>::timingAccess(PacketPtr pkt)
return true;
}
int lat = hitLatency;
Cycles lat = hitLatency;
BlkType *blk = NULL;
PacketList writebacks;
@ -505,7 +505,7 @@ Cache<TagStore>::timingAccess(PacketPtr pkt)
if (needsResponse) {
pkt->makeTimingResponse();
cpuSidePort->schedTimingResp(pkt, curTick()+lat);
cpuSidePort->schedTimingResp(pkt, clockEdge(lat));
} else {
/// @todo nominally we should just delete the packet here,
/// however, until 4-phase stuff we can't because sending
@ -637,7 +637,7 @@ template<class TagStore>
Tick
Cache<TagStore>::atomicAccess(PacketPtr pkt)
{
int lat = hitLatency;
Cycles lat = hitLatency;
// @TODO: make this a parameter
bool last_level_cache = false;
@ -657,7 +657,7 @@ Cache<TagStore>::atomicAccess(PacketPtr pkt)
if (!last_level_cache) {
DPRINTF(Cache, "forwarding mem-inhibited %s on 0x%x\n",
pkt->cmdString(), pkt->getAddr());
lat += memSidePort->sendAtomic(pkt);
lat += ticksToCycles(memSidePort->sendAtomic(pkt));
}
} else {
DPRINTF(Cache, "rcvd mem-inhibited %s on 0x%x: not responding\n",
@ -693,7 +693,7 @@ Cache<TagStore>::atomicAccess(PacketPtr pkt)
CacheBlk::State old_state = blk ? blk->status : 0;
#endif
lat += memSidePort->sendAtomic(bus_pkt);
lat += ticksToCycles(memSidePort->sendAtomic(bus_pkt));
DPRINTF(Cache, "Receive response: %s for addr %x in state %i\n",
bus_pkt->cmdString(), bus_pkt->getAddr(), old_state);
@ -821,7 +821,7 @@ template<class TagStore>
void
Cache<TagStore>::handleResponse(PacketPtr pkt)
{
Tick time = curTick() + hitLatency;
Tick time = clockEdge(hitLatency);
MSHR *mshr = dynamic_cast<MSHR*>(pkt->senderState);
bool is_error = pkt->isError();
@ -901,7 +901,7 @@ Cache<TagStore>::handleResponse(PacketPtr pkt)
// responseLatency is the latency of the return path
// from lower level caches/memory to an upper level cache or
// the core.
completion_time = responseLatency +
completion_time = responseLatency * clock +
(transfer_offset ? pkt->finishTime : pkt->firstWordTime);
assert(!target->pkt->req->isUncacheable());
@ -917,13 +917,13 @@ Cache<TagStore>::handleResponse(PacketPtr pkt)
// responseLatency is the latency of the return path
// from lower level caches/memory to an upper level cache or
// the core.
completion_time = responseLatency + pkt->finishTime;
completion_time = responseLatency * clock + pkt->finishTime;
target->pkt->req->setExtraData(0);
} else {
// not a cache fill, just forwarding response
// responseLatency is the latency of the return path
// from lower level cahces/memory to the core.
completion_time = responseLatency + pkt->finishTime;
completion_time = responseLatency * clock + pkt->finishTime;
if (pkt->isRead() && !is_error) {
target->pkt->setData(pkt->getPtr<uint8_t>());
}
@ -1173,7 +1173,7 @@ doTimingSupplyResponse(PacketPtr req_pkt, uint8_t *blk_data,
// invalidate it.
pkt->cmd = MemCmd::ReadRespWithInvalidate;
}
memSidePort->schedTimingSnoopResp(pkt, curTick() + hitLatency);
memSidePort->schedTimingSnoopResp(pkt, clockEdge(hitLatency));
}
template<class TagStore>
@ -1366,7 +1366,7 @@ Cache<TagStore>::CpuSidePort::recvTimingSnoopResp(PacketPtr pkt)
}
template<class TagStore>
Tick
Cycles
Cache<TagStore>::snoopAtomic(PacketPtr pkt)
{
if (pkt->req->isUncacheable() || pkt->cmd == MemCmd::Writeback) {

47
src/mem/cache/prefetch/Prefetcher.py vendored
View file

@ -1,8 +1,48 @@
from m5.SimObject import SimObject
# 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.
#
# 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
from ClockedObject import ClockedObject
from m5.params import *
from m5.proxy import *
class BasePrefetcher(SimObject):
class BasePrefetcher(ClockedObject):
type = 'BasePrefetcher'
abstract = True
size = Param.Int(100,
@ -13,8 +53,7 @@ class BasePrefetcher(SimObject):
"Squash prefetches with a later time on a subsequent miss")
degree = Param.Int(1,
"Degree of the prefetch depth")
latency = Param.Latency('10t',
"Latency of the prefetcher")
latency = Param.Cycles('1', "Latency of the prefetcher")
use_master_id = Param.Bool(True,
"Use the master id to separate calculations of prefetches")
data_accesses_only = Param.Bool(False,

8
src/mem/cache/prefetch/base.cc vendored
View file

@ -45,7 +45,7 @@
#include "sim/system.hh"
BasePrefetcher::BasePrefetcher(const Params *p)
: SimObject(p), size(p->size), latency(p->latency), degree(p->degree),
: 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()))
@ -212,11 +212,11 @@ BasePrefetcher::notify(PacketPtr &pkt, Tick time)
std::list<Addr> addresses;
std::list<Tick> delays;
std::list<Cycles> delays;
calculatePrefetch(pkt, addresses, delays);
std::list<Addr>::iterator addrIter = addresses.begin();
std::list<Tick>::iterator delayIter = delays.begin();
std::list<Cycles>::iterator delayIter = delays.begin();
for (; addrIter != addresses.end(); ++addrIter, ++delayIter) {
Addr addr = *addrIter;
@ -241,7 +241,7 @@ BasePrefetcher::notify(PacketPtr &pkt, Tick time)
prefetch->req->setThreadContext(pkt->req->contextId(),
pkt->req->threadId());
prefetch->time = time + (*delayIter); // @todo ADD LATENCY HERE
prefetch->time = time + clock * *delayIter;
// We just remove the head if we are full
if (pf.size() == size) {

8
src/mem/cache/prefetch/base.hh vendored
View file

@ -41,11 +41,11 @@
#include "base/statistics.hh"
#include "mem/packet.hh"
#include "params/BaseCache.hh"
#include "sim/sim_object.hh"
#include "sim/clocked_object.hh"
class BaseCache;
class BasePrefetcher : public SimObject
class BasePrefetcher : public ClockedObject
{
protected:
@ -64,7 +64,7 @@ class BasePrefetcher : public SimObject
int blkSize;
/** The latency before a prefetch is issued */
Tick latency;
const Cycles latency;
/** The number of prefetches to issue */
unsigned degree;
@ -133,7 +133,7 @@ class BasePrefetcher : public SimObject
virtual void calculatePrefetch(PacketPtr &pkt,
std::list<Addr> &addresses,
std::list<Tick> &delays) = 0;
std::list<Cycles> &delays) = 0;
std::list<PacketPtr>::iterator inPrefetch(Addr address);

2
src/mem/cache/prefetch/ghb.cc vendored
View file

@ -40,7 +40,7 @@
void
GHBPrefetcher::calculatePrefetch(PacketPtr &pkt, std::list<Addr> &addresses,
std::list<Tick> &delays)
std::list<Cycles> &delays)
{
Addr blk_addr = pkt->getAddr() & ~(Addr)(blkSize-1);
int master_id = useMasterId ? pkt->req->masterId() : 0;

2
src/mem/cache/prefetch/ghb.hh vendored
View file

@ -57,7 +57,7 @@ class GHBPrefetcher : public BasePrefetcher
~GHBPrefetcher() {}
void calculatePrefetch(PacketPtr &pkt, std::list<Addr> &addresses,
std::list<Tick> &delays);
std::list<Cycles> &delays);
};
#endif // __MEM_CACHE_PREFETCH_GHB_PREFETCHER_HH__

2
src/mem/cache/prefetch/stride.cc vendored
View file

@ -40,7 +40,7 @@
void
StridePrefetcher::calculatePrefetch(PacketPtr &pkt, std::list<Addr> &addresses,
std::list<Tick> &delays)
std::list<Cycles> &delays)
{
if (!pkt->req->hasPC()) {
DPRINTF(HWPrefetch, "ignoring request with no PC");

2
src/mem/cache/prefetch/stride.hh vendored
View file

@ -75,7 +75,7 @@ class StridePrefetcher : public BasePrefetcher
~StridePrefetcher() {}
void calculatePrefetch(PacketPtr &pkt, std::list<Addr> &addresses,
std::list<Tick> &delays);
std::list<Cycles> &delays);
};
#endif // __MEM_CACHE_PREFETCH_STRIDE_PREFETCHER_HH__

2
src/mem/cache/prefetch/tagged.cc vendored
View file

@ -43,7 +43,7 @@ TaggedPrefetcher::TaggedPrefetcher(const Params *p)
void
TaggedPrefetcher::
calculatePrefetch(PacketPtr &pkt, std::list<Addr> &addresses,
std::list<Tick> &delays)
std::list<Cycles> &delays)
{
Addr blkAddr = pkt->getAddr() & ~(Addr)(blkSize-1);

2
src/mem/cache/prefetch/tagged.hh vendored
View file

@ -49,7 +49,7 @@ class TaggedPrefetcher : public BasePrefetcher
~TaggedPrefetcher() {}
void calculatePrefetch(PacketPtr &pkt, std::list<Addr> &addresses,
std::list<Tick> &delays);
std::list<Cycles> &delays);
};
#endif // __MEM_CACHE_PREFETCH_TAGGED_PREFETCHER_HH__

4
src/mem/cache/tags/fa_lru.cc vendored
View file

@ -42,7 +42,7 @@
using namespace std;
FALRU::FALRU(unsigned _blkSize, unsigned _size, unsigned hit_latency)
FALRU::FALRU(unsigned _blkSize, unsigned _size, Cycles hit_latency)
: blkSize(_blkSize), size(_size), hitLatency(hit_latency)
{
if (!isPowerOf2(blkSize))
@ -159,7 +159,7 @@ FALRU::invalidate(FALRU::BlkType *blk)
}
FALRUBlk*
FALRU::accessBlock(Addr addr, int &lat, int context_src, int *inCache)
FALRU::accessBlock(Addr addr, Cycles &lat, int context_src, int *inCache)
{
accesses++;
int tmp_in_cache = 0;

8
src/mem/cache/tags/fa_lru.hh vendored
View file

@ -85,7 +85,7 @@ class FALRU : public BaseTags
/** The size of the cache. */
const unsigned size;
/** The hit latency of the cache. */
const unsigned hitLatency;
const Cycles hitLatency;
/** Array of pointers to blocks at the cache size boundaries. */
FALRUBlk **cacheBoundaries;
@ -155,7 +155,7 @@ public:
* @param size The size of the cache.
* @param hit_latency The hit latency of the cache.
*/
FALRU(unsigned blkSize, unsigned size, unsigned hit_latency);
FALRU(unsigned blkSize, unsigned size, Cycles hit_latency);
~FALRU();
/**
@ -181,7 +181,7 @@ public:
* @param inCache The FALRUBlk::inCache flags.
* @return Pointer to the cache block.
*/
FALRUBlk* accessBlock(Addr addr, int &lat, int context_src, int *inCache = 0);
FALRUBlk* accessBlock(Addr addr, Cycles &lat, int context_src, int *inCache = 0);
/**
* Find the block in the cache, do not update the replacement data.
@ -205,7 +205,7 @@ public:
* Return the hit latency of this cache.
* @return The hit latency.
*/
int getHitLatency() const
Cycles getHitLatency() const
{
return hitLatency;
}

17
src/mem/cache/tags/iic.cc vendored
View file

@ -220,11 +220,11 @@ IIC::regStats(const string &name)
IICTag*
IIC::accessBlock(Addr addr, int &lat, int context_src)
IIC::accessBlock(Addr addr, Cycles &lat, int context_src)
{
Addr tag = extractTag(addr);
unsigned set = hash(addr);
int set_lat;
Cycles set_lat;
unsigned long chain_ptr = tagNull;
@ -232,11 +232,11 @@ IIC::accessBlock(Addr addr, int &lat, int context_src)
setAccess.sample(set);
IICTag *tag_ptr = sets[set].findTag(tag, chain_ptr);
set_lat = 1;
set_lat = Cycles(1);
if (tag_ptr == NULL && chain_ptr != tagNull) {
int secondary_depth;
tag_ptr = secondaryChain(tag, chain_ptr, &secondary_depth);
set_lat += secondary_depth;
set_lat += Cycles(secondary_depth);
// set depth for statistics fix this later!!! egh
sets[set].depth = set_lat;
@ -250,9 +250,7 @@ IIC::accessBlock(Addr addr, int &lat, int context_src)
}
}
// @todo: is hashDelay is really cycles, then
// multiply with period
set_lat = set_lat * hashDelay + hitLatency;
set_lat = Cycles(set_lat * hashDelay + hitLatency);
if (tag_ptr != NULL) {
// IIC replacement: if this is not the first element of
// list, reorder
@ -263,8 +261,9 @@ IIC::accessBlock(Addr addr, int &lat, int context_src)
hitDepthTotal += sets[set].depth;
tag_ptr->status |= BlkReferenced;
lat = set_lat;
if (tag_ptr->whenReady > curTick() && tag_ptr->whenReady - curTick() > set_lat) {
lat = tag_ptr->whenReady - curTick();
if (tag_ptr->whenReady > curTick() &&
cache->ticksToCycles(tag_ptr->whenReady - curTick()) > set_lat) {
lat = cache->ticksToCycles(tag_ptr->whenReady - curTick());
}
tag_ptr->refCount += 1;

8
src/mem/cache/tags/iic.hh vendored
View file

@ -176,7 +176,7 @@ class IIC : public BaseTags
/** The associativity of the primary table. */
const unsigned assoc;
/** The base hit latency. */
const unsigned hitLatency;
const Cycles hitLatency;
/** The subblock size, used for compression. */
const unsigned subSize;
@ -278,9 +278,9 @@ class IIC : public BaseTags
/** The associativity of the primary table. */
unsigned assoc;
/** The number of cycles for each hash lookup. */
unsigned hashDelay;
Cycles hashDelay;
/** The number of cycles to read the data. */
unsigned hitLatency;
Cycles hitLatency;
/** The replacement policy. */
Repl *rp;
/** The subblock size in bytes. */
@ -420,7 +420,7 @@ class IIC : public BaseTags
* @param lat The access latency.
* @return A pointer to the block found, if any.
*/
IICTag* accessBlock(Addr addr, int &lat, int context_src);
IICTag* accessBlock(Addr addr, Cycles &lat, int context_src);
/**
* Find the block, do not update the replacement data.

6
src/mem/cache/tags/lru.cc vendored
View file

@ -116,7 +116,7 @@ LRU::~LRU()
}
LRU::BlkType*
LRU::accessBlock(Addr addr, int &lat, int master_id)
LRU::accessBlock(Addr addr, Cycles &lat, int master_id)
{
Addr tag = extractTag(addr);
unsigned set = extractSet(addr);
@ -128,8 +128,8 @@ LRU::accessBlock(Addr addr, int &lat, int master_id)
DPRINTF(CacheRepl, "set %x: moving blk %x to MRU\n",
set, regenerateBlkAddr(tag, set));
if (blk->whenReady > curTick()
&& blk->whenReady - curTick() > hitLatency) {
lat = blk->whenReady - curTick();
&& cache->ticksToCycles(blk->whenReady - curTick()) > hitLatency) {
lat = cache->ticksToCycles(blk->whenReady - curTick());
}
blk->refCount += 1;
}

6
src/mem/cache/tags/lru.hh vendored
View file

@ -68,7 +68,7 @@ class LRU : public BaseTags
/** The associativity of the cache. */
const unsigned assoc;
/** The hit latency. */
const unsigned hitLatency;
const Cycles hitLatency;
/** The cache sets. */
CacheSet *sets;
@ -139,7 +139,7 @@ public:
* @param lat The access latency.
* @return Pointer to the cache block if found.
*/
BlkType* accessBlock(Addr addr, int &lat, int context_src);
BlkType* accessBlock(Addr addr, Cycles &lat, int context_src);
/**
* Finds the given address in the cache, do not update replacement data.
@ -221,7 +221,7 @@ public:
* Return the hit latency.
* @return the hit latency.
*/
int getHitLatency() const
Cycles getHitLatency() const
{
return hitLatency;
}

8
tests/configs/inorder-timing.py
View file

@ -33,8 +33,8 @@ m5.util.addToPath('../configs/common')
class MyCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
tgts_per_mshr = 5
@ -44,8 +44,8 @@ class MyL1Cache(MyCache):
cpu = InOrderCPU(cpu_id=0)
cpu.addTwoLevelCacheHierarchy(MyL1Cache(size = '128kB'),
MyL1Cache(size = '256kB'),
MyCache(size = '2MB', hit_latency='10ns',
response_latency='10ns'))
MyCache(size = '2MB', hit_latency = 20,
response_latency = 20))
cpu.clock = '2GHz'

10
tests/configs/memtest.py
View file

@ -34,8 +34,8 @@ from m5.objects import *
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 12
tgts_per_mshr = 8
@ -47,8 +47,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -65,7 +65,7 @@ system = System(cpu = cpus, funcmem = SimpleMemory(in_addr_map = False),
# l2cache & bus
system.toL2Bus = CoherentBus(clock="2GHz", width=16)
system.l2c = L2(size='64kB', assoc=8)
system.l2c = L2(clock = '2GHz', size='64kB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
# connect l2c to membus

7
tests/configs/o3-timing-checker.py
View file

@ -42,8 +42,8 @@ m5.util.addToPath('../configs/common')
class MyCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
tgts_per_mshr = 5
@ -57,6 +57,9 @@ cpu.addCheckerCpu()
cpu.addTwoLevelCacheHierarchy(MyL1Cache(size = '128kB'),
MyL1Cache(size = '256kB'),
MyCache(size = '2MB'))
# @todo Note that the L2 latency here is unmodified and 2 cycles,
# should set hit latency and response latency to 20 cycles as for
# other scripts
cpu.clock = '2GHz'
system = System(cpu = cpu,

12
tests/configs/o3-timing-mp.py
View file

@ -35,8 +35,8 @@ m5.util.addToPath('../configs/common')
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -48,8 +48,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -61,8 +61,8 @@ cpus = [ DerivO3CPU(cpu_id=i) for i in xrange(nb_cores) ]
system = System(cpu = cpus, physmem = SimpleMemory(), membus = CoherentBus())
# l2cache & bus
system.toL2Bus = CoherentBus()
system.l2c = L2(size='4MB', assoc=8)
system.toL2Bus = CoherentBus(clock = '2GHz')
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
# connect l2c to membus

7
tests/configs/o3-timing.py
View file

@ -33,8 +33,8 @@ m5.util.addToPath('../configs/common')
class MyCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
tgts_per_mshr = 5
@ -46,6 +46,9 @@ cpu = DerivO3CPU(cpu_id=0)
cpu.addTwoLevelCacheHierarchy(MyL1Cache(size = '128kB'),
MyL1Cache(size = '256kB'),
MyCache(size = '2MB'))
# @todo Note that the L2 latency here is unmodified and 2 cycles,
# should set hit latency and response latency to 20 cycles as for
# other scripts
cpu.clock = '2GHz'
system = System(cpu = cpu,

18
tests/configs/pc-o3-timing.py
View file

@ -39,8 +39,8 @@ mem_size = '128MB'
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -52,8 +52,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -64,8 +64,8 @@ class L2(BaseCache):
class PageTableWalkerCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
size = '1kB'
tgts_per_mshr = 12
@ -76,8 +76,8 @@ class PageTableWalkerCache(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -94,7 +94,7 @@ system.kernel = FSConfig.binary('x86_64-vmlinux-2.6.22.9')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/pc-simple-atomic.py
View file

@ -39,8 +39,8 @@ mem_size = '128MB'
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -52,8 +52,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -64,8 +64,8 @@ class L2(BaseCache):
class PageTableWalkerCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
size = '1kB'
tgts_per_mshr = 12
@ -77,8 +77,8 @@ class PageTableWalkerCache(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -96,7 +96,7 @@ system.kernel = FSConfig.binary('x86_64-vmlinux-2.6.22.9')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/pc-simple-timing.py
View file

@ -40,8 +40,8 @@ mem_size = '128MB'
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -53,8 +53,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -65,8 +65,8 @@ class L2(BaseCache):
class PageTableWalkerCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
size = '1kB'
tgts_per_mshr = 12
@ -77,8 +77,8 @@ class PageTableWalkerCache(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -95,7 +95,7 @@ system.kernel = FSConfig.binary('x86_64-vmlinux-2.6.22.9')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

14
tests/configs/realview-o3-checker.py
View file

@ -46,8 +46,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -59,8 +59,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -71,8 +71,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -89,7 +89,7 @@ system.cpu = cpu
cpu.addCheckerCpu()
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/realview-o3-dual.py
View file

@ -37,8 +37,8 @@ from Benchmarks import *
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -74,16 +74,16 @@ class IOCache(BaseCache):
cpus = [DerivO3CPU(cpu_id=i) for i in xrange(2) ]
#the system
system = FSConfig.makeArmSystem('timing', "RealView_PBX", None, False)
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave
system.cpu = cpus
#create the l1/l2 bus
system.toL2Bus = CoherentBus()
system.toL2Bus = CoherentBus(clock = '2GHz')
#connect up the l2 cache
system.l2c = L2(size='4MB', assoc=8)
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
system.l2c.mem_side = system.membus.slave

14
tests/configs/realview-o3.py
View file

@ -37,8 +37,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -78,7 +78,7 @@ system = FSConfig.makeArmSystem('timing', "RealView_PBX", None, False)
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/realview-simple-atomic-dual.py
View file

@ -37,8 +37,8 @@ from Benchmarks import *
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -74,16 +74,16 @@ class IOCache(BaseCache):
cpus = [AtomicSimpleCPU(cpu_id=i) for i in xrange(2) ]
#the system
system = FSConfig.makeArmSystem('atomic', "RealView_PBX", None, False)
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave
system.cpu = cpus
#create the l1/l2 bus
system.toL2Bus = CoherentBus()
system.toL2Bus = CoherentBus(clock = '2GHz')
#connect up the l2 cache
system.l2c = L2(size='4MB', assoc=8)
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
system.l2c.mem_side = system.membus.slave

14
tests/configs/realview-simple-atomic.py
View file

@ -36,8 +36,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -49,8 +49,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -61,8 +61,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -77,7 +77,7 @@ system = FSConfig.makeArmSystem('atomic', "RealView_PBX", None, False)
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/realview-simple-timing-dual.py
View file

@ -37,8 +37,8 @@ from Benchmarks import *
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -74,16 +74,16 @@ class IOCache(BaseCache):
cpus = [TimingSimpleCPU(cpu_id=i) for i in xrange(2) ]
#the system
system = FSConfig.makeArmSystem('timing', "RealView_PBX", None, False)
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave
system.cpu = cpus
#create the l1/l2 bus
system.toL2Bus = CoherentBus()
system.toL2Bus = CoherentBus(clock = '2GHz')
#connect up the l2 cache
system.l2c = L2(size='4MB', assoc=8)
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
system.l2c.mem_side = system.membus.slave

14
tests/configs/realview-simple-timing.py
View file

@ -37,8 +37,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -78,7 +78,7 @@ system = FSConfig.makeArmSystem('timing', "RealView_PBX", None, False)
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

12
tests/configs/simple-atomic-mp.py
View file

@ -34,8 +34,8 @@ from m5.objects import *
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -47,8 +47,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ system = System(cpu = cpus,
membus = CoherentBus())
# l2cache & bus
system.toL2Bus = CoherentBus()
system.l2c = L2(size='4MB', assoc=8)
system.toL2Bus = CoherentBus(clock = '2GHz')
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
# connect l2c to membus

12
tests/configs/simple-timing-mp.py
View file

@ -34,8 +34,8 @@ from m5.objects import *
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -47,8 +47,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -60,8 +60,8 @@ cpus = [ TimingSimpleCPU(cpu_id=i) for i in xrange(nb_cores) ]
system = System(cpu = cpus, physmem = SimpleMemory(), membus = CoherentBus())
# l2cache & bus
system.toL2Bus = CoherentBus()
system.l2c = L2(size='4MB', assoc=8)
system.toL2Bus = CoherentBus(clock = '2GHz')
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
# connect l2c to membus

7
tests/configs/simple-timing.py
View file

@ -32,8 +32,8 @@ from m5.objects import *
class MyCache(BaseCache):
assoc = 2
block_size = 64
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
mshrs = 10
tgts_per_mshr = 5
@ -43,7 +43,8 @@ class MyL1Cache(MyCache):
cpu = TimingSimpleCPU(cpu_id=0)
cpu.addTwoLevelCacheHierarchy(MyL1Cache(size = '128kB'),
MyL1Cache(size = '256kB'),
MyCache(size = '2MB', hit_latency='10ns', response_latency ='10ns'))
MyCache(size = '2MB', hit_latency= 20,
response_latency = 20))
system = System(cpu = cpu,
physmem = SimpleMemory(),
membus = CoherentBus())

14
tests/configs/tsunami-inorder.py
View file

@ -37,8 +37,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -82,7 +82,7 @@ system = FSConfig.makeLinuxAlphaSystem('timing')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/tsunami-o3-dual.py
View file

@ -37,8 +37,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -78,14 +78,14 @@ system = FSConfig.makeLinuxAlphaSystem('timing')
system.cpu = cpus
#create the l1/l2 bus
system.toL2Bus = CoherentBus()
system.iocache = IOCache()
system.toL2Bus = CoherentBus(clock = '2GHz')
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave
#connect up the l2 cache
system.l2c = L2(size='4MB', assoc=8)
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
system.l2c.mem_side = system.membus.slave

14
tests/configs/tsunami-o3.py
View file

@ -37,8 +37,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 20
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -79,7 +79,7 @@ system = FSConfig.makeLinuxAlphaSystem('timing')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/tsunami-simple-atomic-dual.py
View file

@ -36,8 +36,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -49,8 +49,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -61,8 +61,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -74,16 +74,16 @@ class IOCache(BaseCache):
cpus = [ AtomicSimpleCPU(cpu_id=i) for i in xrange(2) ]
#the system
system = FSConfig.makeLinuxAlphaSystem('atomic')
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave
system.cpu = cpus
#create the l1/l2 bus
system.toL2Bus = CoherentBus()
system.toL2Bus = CoherentBus(clock = '2GHz')
#connect up the l2 cache
system.l2c = L2(size='4MB', assoc=8)
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
system.l2c.mem_side = system.membus.slave

14
tests/configs/tsunami-simple-atomic.py
View file

@ -36,8 +36,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -49,8 +49,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -61,8 +61,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -78,7 +78,7 @@ system = FSConfig.makeLinuxAlphaSystem('atomic')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave

18
tests/configs/tsunami-simple-timing-dual.py
View file

@ -36,8 +36,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -49,8 +49,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -61,8 +61,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -74,16 +74,16 @@ class IOCache(BaseCache):
cpus = [ TimingSimpleCPU(cpu_id=i) for i in xrange(2) ]
#the system
system = FSConfig.makeLinuxAlphaSystem('timing')
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave
system.cpu = cpus
#create the l1/l2 bus
system.toL2Bus = CoherentBus()
system.toL2Bus = CoherentBus(clock = '2GHz')
#connect up the l2 cache
system.l2c = L2(size='4MB', assoc=8)
system.l2c = L2(clock = '2GHz', size='4MB', assoc=8)
system.l2c.cpu_side = system.toL2Bus.master
system.l2c.mem_side = system.membus.slave

14
tests/configs/tsunami-simple-timing.py
View file

@ -37,8 +37,8 @@ import FSConfig
# ====================
class L1(BaseCache):
hit_latency = '1ns'
response_latency = '1ns'
hit_latency = 2
response_latency = 2
block_size = 64
mshrs = 4
tgts_per_mshr = 8
@ -50,8 +50,8 @@ class L1(BaseCache):
class L2(BaseCache):
block_size = 64
hit_latency = '10ns'
response_latency = '10ns'
hit_latency = 20
response_latency = 20
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
@ -62,8 +62,8 @@ class L2(BaseCache):
class IOCache(BaseCache):
assoc = 8
block_size = 64
hit_latency = '50ns'
response_latency = '50ns'
hit_latency = 50
response_latency = 50
mshrs = 20
size = '1kB'
tgts_per_mshr = 12
@ -79,7 +79,7 @@ system = FSConfig.makeLinuxAlphaSystem('timing')
system.cpu = cpu
#create the iocache
system.iocache = IOCache()
system.iocache = IOCache(clock = '1GHz')
system.iocache.cpu_side = system.iobus.master
system.iocache.mem_side = system.membus.slave