This change refactors the Looppoint files. While functionally
equivalent, this classes have been moved and altered to be easier to
handle going forward. The following changes have been made:
- New classes have been added to represent the data structure of the
Looppoint JSON. This simplifies the parsing of JSON files and makes it
handle Looppoint data structures. Ultimately this is hidden from the
user via the new 'gem5.resources.Looppoint' class which will be the
front-facing class for Looppoint interactions.
- The `LooppointCheckpoint` class has been replaced with
`LooppointCsvLoader`. This new class takes in a CSV pintpoints file
to load necessary looppoint data.
- The `LoopPointRestore` class has been replaced by
`LooppointJsonLoader`.
- All Looppoint classes have been moved to `gem5.resources`. This will
make it easier when we add Looppoints as specific gem5 resources.
Change-Id: I11dd1fe8f76658db220320584270d57cb37a3c62
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67611
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Added looppoint_save_checkpoint_generator to take checkpoints for
LoopPoint methodology.
Users can decide to update the relative counts storing in the LoopPoint
module and exit when all the target PC-count pairs are encountered or
not.
Change-Id: Id1cf1516f4fa838e20a67530e94b361e42ca09f3
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67197
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Added a set_se_looppoint_workload function to take in information for
workload and a stdlib LoopPoint object that stores all the information
the workload needed to run the LoopPoint sampling method.
Added a get_looppoint function to return the stdlib LoopPoint object.
Change-Id: I7afc5c4c743256f7df97345f331b6f72b7a5fd07
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67196
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
LoopPoint is a multithreaded workload sampling method that targets
PCs and PC execution counts.
The main idea for LoopPoint is to base the beginning and end of the
simjulation sample on the number of times a particular loop (PC) has
been executed globally across all threads in a region that partitioned
with a set length of instruction counts. This in some senses
generalizes SimPoint which use the instruction count of a single
thread.
The link to the paper: https://ieeexplore.ieee.org/document/9773236
The LoopPointCheckpoint is designed to take in LoopPoint data file
and generate the information needed to take checkpoints for LoopPoint
regions(warmup regions+simulation region)
The LoopPointRestore is designed to take in the LoopPOint data file
and generate information needed to to restore a checkpoint taken by
the LoopPOintCheckpoint.
The LoopPoint is the parent class for LoopPointCheckpoint and
LoopPointRestore.
Change-Id: I595b0ff9d350c7c496639748a9c63ecc61fbaec9
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67195
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
PcCountTracker is a probelistener that connects to one core and listens
for a list of Program Counter addresses(PCs). It notifys the
PcCountTrackerManager every time it encounters a Program Counter
address in the list.
PcCountTrackerManager is a SimObject that is responsible for keeping
track of a list of PC-count pairs and the number of time a particular
PC has been executed globally.
This patch adds a way to track the number of times a set of specific
PCs have been executed.
Change-Id: I8f47bfa7e29aa2bb6ab817417266033439b85d51
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67194
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
This commit introduces a PcCountPair type that stores a Program Counter
address and an integer of counts for the Program Counter address.
The PcCountPair can be used in the same way and hashable in both C++
and Python.
Change-Id: I66d93e2c6a1d286cb9dd795ba97f8d887f67d503
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67193
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
This patches does the following:
- Adds 'SimpointResource' which encapsulates Simpoint data and
functionality. It replaces the old 'gem5.util.simpoint.SimPoint'
class. Simpoints can be loaded from gem5-resources using the
`obtain_resource` function.
- Adds 'SimpointDirectoryResource'. This inherits form
'SimpointResource'. While 'SimpointResource' takes raw Simpoint data
via parameters, 'SimpointDirectoryResource' assumes the data exists
in files, in a directory.
- Updates the
"configs/example/gem5_library/checkpoints/simpoints-se-checkpoint.py"
and
"configs/example/gem5_library/checkpoints/simpoints-se-restory.py"
example files to utilize this new Simpoint resource classes.
**Note**: While the old "SimPoint" class
("src/python/gem5/util/simpoint.py") is marked as deprecated, it may be
difficult to utilize given updates to the APIs in the gem5 stdlib Cores
and Simulator modules.
Change-Id: I9bed5c643ffc735838c9f22a58c53547941010e7
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67339
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
With this patch, when a resource entry does not specify a "url" field,
there is no file downloaded. This is necessary infrastructure for
gem5-resources which do not have specific files/directories to be
downloaded but exist solely in the resources.json file.
Change-Id: I0d92e830bfcef750119078b8c226b0659ba7f6cb
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67338
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
This patch allows for the "version" field in the resources.json file to
be `null` (translated to `None` in the Python JSON package) or not
declared. In this case the resources.json file will be used regardless
as to what version the gem5 binary is set. This is useful for testing
purposes.
Tests have been updated to utilize this where possible.
Change-Id: I9d8ae18cb3e61d58bc822bad30853fa3442cb33f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67337
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
This commit specializes the Resource class into specific sub-types.
The `Resource`, `CustomResource` and `CustomDiskImageResource` classes
have been deprecated in favor of the `AbstractResource` subclasses.
Custom Resources can be created via the resource specialization
constructor. Resources can be obtained via the gem5-resource
infrastructure with the `obtain_resource` function.
Fully implemented:
- DiskImageResource
- BinaryResource
- KernelResource
- BootloaderResource
- FileResource
- DirectoryResource
Partially implemented:
- SimpointResource
- CheckpointResource
While the schema of the resource.json file has changed, efforts have
been made to ensure backwards compatibility is maintained during this
transition.
Tests are included in this commit to verify this feature works as
expected.
**Note:** The Simpoint tests are disabled in this commit, to be
reenabled when Simpoint resource specialization is fully incorporated
here:
https://gem5-review.googlesource.com/c/public/gem5/+/67339
Change-Id: I77277ecaffc7abc86db08526aacc0b606ef04fe8
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67175
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Add an import to m5.internal.params which became necessary after:
95f9017c2e configs,python: Clean some cruft out of m5.objects.
This import is necessary but also causes problems when scons calls
build_tools/sim_object_param_struct_hh.py to generate
params/SimObject.hh. m5.internal.params itself imports _m5 and _m5 is
unavalailable resulting in an ImportError. This is bening and we can
safely ignore it.
Change-Id: I3809e81284e730fb9c9e0e7e91bd61b801d73f90
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67797
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Extensible is for carrying additional user-defined
information. Each type of the extension will have a unique
extension ID and there is a linked list of extension in every
Extensible object. There will be most one extension with the same type in
the linked list. With the shared_ptr, the extension will be
deleted automatically. That is, the caller should allocate
the extension and add into the packet.
Change-Id: I54729536a305c91c751d5fb059bd2f9a3db05523
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/62892
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
According to RISC V spec Volumn I, Section 11.1, the CSR will be
written only if RS1 != 0 or imm != 0. However, after the change
of CL(https://gem5-review.googlesource.com/c/public/gem5/+/67717),
it will cause IllegalInstFault to write status CSR if we don't
change the data.
Example of Instruction Fault for mstatus
```
addi a5, zero, 8
csrc mstatus, a5
```
It will cause instruction fault if mstatus value is 0 due to
"newdata_all == olddata_all". We can just simply check if
the data value is changed out of mask.
Change-Id: Iab4ce7ac646a9105dc04e69c24d084572e28ebab
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67897
Reviewed-by: Yu-hsin Wang <yuhsingw@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The initial value of register is set in constructor but there is no
standard way to assign the initial value and default value at the same
time out of that. So we decided to add an extra method to set the
initialValue to current register value. The usecase would be:
reg.get().field1 = val1;
reg.get().field2 = val2;
reg.resetInitialValue();
Change-Id: Ibc5454e2945cc6aff943e6599043edd8ca442f5f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67917
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
The VGPR-offset used when SGPR-base addressing is used can be signed in
Vega. These are global instructions of the format:
`global_load_dword v0, v1, s[0:1]`. This is not explicitly stated in the
ISA manual however based on compiler output the offset can be negative.
This changeset assigns the offset to a signed 32-bit integer and the
compiler takes care of the signedness in the expression which calculates
the final address. This fixes a bad address calculation in a rocPRIM
unit test.
Change-Id: I271edfbb4c6344cb1a6a69a0fd3df58a6198d599
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67412
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
This port will stop execution on the CPU when raised. When lowered, it
will allow execution to reset the state of the CPU and allow execution
to resume. The state could theoretically be reset when the reset state
starts, but then it wouldn't reflect the most up to date condition of
the CPU when resuming. For instance, if a reset vector was set
somehow, that wouldn't be updated if it was changed while reset was
asserted. The tradeoff is that the state won't look like it will when
execution resumes while reset is held (to GDB for instance), but that
seems like a more obvious and less common sort of problem.
This signal is managed by the BaseCPU itself, but is backed by a
virtual method which can be overridden by other CPU types which may
not work the same way or have the same components. For instance, a
fast model CPU could toggle reset lines on the underlying model and
let it handle resetting all the state.
The fast models in particular already have a generic reset line with
the same name, but they have it at the level of the fast model which
may have multiple cores within it, each represented by a gem5 CPU.
It isn't implemented here, but there could be some sort of cooperation
between these signals where the reset at the core level is considered
an "or" of the cluster level reset and the individual core level
resets. At least in the A76 model, there are resets for each individual
core within the cluster as well, which the generic reset toggles.
Another option would be to get rid of the whole cluster reset pin, and
make the user gang the resets for each of the cores together to
whatever reset signal they're using. That's effectively what the
cluster level reset is doing, but within the C++ of the model wrapper
instead of in the python config.
Change-Id: Ie6b4769298ea224ec5dc88360cbb52ee8fbbf69c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67574
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Roger Chang <rogerycchang@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Reviewed-by: Yu-hsin Wang <yuhsingw@google.com>
In the current implementation pollqueue is not thread safe. The design
of multi threads handle async events is thus causing issue in parallel
environment. Given the low rate of async events, it should be OK to only
handle them in the main thread to avoid unexpected racing issues.
Change-Id: Iddd512235e84e9d77f60985bb1771aa4cc693004
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67533
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Reviewed-by: Yu-hsin Wang <yuhsingw@google.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
These are the "reset" and "po_reset" lines. It seems reasonable that
these are the normal reset and the power on reset signals, but that's
not spelled out in the fast model "lisa" file, nor does it explain
exactly what the difference is between them.
Change-Id: I686b4d973fc3cfff8a3ec05f8c95ee2cb6ff6698
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67575
Reviewed-by: Jui-min Lee <fcrh@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This CL enables forwarding stream ID from amba_pv to gem5 world.
The stream ID information is originally stored in master_id of
pv::TransactionAtrribute, then it will be stored to m_id of
amba_pv::amba_pv_extension.
This CL brings the information to stream ID field of
Gem5SystemC::ControlExtension. Then the information can be set to stream
ID of the gem5 packet's request.
After bringing the information to gem5, we can identify the packet's
stream ID from gem5 side. One example usage is PL330. In PL330_DMAC, each
transaction is associated with a stream ID. If we can identitfy the
stream ID, we can, for example, set attribute to specific DMAC channel.
Change-Id: I943ce49fde57b0bcfc18b58c7566eec61cc676f4
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/67591
Reviewed-by: Yu-hsin Wang <yuhsingw@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
