Currently gem5 assumes the amdgpu device to be Vega10. In order to
support more devices we need to handle situations where different
registers and addresses have the same functionality but different
offsets on different devices.
This changeset adds an NBIO class to handle device discovery and driver
initialization related tasks, pulling them out of the AMDGPUDevice
class. The offsets used for MMIOs are reworked slightly to use offsets
rather than absolute addresses. This is because we cannot determine the
absolute address in the constructor since the BAR has not been assigned
by the OS yet.
Change-Id: I14b364374e086e185978334425a4e265cf2760d0
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/70041
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Currently the amdgpu simulated device is assumed to be a Vega10. As a
result there are a few things that are hardcoded. One of those is the
number of SDMAs. In order to add a newer device, such as MI100+, we need
to enable a flexible number of SDMAs.
In order to support a variable number of SDMAs and with the MMIO offsets
of each device being potentially different, the MMIO interface for SDMAs
is changed to use an SDMA class method dispatch table with forwards a
32-bit value from the MMIO packet to the MMIO functions in SDMA of the
format `void method(uint32_t)`. Several changes are made to enable this:
- Allow the SDMA to have a variable MMIO base and size. These are
configured in python.
- An SDMA class method dispatch table which contains the MMIO offset
relative to the SDMA's MMIO base address.
- An updated writeMMIO method to iterate over the SDMA MMIO address
ranges and call the appropriate SDMA MMIO method which matches the
MMIO offset.
- Moved all SDMA related MMIO data bit twiddling, masking, etc. into
the MMIO methods themselves instead of in the writeMMIO method in
SDMAEngine.
Change-Id: Ifce626f84d52f9e27e4438ba4e685e30dbf06dbc
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/70040
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
If an MMIO was previously written and the driver reads it, we should
return the value that was previously read. This overwrites the MMIO
trace value which is the last resort fallback for finding an MMIO value.
This is needed to initialize newer GPU devices in gem5.
Change-Id: Ida2435290b706288e88518b5d920691cdb6dcc09
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/70039
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
During GPUFS checkpoint restore, doorbells callbacks are created based
on certain MQD attributes. These callbacks are required to create new
SDMA doorbells. If these attributes are not present in the checkpoint,
the restore hangs indefinitely waiting for ioctl calls that access these
doorbells to finish execution. This commit adds the attributes required
for checkpoint restore to proceed.
Change-Id: Id3d1b7a2627d4c50133d923096495957a233f675
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/70077
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Matthew Poremba <matthew.poremba@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Matthew Poremba <matthew.poremba@amd.com>
This change addresses an error in the compiler tests:
https://jenkins.gem5.org/job/compiler-checks/573/
For clang versions 6 through 10, as well as GCC 7,
in order to use the "filesystem" module, you must
include the experimental namespace. In all newer
versions, you can use the "filesystem" module as is.
Because of this, include guards to handle this. They include
"<experimental/filesystem>" for the older clang versions and
the "<filesystem>" for all other versions.
As opposed to checking by version, we now check if the
filesystem library has been defined before using it.
Change-Id: I8fb8d4eaa33f3edc29b7626f44b82ee66ffe72be
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69778
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
The generation of the device tree in an arm system requires knowing
cpu-release-addr property, which is only available after setting up
the bootloader.
cpu-release-addr specifies where the secondary CPUs spin/sleep(?) before
being waken up by the kernel.
The incorrect cpu-release-addr causes booting an arm system using the
standard library with the arm's provided bootloader+linux_kernel to
fail to regconize more than 1 core.
Change-Id: Ice0e38492e2f77020b0e30c42dd4e8b7ee58e598
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/70017
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Richard Cooper <richard.cooper@arm.com>
For non-KVM CPUs the VBIOS memory falls into an I/O hole and therefore
gets routed to the PIO bus in gem5. This gets routed to the GPU in the
case of a ROM write. We write to the ROM as a way to "load" the VBIOS
without creating holes in the KVM VM.
This write method allows the same scripts as KVM to be used by writing
to the ROM area and overwriting what might already be there from the
--gpu-rom option.
Change-Id: I8c2d2aa05a823569a774dfdd3bf2d2e773f38683
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/70037
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Previously the CPU type and memory modes were hardcoded for KVM, because
there was a deadlock bug. After some recent testing, this deadlock bug
no longer exists with the simple CPU models. Thus, changing the configs
to allow for other CPU models as a first step toward lifting the KVM
requirement from GPUFS.
Change-Id: Ib616c3ef60f173871421b55a8bb73b25ce2990b5
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69979
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
This file is a required input to the simulator for GPUFS. There seems to
be confusion from several users who are not providing this input. This
usually results in the amdgpu driver failing to load, leading to the
application under test exiting along with it.
This changeset adds a simple md5 hashsum check to compare against the
known good MMIO trace located in the gem5-resources repository.
Change-Id: I59819fc795a6bc4bc6badbd4d120db1246498987
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69978
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
The dmesg level is currently set to 3 which will not display errors if
the amdgpu driver fails to load. Changing to level 8 will show errors in
the gem5 terminal and is not too spammy. This will help GPUFS developers
with bug reports since we would actually be able to observe an error.
Currently if the driver fails to load, there is no way to detect it and
applications will attempt to run, usually failing on getting device
properties.
Change-Id: I56b9581c1a12a8ce329066d18d6a072d006c096d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69977
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Currently, taking a checkpoint with a ruby cache involves moving all
the dirty data in cache to memory. This is done by keeping **only**
simulating the cache until all dirty data are flushed to the memory
before taking the checkpoint.
However, when the cache does not have dirty data, it is a problem if
we keep simulating the cache. E.g., calling checkpoint caused the gem5
"empty event queue" assertion fault when running the ruby cache in
atomic_noncaching mode. Since the mode bypasses the cache, all blocks
are invalid and do not contain dirty data. Subsequently, there is no
event placed to the event queue when we keep **only** simulating the
cache before taking the checkpoint.
This patch fixes this problem by checking if there is any actionable
item when trying to move dirty data to memory. If there is no block
contains dirty data, we simply choose not to continue simulating the
cache before taking the checkpoint.
Change-Id: Idfa09be51274c7fc8a340e9e33167f5b32d1b866
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69897
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
The SwitchableProcessors in the standard library have switched-in and
switched-out cores. The `get_cores` API in the stdlib only returns
switched-in cores. In most uses this is desirable.
In the case of setting workloads in SE mode it's necessary to set the
workload to every core, switched-in and switched-out. As the `get_cores`
function was used for this, SwitchableProcessors were failing when used
in SE Mode.
This patch checks the processor type and, if a SwitchableProcessor, uses
the SwitchableProcessor's special `_all_cores` function which gets all
the cores, regardless as to their switched-in/switched-out status.
Issue-on: https://gem5.atlassian.net/browse/GEM5-1320
Change-Id: I0b7a699ac6196e827667955bef7afa37b2648744
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/68997
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Removed the calls to `sys.exit()` from the Arm simple configs. These
calls terminate gem5's embedded Python interpreter and gem5 at the end
of the config script, preventing gem5 from dropping into the
interactive IPython shell when the `--interactive` option has been
specified.
Change-Id: I0c350b0d107f297691255361d25c566c889f9469
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69687
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Only the GICv3 model has a `gicv4` parameter, causing the current
`baremetal.py` config to throw an exception when used with the
VExpress_GEM5_V1 platform containing a GICv2.
This patch checks for the existence of the `gicv4` parameter, allowing
all VExpress platforms to be used.
Change-Id: I72667a9caee64fa497bda516217cd424050eb242
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69685
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
gem5 supports Tarmac trace generation for Arm simulations, but there
are no examples of how to use this feature.
This patch adds a `--tarmac-gen` option to three of the simple Arm
configs. Tarmac generation is useful for out-of-the-box users, and
this patch also provides an example of how to use the Tarmac
generation feature.
Change-Id: I0d3c523b5c0bb6d94de93bc502e4451622fb635d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69684
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Converted the generic DPRINTF messages for the GICv2 register reads
and writes (showing only the memory mapped address) to finer grained
DPRINTF messages showing the names of the mapped registers being
accessed.
This change is intended to make it easier to debug the GIC setup from
the gem5 debug trace.
Change-Id: Ic418b2ea8438fed6a5a810ebc0b686cd4c891cb0
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69681
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
The Tarmac v8 Register ("R") record serialisation formats the
underlying 64-bit storage using a format string field width specifier.
This sets a minimum number of hex characters for the value, rather
than a maximum number of characters.
Because of this, when formatting a narrowed view of a larger
register (e.g. the 32-bit w0 view of the 64-bit x0 register), if any
of the upper bits in the underlying storage are set, then the number
of hex characters used will be the minimum number required to
represent the full value. This could result in irregular formatting,
for example an odd number of hex characters.
This irregular formatting can cause parsing warnings or failures in
some Tarmac tools, for example the Arm Tarmac Trace Utilities [1].
This patch modifies the "R" record formatting to first mask off the
upper bits of the value in the underlying storage to ensure that the
correct number of hex characters are used for the size of the register
being serialised.
[1] https://github.com/ARM-software/tarmac-trace-utilities
Change-Id: Idbd80553d3bcdb56fa9edddd48440ab7d4dff073
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69680
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
A flag, introduced in this patch:
https://gem5-review.googlesource.com/c/public/gem5/+/68518
allowed users to pass "no-duplicate-sources" to a gem5 compilation to
not symlink sources in the build directory.
In this patch "src" was added as a shared top-level header directory.
This means that the header files are not copied to the "build" directory
whether or not "no-duplicate-sources" is set.
This patch ensures the "src" directory is only added as a shared
top-level headers directory in the case where "no-duplicate-sources" is
set.
In addition, the "duplicate_sources" parameter (the destination for the
"no-duplicate-sources") was "None" by default, and only set to False
when the flag was used. `default=True` has been added so
"duplicate_sources" can be used as a boolean.
This bug was a cause of a Nightly build error:
https://jenkins.gem5.org/job/nightly/570
In this error, building ext/sst resulted in an error as the Makefile
depends on adding "build/RISCV" to the include path. Without the header
files in the "build" directory, building SST failed. Though, ext/stt
should probably not be using header files in the "build/RISCV"
directory. This will be fixed in another change.
Change-Id: I786486a177fe17a67f3b939c539eecdcbfcaeaf2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69717
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Normally this would just generate a warning, but a warning is easy to
miss, and truncating the path to fit would be surprising. Since the max
length isn't likely to change, a path which has to be truncated is
essentially fundementally wrong, and could be defined as something
else which is short enough before being used in the config.
Note that this only applies to either the abstract path which is just
a string, or the file name and not the directory path on a file based
socket.
Change-Id: I8702cf02c03053b5d0b6133f25b0e588de666f15
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69677
Maintainer: Gabe Black <gabeblack@google.com>
Reviewed-by: Earl Ou <shunhsingou@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
These can either be set to an integer, in which case it's interpreted
as a TCP port, or a string, in which case it's treated as a unix domain
socket. If the unix domain socket is prefixed with a "@", it will be
treated as an abstract socket.
When stored in the ini file, there is always a prefix added to make
parsing the string more systematic and less ambiguous. A port number is
prefixed with "#", an abstract socket with "@", and a socket file with
the prefix "P" for "path".
Change-Id: I1fc7a579074e849b3becd936238c62fb0d9a2087
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69165
Reviewed-by: Yu-hsin Wang <yuhsingw@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Use a HostSocket parameter to accept connections, rather than a hand
implementation for unix domain sockets. This consolidates this code
with the code derived from it in ListenSocket, and also makes it
possible to connect to the SharedMemoryServer over an AF_INET socket.
Change-Id: I8e05434d08cffaebdf6c68a967e2ee7613c10a76
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69168
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jui-min Lee <fcrh@google.com>
gem5 officially supports GCC 7+. In GCC 7 the "filesystem" module was
added but only in the "experimental" namespace as
"<experimental/filesystem>". In GCC 8+ the module can be found as
"<filesystem>".
Because of this, include guards to handle this. They include
"<experimental/filesystem>" for the GCC v7 case and the "<filesystem>"
for all other versions.
This bug was partially responsible for this compiler tests failures:
https://jenkins.gem5.org/job/compiler-checks/570
Note: gem5 does not support GCC versions <7. Thus the
"#if __GNUC__ >=8 <GCC 8+ code> #else <GCC 7 code> #endif" logic is
valid.
Change-Id: I31db5488f272f9652edebf24ecefca3722369076
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69598
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
This commit is in response to a failure in the compiler tests:
https://jenkins.gem5.org/job/compiler-checks/570/
GCC versions <9 failed to compile gem5 with the following error:
```
socket.cc:(.text+0x32d5): undefined reference to `std::filesystem::__cxx11::path::_M_split_cmpts()'
socket.cc:(.text+0x32e5): undefined reference to `std::filesystem::create_directory(std::filesystem::__cxx11::path const&, std::error_code&)'
socket.cc:(.text+0x3370): undefined reference to `std::filesystem::current_path[abi:cxx11](std::error_code&)'
socket.cc:(.text+0x33cc): undefined reference to `std::filesystem::__cxx11::path::_M_split_cmpts()'
socket.cc:(.text+0x33d7): undefined reference to `std::filesystem::current_path(std::filesystem::__cxx11::path const&, std::error_code&)'
socket.cc:(.text+0x3468): undefined reference to `std::filesystem::current_path(std::filesystem::__cxx11::path const&, std::error_code&)'
```
This was due to this patch:
https://gem5-review.googlesource.com/c/public/gem5/+/69164 which
utilized `std::filesystem`. Prior to to GCC v9, explicit linking with
the stdc++fs library is required due to `std::filesystem` being an
experimental implementation prior to GGC v9.
Change-Id: I584e29f100cb59b40bd155c212e1814f6d8fbb99
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69597
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
The "cwd" string is used to store the previous working directory, if
m5term needs to change it to maximize the space it has to create a
unix domain socket using a relative path. That string was not being
freed, which is a small memory leak. This change frees that memory.
Change-Id: I1ad901e7c59614dd938165cae91c3f666d86e872
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69557
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
I modified the spec-2017 benchmark file to include a
processor.switch() statement (which had been removed
in an earlier commit). I also replaced the use of
get_roi_ticks() with get_tick_stopwatch() as this
particular workload is not annotated with "workbegin"
and "workend" annotations.
Lastly, I fixed a minor bug in the spec-2006 file
which printed the total simulated ticks as zero due
to incorrect indexing of get_tick_stopwatch().
Change-Id: If0b6f88d6701a11f32e2e69994582524e0d7097b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/69537
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Ayaz Akram <yazakram@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
This adds a new scons flag --no-duplicate-sources to build without
linking source files to the build directory.
I find this very helpful when using CLion, since I can now generate a
compilation database using
`bear scons build/ALL/gem5.debug --no-duplicate-sources` and CLion will
now correctly semantically analyze all the files inside src/.
It also ensures that clicking on a build warning/error now opens the
real source file rather than a symlink.
This is not enabled by default since it's possible that certain use
cases are not working correctly, but the basic testing I've done so
far appears to work just fine.
It appears that with this change the `<root>/src` directory is no longer
added to `PYTHONPATH` when running `tests/main.py`, so this change
depends on https://gem5-review.git.corp.google.com/c/public/gem5/+/68757
Change-Id: Iddc9bf9c8211e68e5432c0a07f5c95f427c1ca16
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/68518
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
