Add some more stringent Python tests that ensure that we can link with
and run applications that embed Python. This is implemented by running
building a small c++ program that embeds Python using PyBind11. The
program is run by the build system and prints the version of the
Python interpreter. The version information is then used by the build
system to ensure that the installed version is supported.
Change-Id: I727e0832f171362f5506247c022bea365068a0f6
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/36383
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The parseParam and showParam functions partially worked using template
specialization, and partially worked using function overloading. The
template specialization could be resolved later once other functions
were added, but the regular function overloads could not. That meant
that it was practically impossible to add new definitions of those two
functions local to the types they worked with.
Also, because C++ does not allow partial specialization of template
functions, it would not be possible to truly use specialization to wire
in BitUnion types.
To fix these problems, these functions have been turned into structs
which wrap static functions. These can be partially specialized as
desired, making them compatible with BitUnions. Also, it's not possible
to overload structures like it is with functions, so only specialization
is considered, not overloading.
While making these changes, these functions (now structs) were also
reworked so that they share implementation more, and are generally
more streamlined.
Given the fact that the previous parseParam and showParam functions
could not actually be expanded beyond serialize.hh, and were not
actually called directly by any code outside of that file, they should
have never been considered part of the API.
Now that these structs actually *can* be specialized outside of this
file, they should be considered part of the interface.
Change-Id: Ic8e677b97fda8378ee1da1f3cf6001e02783fde3
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/36280
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Richard Cooper <richard.cooper@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
These had been written specifically for the vector, list, set, and C
style array types. This change reworks them to share an implementation,
and to work with more general types. The arrayParamOut method requires
std::begin() and std::end() to accept that type, and the arrayParamIn
method requires either insert or push_back, or the type to be an array.
Also fix up a couple of files which accidentally depended on includes in
the serialize headers which are no longer necessary.
Change-Id: I6ec4fe3bb900603bbb4e35c4efa620c249942452
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/36277
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
As gem5 has started to use Python2 incompatible features, compiling
gem5 in a Python2 environment results in an error.
This commit addresses this issue by raising an Exception when scons
is run in a Python2 environment, and adding a few pointers on how to
install Python3 and on how to use scons in a Python3 environment. The
solution works in a system where both Python2 and Python3 are
installed.
JIRA: https://gem5.atlassian.net/browse/GEM5-797
Change-Id: I98d4a39f586f39d9253ab2517b77e86c5ed19466
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/36157
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This is a scoped enum meant to be used mainly in the python world
for DTB autogeneration. By making an ArmInterruptPin self aware of
its own type, we can use it in the C++ world when modelling devices.
For example if a device spec is enforcing a specific triggering behaviour,
its gem5 implementation can query the interrupt type and panic if its
expectations are not met. In this way we are sure what the Linux kernel
sees in the DTB is in sync with how the model really behaves
Change-Id: I66ae3cfbc7b1ed94804f1f882c12eb31f70840da
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/35395
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
In python, the BARs had been configured using three arrays and a scalar
parameter. The arrays tracked the BAR value in the config, whether the
BAR was for a "legacy" IO range, and the size of the BAR, and the
scalar parameter was an offset for the "legacy" IO addresses to map
into the host physical address space. The nature of a BAR was implied
by its raw config space value, with each of the control bits (IO vs.
memory, 64 bit, reserved bits) encoded directly in the value.
Now, the BARs are represented by objects which have different types
depending on what type of BAR they are. There's one for IO, one for
memory, one for the upper 32 bits of a 64 bit BAR (so indices work
out), and one for legacy IO ranges. Each type has parameters which
are appropriate for it, and they're parameters are all grouped together
as a unit instead of being spread across all the previous values.
The legacy IO offset has been removed, since these addresses can be
offset like any other IO address. They can be represented naturally
in the config using their typical IO port numbers, and still be turned
into an address that gem5 will handle correctly in the back end.
Unfortunately, this exposes a problem in the config system where
a VectorParam can't be overwritten successfully one element at a time,
at least when dealing with SimObject classes. It might work with
actual SimObjects in a config, but I haven't tried it. If you were
to do that to, for instance, update the BARs for x86 so that they
used legacy IO ports for the IDE controller, it would complain that
you were trying to instantiate orphaned nodes. Replacing the whole
VectorParam with a new list of BAR objects seems to work, so that's
what's implemented in this change.
On the C++ side, BARs in the config space are treated as flat values
on reads, and are stored in the config structure associated with each
PCI device. On writes, the value is first passed to the BAR object,
and it has a chance to mask any bits which are fixed in hardware and
update its idea of what range it corresponds to in memory.
When sending AddrRanges up to the parent bus to set up routing, the
BARs generate each AddrRange if and only if their type has been
enabled in the config space command register. The BAR object which
represents the upper 32 bits of a 64 bit BAR does not claim to be
IO or memory, and so doesn't contribute a range. It communicates with
the BAR which represents the lower 32 bits, so that that BAR has the
whole base address.
Since the IO or memory BAR enable bits in the command register are now
handled by the PCI device base class, the IDE controller no longer has
to handle that manually. It does still need to keep track of whether
the bus master functionality has been enabled though, which it can
check when those registers are accessed.
There was already a mechanism for decoding addresses based on BARs
in the PCI device base class, but it was overly complicated and not
used consistently across devices. It's been consolidated, and used in
most places where it makes sense.
Finally, a few unnecessary values have been dropped from the base PCI
device's and IDE controller's checkpoint output. These were just local
copies of information already in the BARs, which in turn are already
stored along with the data in the device's config space.
Change-Id: I16d5f8cdf86d7a2d02a6b04d1f9e1b3eb1dd189d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/35516
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
There is a flow of packets as so:
WriteResp -> WriteReq -> WriteCompleteResp
These packets share some variables, in particular senderState and a
status vector.
One issue was the WriteResp packet decremented the status vector, which
was used by the WriteCompleteResp packets to determine when to handle
the global memory response. This could lead to multiple
WriteCompleteResp packets attempting to handle the global memory
response.
Because of that, the WriteCompleteResp packets needed to handle the
status vector. this patch moves WriteCompleteResp packet handling back
into ComputeUnit::DataPort::processMemRespEvent from
ComputeUnit::DataPort::recvTimingResp. This helps remove some redundant
code.
This patch has the WriteResp packet return without doing any status
vector handling, and without deleting the senderState, which had
previously caused a segfault.
Another issue was WriteCompleteResp packets weren't being issued for
each active lane, as the coalesced request was being issued too early.
In order to fix that, we have to ensure every active lane puts their
request into their applicable coalesced request before issuing the
coalesced request. Because of that change, we change the issuing of
CoalescedRequests from GPUCoalescer::coalescePacket to
GPUCoalescer::completeIssue.
That change involves adding a new variable to store the
CoalescedRequests that are created in the calls to coalescePacket. This
variable is a map from instruction sequence number to coalesced
requests.
Additionally, the WriteCompleteResp packet was attempting to access
physical memory in hitCallback while not having any data, which
caused a crash. This can be resolved either by not allowing
WriteCompleteResp packets to access memory, or by copying the data
from the WriteReq packet. This patch denies WriteCompleteResp packets
memory access in hitCallback.
Finally, in VIPERCoalescer::writeCompleteCallback there was a map
that held the WriteComplete packets, but no packets were ever being
removed. This patch removes packets that match the address that was
passed in to the function.
Change-Id: I9a064a0def2bf6c513f5295596c56b1b652b0ca4
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33656
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
These were checking for gcc version 3, well below the minimum version we
support, and were hard wired to be enabled anyway. This change gets rid
of the check and the dead code on the hard wired off branch.
Also, this change cleans up the style in the surviving code and
simplifies it slightly.
Change-Id: I8df73a378f478413c111a4dea962450a37fb4092
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/35977
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Gabe Black <gabeblack@google.com>
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>
On the O3 CPU, when the number of threads on the CPU (SMT) is too low to
hold all the old style CPU workload items, then it would increase the
number of threads to match. There are three problems with this.
1. This behavior was only implemented on O3.
2. It could silently hide a bug in the config where the number of
workload items was accidentally too big.
3. It makes the DerivO3CPUParams struct tamper with itself in the
create() method, which means not even config.ini will accurately
reflect the actual config of the system.
Change-Id: I0aab70d4b98093f7f14156ca437e763f031049ab
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/35937
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
