The SVE patches made registers longer by increasing NumVecElemPerVecReg,
but the GDB XML was not updated to account for that, and as a result GDB
connections were failing with:
Remote 'g' packet reply is too long
This commit introduces NumVecElemPerSimdVecReg which counts only the SIMD
register sizes to get it back working. SVE GDB support is not added here.
Change-Id: I4191b9f1999ae02b0308863db4cc9b5b16a27d6d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20468
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The GICv3 update methods are method which are invoked anytime the model
needs to evaluate a change in its state, which most of the time means
managing the state of an interrupt (forwarding it to a PE, deasserting
it, etc).
The way it is currently done is a little bit obscure and doesn't
handle correctly IRQ prioritization.
Example:
An IRQ which is handled by the redistributor (PPI or LPI) was not
competing with any pending interrupts coming from the distributor (SPIs)
once raised by a peripheral.
Also the way the pending state of an interrupt was removed at the
cpu interface level wasn't happening in place where this was actually
happening (E.g. when activating it), but happened with a weird
fullUpdate semantic, where if there was a pending interrupt in a
cpu interface, all cpu interfaces had their pending interrupt (if any)
been disabled.
With this patch, state update always starts at the distributor, and
it goes down until the cpu interface where a Gicv3CPUInterface::update
method selects the winning interrupt coming from distributor/redistributor
to be forwarded to the PE.
Change-Id: I1c517cbc4bf107cc2d7ae7beb2692e3cf5187a40
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20614
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The patch is fixing the following aspects of SGIs
* The conditons over which an SGI can be forwarded to a PE
* SGIs in AArch32 (see below)
It is in fact refactoring SGI generation under a common method in the
cpu interface. It is abandoning the implicit fallthrough mechanism not
only for cosmetic reasons, but also because checking "misc_reg ==" was
only working if the register was an AArch64 one (e.g.
MISCREG_ICC_SGI0R_EL1) and not the AArch32 counterpart (MISCREG_SGI0R).
Change-Id: I6fedfb80388666f4f1d20f6abef378a9f093aa83
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20610
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The current default base address for GICv3 ITS stated in RealView is
0x2c120000. The redistributors base address is 0x2c010000; each
instantiated core has an associated redistributor with memory region
size 0x40000 (with GICv4 extension, enabled by default). With 8 cores,
the redistributor range spans to 0x2c210000, creating a conflict with
the ITS address space.
This patch changes the ITS base address to 0x2e010000 which guarantees
no overlapping with the redistributor.
Change-Id: I7dc1af9e69ac037f85ae96f0985554f1fb8372a0
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20608
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The SMMUv3SlaveInterface is using the xlateSlotsRemaining to model a
limit on the number of translation requests it can receive from the
master device.
Patch
https://gem5-review.googlesource.com/c/public/gem5/+/19308/2
moved the resource acquire/release inside the SMMUTranslationProcess
constructor/destructor, for the sake of having a unique place for
calling the signalDrainDone.
While this is convenient, it breaks the original implementation,
which was freeing resources AFTER a translation has completed, but
BEFORE the final memory access (with the translated PA) is performed.
In other words the xlateSlotsRemaining is only modelling translation
slots and should be release once the PA gets produced.
The patch fixes this mismatch by restoring the resource release in
the right place (while keeping the acquire in the constructor)
and by adding a pendingMemAccess counter, which is keeping track
of a complete device memory request (translation + final access)
and will be used by the draining logic
Change-Id: I708fe2d0b6c96ed46f3f4f9a0512f8c1cc43a56c
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Adrian Herrera <adrian.herrera@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20260
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Non-temporal quadword/double-quadword move instructions.
This change ignores the non-temporal hint and instructions are
implemented to send cacheable request to memory.
This would have some "performance" impact (i.e. having some cache
pollution) to get better "correctness" in behavior.
Change-Id: I2052ac0970f61a54bafb7332762debcb7103202d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20288
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
data_write_req byteEnable which is used in ARM SVE partial writes was not
being zeroed between writes.
As a result, non-SVE memory write instructions such as STP that followed
SVE memory write instructions could still have the write mask active.
This could lead to wrong simulation behaviour, and to an assertion failure:
src/mem/packet.hh:1211: void Packet::writeData(uint8_t*) const: Assertion
`req->getByteEnable().size() == getSize()' failed. '`
Change-Id: I74b5a82675e9923b0ffdf2c1dd9afb00c91cb204
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20448
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
When creating a base class which needs to be a SimObject, it's
necessary to decide ahead of time whether to use PioDevice or
BasicPioDevice in the hierarchy because they inherit from SimObject. If
they were added into the hierarchy later, then the original class would
inherit from SimObject, as would PioDevice. That would create a diamond
inheritance structure which would require virtual inheritance, and
that's a can of worms we'd rather not get into.
A big part of the PioPort mechanism is the PioPort itself which holds
a pointer to its parent device and delegates reads/writes to it. It
does that with a PioDevice pointer, and PioDevice declares virtual
functions for all the callbacks the port can call into.
Instead of that, this change templatizes PioPort based on the class of
the device that holds it. That will let you use a PioPort on *any*
class, as long as it has the methods PioPort depends on. That removes
the need to create an inheritance diamond to add a PioPort down the
line since PioDevice is no longer strictly required.
The PioDevice and BasicPioDevice classes are still around since they
still provide some additional functionality and there are existing
classes which depend on them.
Change-Id: I753afc1e0fa54b91217d54c1f8743c150537e960
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20568
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.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>
When using the HDF5 stats backend, we currently pass the filename from
the command line straight to the HDF5 library. This behaviour is
different from the expected behaviour when using plain text stat files
where relative paths are resolved relative to simout. This change adds
support for resolving hdf5 paths relative to simout.
Change-Id: I09cb0b7468735e697125eda7f04c5a85725ca8d0
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20508
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
This changeset add support for stat dumps in the HDF5 file
format. HDF5 is a binary data format that represents data in a
file-system-like balanced tree. It has native support for
N-dimensional arrays and binary data (e.g., frame buffers).
It has the following benefits over traditional text stat files:
* Efficient storage of time series (multiple stat dumps)
* Fast lookup of stats
* Plenty of existing tooling (e.g., Python libraries and graphical
viewers)
* File format can be used to store frame buffers together with
normal stats.
Drawbacks:
* Large startup cost (single stat dump larger than text equivalent)
* Stat dumps are slower than text
Known limitations:
* Distributions and histograms aren't supported.
HDF5 stat output can be enabled using the 'h5' URL scheme when
overriding the stat file name on gem5's command line. The following
parameters are supported:
* chunking (unsigned): Number of time steps to pre-allocate
(default: 10)
* desc (bool): Output stat descriptions (default: True)
* formulas (bool): Output derived stats (default: True)
Example gem5 command line:
./build/ARM/gem5.opt \
--stats-file="h5://stats.h5?desc=False;formulas=False" \
configs/example/fs.py
Example Python stat consumer that computes IPC:
import h5py
f = h5py.File('stats.h5', 'r')
group = f['/system/cpu']
for i, c in zip(group['committedInsts'], group['numCycles']):
print i, c, i / c
Change-Id: I351c6cbff2fb7bef9012f47876ba227ed288975b
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/8121
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Ciro Santilli <ciro.santilli@arm.com>
Add support for partial stat dumps by passing an optional 'root'
keyword argument to m5.stats.dump(). Specifying root slightly changes
the semantics of the dump command. For legacy reasons, gem5 only
allows one stat dump per tick. This is likely a limitation introduced
as a hack to prevent automatic dumping at the end of simulation from
interfering with explicit dumping from a simulation script. This
restriction does not apply when specifying a root. However, these stat
dumps will still prevent an additional stat dump in the same tick with
an unspecified root.
N.B.: This new API /only/ works for new-style stats that have an
explicit hierarchy. Legacy stats will not be dumped if a root is
specified.
Change-Id: Idc8ff448b9f70a796427b4a5231e7371485130b4
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19369
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This change makes the stat system aware of the hierarchical nature of
stats. The aim is to achieve the following goals:
* Make the SimObject hierarchy explicit in the stat system (i.e.,
get rid of name() + ".foo"). This makes stat naming less fragile
and makes it possible to implement hierarchical formats like
XML/HDF5/JSON in a clean way.
* Make it more convenient to split stats into a separate
struct/class that can be bound to a SimObject. This makes the
namespace cleaner and makes stat accesses a bit more obvious.
* Make it possible to build groups of stats in C++ that can be used
in subcomponents in a SimObject (similar to what we do for
checkpoint sections). This makes it easier to structure large
components.
* Enable partial stat dumps. Some of our internal users have been
asking for this since a full stat dump can be large.
* Enable better stat access from Python.
This changeset implements solves the first three points by introducing
a class (Stats::Group) that owns statistics belonging to the same
object. SimObjects inherit from Stats::Group since they typically have
statistics.
New-style statistics need to be associated with a parent group at
instantiation time. Instantiation typically sets the name and the
description, other parameters need to be set by overriding
Group::regStats() just like with legacy stats. Simple objects with
scalar stats can typically avoid implementing regStats() altogether
since the stat name and description are both specified in the
constructor.
For convenience reasons, statistics groups can be merged into other
groups. This means that a SimObject can create a stat struct that
inherits from Stats::Group and merge it into the parent group
(SimObject). This can make the code cleaner since statistics tracking
gets grouped into a single object.
Stat visitors have a new API to expose the group structure. The
Output::beginGroup(name) method is called at the beginning of a group
and the Output::endGroup() method is called when all stats, and
sub-groups, have been visited. Flat formats (e.g., the text format)
typically need to maintain a stack to track the full path to a stat.
Legacy, flat, statistics are still supported after applying this
change. These stats don't belong to any group and stat visitors will
not see a Output::beginGroup(name) call before their corresponding
Output::visit() methods are called.
Change-Id: I9025d61dfadeabcc8ecf30813ab2060def455648
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19368
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
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>
No caller uses any of the MasterPort specific properties of these
function's return values, so we can instead return a reference to the
base Port class. This makes it possible for the data and inst ports
to be of any port type, not just gem5 style MasterPorts. This makes
life simpler for, for example, systemc based CPUs which might have TLM
ports.
It also makes it possible for any two CPUs which have compatible ports
to be switched between, as long as the ports they use support being
unbound. Unfortunately that does not include TLM or systemc ports which
are bound permanently.
Change-Id: I98fce5a16d2ef1af051238e929dd96d57a4ac838
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20240
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@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>
