Typically, a memory controller is assigned an address range of the
form [start, end). This address range might be interleaved and
therefore only a non-continuous subset of the addresses in the address
range is handed by this controller.
Prior to this patch, the DRAM controller was unaware of the
interleaving and as a result the address range could affect the
mapping of addresses to DRAM ranks, rows and columns. This patch
changes the DRAM controller, to transform the input address to a
continuous range of the form [0, size). As a result the DRAM
controller always operates on a dense and continuous address range
regardlesss of the system configuration.
Change-Id: I7d273a630928421d1854658c9bb0ab34e9360851
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19328
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Wendy Elsasser <wendy.elsasser@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Often a request that hits on an MSHR has to be deferred as it can't be
serviced by the current response.
For example, a request that requires writable has to be deferred when
the response is expected to bring in a read-only copy of the
block. However, there are cases where the response, although not
expected to do so, brings a writable copy and as a result we also
service deferred targets. In such cases, we promote deferred targets
up until the first that can't be serviced by the current response
(e.g., cache maintainance operation). If the first deferred target is
incompatible we don't promote any targets at all.
Change-Id: Ib3e13be51120b7c0f0053b83b76bde03e1b7dd4e
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/22127
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
The MSHR keeps track of outstanding writes and services them as a
whole line write whenever possible. To do this the outstanding writes
have to be compatible (e.g., not strictly ordered). Prior to this
change, due to this tracking mechanism, the MSHR would not service a
WriteLineReq with flags that do not allow merging as a full line write
even if it was the first target triggering an assertion. This
changeset fixes this bug.
Change-Id: I2cbf5ece0c108c1fcfe6855e8f194408d5ab8ce2
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/22126
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Tested-by: kokoro <noreply+kokoro@google.com>
Modify second chance replacement policy so that entries are inserted
without a second chance. Previously, the second chance bit was set
to true when a cache line was inserted. So the cache line would gain
its second chance when inserting. This is wrong because the cache
block will only get a second chance when it hits.
Here's a quoted citation for the second chance replacement policy:
"Whenever the algorithm examines a page entry, it extracts the associated
usage bit and enters it into the high-order position of a k-bit shift
register after shifting the contents of the register one bit-position
lower. Then if the shift register is nonzero, the page is retained; if the
shift register is zero, the page is replaced by the new page. In either
case the usage bit for the page is turned off and the circular list
pointer is advanced."
(A Paging Experiment with the Multics System, FJ Corbato, 1968)
Change-Id: I0d07e56aa16c67dd36e0d490c3f457f91e46f320
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20882
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Add support in Ruby to use all replacement policies in Classic.
Furthermore, if new replacement policies are added to the
Classic system, the Ruby system will recognize new policies
without any other changes in Ruby system. The following list
all the major changes:
* Make Ruby cache entries (AbstractCacheEntry) inherit from
Classic cache entries (ReplaceableEntry). By doing this,
replacement policies can use cache entries from Ruby caches.
AccessPermission and print function are moved from
AbstractEntry to AbstractCacheEntry, so AbstractEntry is no
longer needed.
* DirectoryMemory and all SLICC files are changed to use
AbstractCacheEntry as their cache entry interface. So do the
python files in mem/slicc/ast which check the entry
interface.
* "main='false'" argument is added to the protocol files where
the DirectoryEntry is defined. This change helps
differentiate DirectoryEntry from CacheEntry because they are
both the instances of AbstractCacheEntry now.
* Use BaseReplacementPolicy in Ruby caches instead of
AbstractReplacementPolicy so that Ruby caches will recognize
the replacement policies from Classic.
* Add getLastAccess() and useOccupancy() function to Classic
system so that Ruby caches can use them. Move lastTouchTick
to ReplacementData struct because it's needed by
getLastAccess() to return the correct value.
* Add a 2-dimensional array of ReplacementData in Ruby caches
to store information for different replacement policies. Note
that, unlike Classic caches, where policy information is
stored in cache entries, the policy information needs to be
stored in a new 2-dimensional array. This is due to Ruby
caches deleting the cache entry every time the corresponding
cache line get evicted.
Change-Id: Idff6fdd2102a552c103e9d5f31f779aae052943f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20879
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
When a writeback needs to be allocated the whenReady field of the
block is not set, and therefore its access latency calculation
uses the previously invalidated value (MaxTick), significantly
delaying execution.
This is fixed by assuming that the data write portion of a write
access is done regardless of previous writes, and that only the
tag latency is important for the critical path latency calculation.
Change-Id: I739132a2deab6eb4c46d084f4ee6dd65177873fd
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20068
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Note that this changes the stat format used by the DRAM
controller. Previously, it would have a structure looking a bit like
this:
- system
- dram: Main DRAM controller
- dram_0: Rank 0
- dram_1: Rank 1
This structure can't be replicated with new-world stats since stats
are confined to the SimObject name space. This means that the new
structure looks like this:
- system
- dram: Main DRAM controller
- rank0: Rank 0
- rank1: Rank 1
Change-Id: I7435cfaf137c94b0c18de619d816362dd0da8125
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/21142
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Wendy Elsasser <wendy.elsasser@arm.com>
The cacheProbe() function will return the victim entry, and it gets
called for multiple times in trigger function in a single miss. This
will cause a problem when we try to add a new replacement policy to
the Ruby system. Certain policy, like RRIP, will modify the block
information every time the getVictim() function gets called. To
prevent future problems, we need to store the victim entry, so that
we only call it once in one miss.
Change-Id: Ic5ca05f789d9bbfb963b8e993ef707020f243702
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/21099
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This change is based on modify the way we move the AtomicOpFunctor*
through gem5 in order to mantain proper ownership of the object and
ensuring its destruction when it is no longer used.
Doing that we fix at the same time a memory leak in Request.hh
where we were assigning a new AtomicOpFunctor* without destroying the
previous one.
This change creates a new type AtomicOpFunctor_ptr as a
std::unique_ptr<AtomicOpFunctor> and move its ownership as needed. Except
for its only usage when AtomicOpFunc() is called.
Change-Id: Ic516f9d8217cb1ae1f0a19500e5da0336da9fd4f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20919
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
It's constructor will now warn that it's deprecated and suggest using
ClockedObject directly. This change also gets rid of the params()
method and the Params typedef since they are functionally equivalent to
the ClockedObject versions.
It also removes the include of mem/port.hh which is not used in
mem_object.hh. This may break code which purposefully or (more likely)
accidentally depended on that transitive include from mem_object.hh.
Change-Id: I6dab3ba626e3f3ab6a6bd86edcf4f5cb4d6d2c45
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20720
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
To ensure that enqueuer observes the practical availability. We
check the message buffer queue size at the start of the cycle.
We also add the size of the stall queue to consider the total
queue size. However, messages can be moved from regular queue
to stall map. This leads to messages being considered twice leading
to false flow control. This patch fixes it by storing the stall map
size at the beginning of the cycle and considering it for checking
availability.
Change-Id: I6ea94f34fe5279b91f74e106d43263e55ec4bf06
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20389
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Make MultiBitSelBloomFilter a generic BloomFilter that maps
multiple entries to an address, and therefore uses multiple
hash functions. This allows the common functionality of both
filters to be merged into one, since they only differ in the
hash functions being used.
Change-Id: I0984067b710a208715f5f2727b8c4312feb6529b
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18873
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This returns a sendFunctional delegate references which can be used
to send functional accesses directly, or more likely when constructing
a PortProxy subclass. In those cases only the functional capabilities
of those ports are needed so there's no reason to require a full port
which supports all three protocols. Also, this removes the last
remaining use of get(Data|Inst)Port which relies on those returning
a port which supports the gem5 protocols, except the default
implementations of this new function. If a CPU doesn't have
traditional gem5 style ports, it can override this function to
do whatever other behavior is necessary and return its real ports
through get(Data|Inst)Port.
Change-Id: Ide4da81e3bc679662cd85902ba6bd537cce54a53
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20237
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
The only part of the MaserPort the PortProxy uses is the sendFunctional
function which is part of the functional protocol. Rather than require
a MasterPort which comes along with a lot of other mechanisms, this
change slightly adjusts the PortProxy to only require that function
through the use of a delegate. That allows lots of flexibility in how
the actual packet gets sent and what sends it.
In cases where code constructs a PortProxy and passes its constructor
an unbound MasterPort, the PortProxy will create a delegate to the
sendFunctional method on its own.
This should also make it easier for objects which don't have
traditional gem5 style ports, for instance systemc models, to implement
just the little bit of the protocol they need, rather than having to
stub out a whole port class, most of which will be ignored.
Change-Id: I234b42ce050f12313b551a61736186ddf2c9e2c7
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20229
Maintainer: Gabe Black <gabeblack@google.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The base Port class can keep track of its peer, and also whether it's
connected. This is partially delegated away from the port subclasses
which still keep track of a cast version of their peer pointer for
their own conveneince, so that it can be used by generic code. Even
with the Port mechanism's new flexibility, each port still has
exactly one peer and is either connected or not based on whether there
is a peer currently.
Change-Id: Id3228617dd1604d196814254a1aadeac5ade7cde
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20232
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
