Because the python environment may already be up and running by the time
static initializers are run, specifically when gem5 is built as a
library and loaded with dlopen, we can't rely on all of the objects
declaring python initialization callbacks having been constructed by the
time the code which would execute them runs.
To address that problem, this change keeps track of whether the
initialization has already happened when a callback is installed, and if
so, runs the callback immediately.
The original implementation also had users install callbacks by
overriding a virtual function in the PythonInitFunc class, and then
statically allocating an instance of that subclass so its constructor
would be called at initialization time. Calling the function manually if
initialization has already happened won't work in that case, because you
can't call a virtual function from a constructor and get the behavior
you'd want.
Instead, this change makes the PythonInitFunc wrap the actual callback
which is outside of the structure itself. Because the callback is not a
virtual function of PythonInitFunc, we can call it in the constructor
without issue.
Also, the Callback type has to be a bare function pointer and not a
std::function<...> because the argument it takes is a pybind11::module_
reference. Pybind11 sets the visibility of all of its code to hidden to
improve binary size, but unfortunately that causes problems when
accepting one as an argument in a publically accessible lambda in g++.
clang doesn't raise a warning, but g++ does which breaks the build. We
could potentially disable this warning, but accepting a function pointer
instead works just as well, since captureless lambdas can be trivially
converted into function pointers, and they don't seem to upset g++.
Change-Id: I3fb321b577090df67c7be3be0e677c2c2055d446
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/54325
Maintainer: Gabe Black <gabe.black@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
GCC11 introduces a new warning, free-nonheap-object, which would check
if the code potentially calls delete with a nonheap object. The
scheduler is a global object, and its events members fall to this case.
Here's a simplified example.
https://godbolt.org/z/q6GqEfETa
We think this is a false positive warning, since we set auto delete to
false in the event constructor. To avoid performance penalty, we want to
keep current implementation. As the result, we disable the warning in
the SConscript.
Change-Id: I606ebfdec0af7c78d7bbb336faa1f587caa62855
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/54064
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Apply the gem5 namespace to the codebase.
Some anonymous namespaces could theoretically be removed,
but since this change's main goal was to keep conflicts
at a minimum, it was decided not to modify much the
general shape of the files.
A few missing comments of the form "// namespace X" that
occurred before the newly added "} // namespace gem5"
have been added for consistency.
std out should not be included in the gem5 namespace, so
they weren't.
ProtoMessage has not been included in the gem5 namespace,
since I'm not familiar with how proto works.
Regarding the SystemC files, although they belong to gem5,
they actually perform integration between gem5 and SystemC;
therefore, it deserved its own separate namespace.
Files that are automatically generated have been included
in the gem5 namespace.
The .isa files currently are limited to a single namespace.
This limitation should be later removed to make it easier
to accomodate a better API.
Regarding the files in util, gem5:: was prepended where
suitable. Notice that this patch was tested as much as
possible given that most of these were already not
previously compiling.
Change-Id: Ia53d404ec79c46edaa98f654e23bc3b0e179fe2d
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/46323
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Matthew Poremba <matthew.poremba@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
As part of recent decisions regarding namespace
naming conventions, all namespaces will be changed
to snake case.
sim_clock::Float became sim_clock::as_float.
"as_float" was chosen because "float" is a reserved
keywords, and this namespace acts as a selector of
how to read the internal variables. Another
possibility to resolve this would be to remove the
namespaces "Float" and "Int" and use unions instead.
Change-Id: I7b3d9c6e9ab547493d5596c7eda080a25509a730
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/45435
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Maintainer: Gabe Black <gabe.black@gmail.com>
This flag was necessary because of self assignments in the ISA parser
where self assignments are often hints to the parser itself, and in one
case because a pybind-ism used to attach the -= operator looked like a
self assignment.
This change narrows the scope of the flag that disables this warning to
only files generated by the ISA parser, and the single file in the
systemc code which uses that operator overload.
Change-Id: Ib64fc72e46f894cba9064afcdbdcc5859c30e745
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40952
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This should help reduce warning spew when building with newer compilers.
The pybind11::module type has been renamed pybind11::module_ to avoid
conflicts with c++20 modules, according to the pybind11 changelog, so
this CL also updates gem5 source to use the new type. There is
supposedly an alias pybind11::module which is for compatibility, but we
still get linker errors without changing to pybind11::module_.
Change-Id: I0acb36215b33e3a713866baec43f5af630c356ee
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40255
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
The create() method on Params structs usually instantiate SimObjects
using a constructor which takes the Params struct as a parameter
somehow. There has been a lot of needless variation in how that was
done, making it annoying to pass Params down to base classes. Some of
the different forms were:
const Params &
Params &
Params *
const Params *
Params const*
This change goes through and fixes up every constructor and every
create() method to use the const Params & form. We use a reference
because the Params struct should never be null. We use const because
neither the create method nor the consuming object should modify the
record of the parameters as they came in from the config. That would
make consuming them not idempotent, and make it impossible to tell what
the actual simulation configuration was since it would change from any
user visible form (config script, config.ini, dot pdf output).
Change-Id: I77453cba52fdcfd5f4eec92dfb0bddb5a9945f31
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/35938
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The queue in systemC scheduler is implemented as a std::map. This provides
the best big-O solution. However, most of simulation usecases has very
small number of pending events. This is expected as we usually only trigger a
few new events after some events are processed. In such scenario, we
should optimize for insert/erase instead of search. This change use
std::list instead of std::map.
As a proof, we can find that gem5's original event_queue is also
implemented as a list instead of tree.
We see 5% speed improvement with the example provided by Matthias Jung:
https://gist.github.com/myzinsky/557200aa04556de44a317e0a10f51840
Change-Id: I75c30df9134e94df42fd778115cf923488ff5886
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34515
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This mechanism had just been plumbed into the regular request_update,
but that doesn't have any thread safety which is the whole point of
async_request_update. This new mechanism puts async update requests
into their own list which is checked any time normal updates happen.
The delta cycle which triggers those updates must happen through some
other means which will usually be ok. The exact timing of the update
is undefined, so it would be legal for it to either not be recognized
before the impending end of the simulation, or for it to get picked up
by subsequent activity. If there isn't subsequent activity but the
simulation also doesn't end, for instance if there are only gem5 events
left, then that update could be lost. That is an unresolved issue.
It would be nice to schedule a "ready" event if async updates were
added which would ensure they wouldn't starve. Unfortunately that
requires the event queue lock, and in practice it's been found that a
systemc process might block, effectively holding the event queue lock,
while it waits for some asyncrhonous update to give it something to do.
This effectively deadlocks the system since the update is blocked on
the lock the main thread holds, and the main thread is blocked waiting
for the update.
Change-Id: I580303db01673faafc2e63545b6a69b3327a521c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18288
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
In those cases, there's no sc_main to return control to. The python
config script is serving more or less the same purpose, so we can
return control to there instead.
Change-Id: I3cf0623ae51d989b883fb8556ebbf44651bbec99
Reviewed-on: https://gem5-review.googlesource.com/c/16445
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
When running without sc_main, sc_start won't be called, and therefore
runToTime and maxTick won't be initialized. To avoid the scheduler
getting confused and behaving erratically, those values should be
initialized to something that makes sense in situations where there's
no sc_main.
Change-Id: I6ddd7db9ecb36d716eb5ef75e1c38bb99a386092
Reviewed-on: https://gem5-review.googlesource.com/c/16443
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
If sc_main hasn't run, for instance if there isn't an sc_main and gem5
is orchestrating the simulation directly, then exceptions shouldn't be
thrown to the sc_main fiber since it isn't running and may not be able
to run since sc_main may not even exist.
Instead, we need to check whether it makes sense to throw to sc_main,
and if not pass the exception directly to the report handler since
there likely won't be anyone to catch it if we just throw it from the
scheduler or into general purpose gem5.
Since the name throwToScMain is no longer a complete description for
what that function does, this change renames it to throwUp, since it
will now throw exceptions up the stack, either to sc_main or to the
conceptual top level by going directly to the report handler.
Change-Id: Ibdc92c9cf213ec6aa15ad654862057b7bf2e1c8e
Reviewed-on: https://gem5-review.googlesource.com/c/16442
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
We were supposed to do one update phase before adding all processes
which didn't have dont_initialize() called to the run queue. We were
doing that in the opposite order.
Change-Id: I6dd4108040d455dc78029da029a2eb5e7af819cb
Reviewed-on: https://gem5-review.googlesource.com/c/14915
Reviewed-by: Matthias Jung <jungma@eit.uni-kl.de>
Maintainer: Gabe Black <gabeblack@google.com>
This is a reasonable size for a stack, and the default size for a stack
on Linux as determined by some quick Googling. The sc_main fiber would
normally use the primary program stack if run under the standard
systemc implementation, and so might expect to have more room to play
with.
Change-Id: Ie12344939e7b249da203630ebc7dc773a387d716
Reviewed-on: https://gem5-review.googlesource.com/c/14396
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>
Some build failures has been seen after USE_SYSTEMC being True by
default and that has been caused by double definition of _XOPEN_SOURCE
and _POSIX_C_SOURCE in some python versions (like 2.7.5, 2.7.13) and
/usr/include/features.h (used by gcc)
Python definition should preceed features.h one, since the latter will
manually #undef them before #define them.
Change-Id: I774711aaf8145df9ad7677a393a60cf3662d6816
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/14095
Maintainer: Gabe Black <gabeblack@google.com>
Now that that's managed in c++, we can do that directly without having
to depend on the python code being available, the code which lets us
call from c++ to python, or for the embedded python interpretter to
have started running and have loaded the appropriate modules.
Change-Id: Ied110d8f22181095f8c0c645636a9bd67964263e
Reviewed-on: https://gem5-review.googlesource.com/c/14056
Maintainer: Gabe Black <gabeblack@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
By pulling out the sc_main fiber (scMainFiber), we can make it
available to different entities in the simulator and avoid having to
have parallel bookkeeping.
Also this will make it possible to hook into sc_main without putting
the code in sc_main.cc.
Change-Id: I7689441424238e9b2e4d2b48e945dea35fd8cc5d
Reviewed-on: https://gem5-review.googlesource.com/c/13977
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>
The previous implementation dereferenced a null pointer to create a
reference which would then have its address taken in the sc_bind_proxy
constructor. clang says that that uses undefined behavior, so this
change adds a default constructor which initializes the two contained
pointers to null explicitly.
We have to hope systemc code doesn't play around with sc_bind_proxy too
much and doesn't accidentally use this constructor unintentionally, but
it seems like the least bad possible solution which makes clang happy.
Change-Id: Ic59603495fe7a406586a18ce44de979f84089bcd
Reviewed-on: https://gem5-review.googlesource.com/c/13879
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
These warnings were removed when the functionality they warned about
was implemented, but there were some leftovers like unnecessary
includes and some helper functions which hid gem5 specific headers
from the ext directory.
Change-Id: Ic886ac0f1264687524e3a7b7eaab8836f318a5a2
Reviewed-on: https://gem5-review.googlesource.com/c/13398
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
