The current SDMA wrap around handling only considers the ring buffer
location as seen by the GPU. Eventually when the end of the SDMA ring
buffer is reached, the driver waits until the rptr written back to the
host catches up to what the driver sees before wrapping around back to
the beginning of the buffer. This writeback currently does not happen at
all, causing hangs for applications with a lot of SDMA commands.
This changeset first fixes the sizes of the queues, especially RLC
queues, so that the wrap around occurs in the correct place. Second, we
now store the rptr writeback address and the absoluate (unwrapped) rptr
value in each SDMA queue. The absolulte rptr is what the driver sends to
the device and what it expects to be written back.
This was tested with an application which basically does a few hundred
thousand hipMemcpy() calls in a loop. It should also fix the issue with
pannotia BC in fullsystem mode.
Change-Id: I53ebdcc6b02fb4eb4da435c9a509544066a97069
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65351
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
The AMDKernelCode object can span potentially span two pages. Currently
the copy loop from device memory only translates once at the base
address.
This changeset translates one cache line at a time before copying and
has the ancillary benefit for cleaning up this code a bit.
Change-Id: I602bc12d8f8c5d3a3e57ab3f42f7dd3df58dc144
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65251
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Per RISC-V ISA Manual, vol II, section 3.1.6.6, page 26, the SD bit is
a read-only bit indicating whether any of FS, VS, and XS fields being
in the respective dirty state.
Per section 3.1.6, page 20, the SD bit is the most significant bit of
the mstatus register for both RV32 and RV64.
Per section 3.1.6.6, page 29, the explicit formula for updating the SD is,
SD = ((FS==DIRTY) | (XS==DIRTY) | (VS==DIRTY))
Previously in gem5, this bit is not updated anywhere in the gem5
implementation. This cause an issue of incorrectly saving the context
before entering the system call and consequently, incorecttly restoring
the context after a system call as described here [1].
Ideally, we want to update the SD after every relevant instruction;
however, lazily updating the Status register upon its read produces
the same effect.
[1] https://gem5-review.googlesource.com/c/public/gem5/+/65272/
Change-Id: I1db0cc619d43bc5bacb1d03f6f214345d9d90e28
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65273
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Per RISC-V ISA Manual, vol II, section 3.1.6.6, page 25, the
FS field of the mstatus register encodes the status of the floating
point unit, including the floating point registers. Per page 27,
microarchitecture can choose to set the FS field to Dirty even if
the floating point unit has not been modified.
Per section 3.1.6, page 20, the FS field is located at bits 14..13
of the mstatus register.
Per section 3.1.6.6, page 27, the FS field is used for saving
context.
Upon a system call, the Linux kernel relies on mstatus for
choosing registers to save for switching to kernel code.
In particular, if the SD bit (updating this bit is also a bug
in gem5 and will be explained in the next commit) is not set
properly due to the FS field being incorrect, the process of saving
the context and restoring the context result in the floating
point registers being zeroed out. I.e., upon the saving context
function call, the floating point registers are not saved, while
in restore context function call, the floating point registers
are overwritten with zero bits.
Previously, in gem5 RISC-V ISA, the FS field is not updated upon
floating point instruction execution. This caused issue on context
saving described above.
This change conservatively updates the FS field to Dirty on
the execution of any floating point instruction.
Change-Id: I8b3b4922e8da483cff3a2210ee80c163cace182a
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65272
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Running workloads likely causes some content to be written to
the disk image, e.g., `m5 readfile`. However, on riscv boards,
the default kernel param specifies the disk image to be read-only.
This change changes this param so that the disk image is
read-write by default.
Change-Id: I414e483ad11d747f34433560e32a8f91a425ce7e
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65194
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This patch:
- Makes this function private.
- Updates the function's documentation.
- Changes the 'init' parameter to 'board_initialized'.
It doesn't make much sense for this function to be exposed directly to
the user as it requires knowing whether the board is initialized or not.
In addition to this I believe it makes more sense for the 'init' logic
to be flipped and renamed "board_initialized' so that this value is True
if the board has been initialized.
The documentation for this function has been updated.
Change-Id: I016c65bde88357111d3e648d7aa99aeb6e31f410
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/64833
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
This change:
- Makes this function private.
- Adds better documentation describing the usage.
- Changes the 'init' param to 'board_initialized'
This function really doesn't make much sense to set directly by an
stdlib user. It requires knowing whether or not the the board has been
initialized which is an annoying detail and will cause error if set
incorrectly.
The logic of the `init` parameter has been flipped to be
`board_initialized`. This makes it clearer what the parameter is
doing and what it's for.
The documentation for this function has been updated to make it clearer
on how the `board_initialized` parameter should be used correctly.
Change-Id: I567a48df06e6327b38673a2c510065d4334657e2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/64832
Reviewed-by: Melissa Jost <mkjost@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
For Statistics the value is returned. E.g.:
```
print(simstats.board.core.some_integer)
> 5
```
For Groups the names of the stats in that group are listed.
E.g.:
```
print(stats.board.core)
> [Group: [some_integer, another_stat, another_group]]
```
Change-Id: I94cea907608fba622f4fc141d5b22ac95d8cde40
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/63271
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
The exclusion in .pre-commit-config.yaml covered all files in
src/python/m5/ext. This excludes src/python/m5/exit/pystats, which we
want covered by black. This commit updates .pre-commit-config.yaml to
only exclude src/python/m5/ext/pyfdt.
This change also runs black on these files.
Change-Id: Iecff45ea2a27a37fc0d00b867d41300aad911c7a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/63711
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Hi, we are security researchers from the Advanced Research Center at Trellix.
We have began a campaign to patch a widespread bug named CVE-2007-4559.
CVE-2007-4559 is a 15 year old bug in the Python tarfile package. By using
extract() or extractall() on a tarfile object without sanitizing input,
a maliciously crafted .tar file could perform a directory path traversal
attack. We found at least one unsantized extractall() in your codebase
and are providing a patch for you via pull request. The patch essentially
checks to see if all tarfile members will be extracted safely and throws
an exception otherwise. We encourage you to use this patch or your own
solution to secure against CVE-2007-4559.
If you have further questions you may contact us through this
projects lead researcher Kasimir Schulz.
Change-Id: I891ac6652cfbd479aed51d64ef6d4e0fe740e06d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65271
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Previously, shared memory server remove old socket *before* filling the
target path into API's data structure. However, the target path might
get truncated hence the path we check against might not be the one we
will be using in the end.
In a case where the path specified by user is free while the truncated
path is in used, gem5 will get a mysterious EADDRINUSE.
We swap the two steps in the CL, so we'll be checking against the actual
path we use, instead of the path user request to use.
Change-Id: Ib34f8b00ea1d2f15dcd4e7b6d2d4a6d6ddc4e411
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65153
Reviewed-by: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
This changeset replicates some of the multiprocessing module
implementation from the python standard library in gem5. The goal of
this and following changesets is to enable users to use a *single* set
of python scripts to run and analyze a suite of gem5 simulations.
We must reimplement some of the multiprocessing module becaue it is not
flexible enough to allow for customized command line parameter to the
"python" executable (gem5 in our case). To get around this, I extended
the Process and context objects to be gem5 specific.
The next steps is to wrap the Process and Pool types with gem5-specific
versions that will improve their usability for our needs. With this
changeset, these objects are usable, but it will require significant
user effort to reach the goal of running/analyzing many different gem5
simulations.
There are some limitation:
- The pool will only work if the max tasks per child is 1
- The functions that are executed must come from another module
As an example, the following code should work after applying this
change.
test.py:
```python
from gem5.utils.multiprocessing import Process, Pool
from sim import info, run_sim
if __name__ == '__m5_main__' or __name__ == '__main__':
info('main line')
p1 = Process(target=run_sim, args=('bob',))
p2 = Process(target=run_sim, args=('jane',))
p1.start()
p2.start()
p2.join()
p1.join()
with Pool(processes=4, maxtasksperchild=1) as pool:
pool.map(run_sim, range(10))
```
sim.py:
```
import os
def info(title):
print(title)
print('module name:', __name__)
print('parent process:', os.getppid())
print('process id:', os.getpid())
def run_sim(name):
info('function g')
from gem5.prebuilt.demo.x86_demo_board import X86DemoBoard
from gem5.resources.resource import Resource
from gem5.simulate.simulator import Simulator
board = X86DemoBoard()
board.set_kernel_disk_workload(
kernel=Resource("x86-linux-kernel-5.4.49"),
disk_image=Resource("x86-ubuntu-18.04-img"),
)
simulator = Simulator(board=board)
simulator.run(max_ticks=10000000)
```
Change-Id: I4348ebaa75d006949ec96e732f5dc2a5173c6048
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/63432
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Richard Cooper <richard.cooper@arm.com>
This change enables OPOST to enable output post-processing. It then
enables ONLCR to prepend newline characters with carriage return so
that start of each line is always left aligned. Note that on some
terminals this might display a redundant ^M.
Change-Id: Ia0b4c61725ab7478e7341273a8279b96e53d9f26
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65152
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The ARM Boot Tests required the compilation of ARM/gem5.opt to run a
quick test that the CHI protocol was functioning correctly with ARM and
the ArmBoard. This test has been removed and the test refactored
slightly to use the ALL/gem5.opt.
The CHI protocol is already tested nightly.
Change-Id: Ibe406348caefa2493860036eb89a20681478ea66
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65195
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
The `main.py` script will build the ISAs required to run tests. Our
compiler tests (see "tests/compiler-tests.sh") are run nightly and
already test to ensure these ISAs are compiled correctly. Compiling
these ISAs as part of this script is therefore redundant. This patch
removes this step to save testing time.
Change-Id: I58636acfd5512886ac11ca84ee96cbdc9e344c68
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65175
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The only tests requiring the the compilation of SPARC, MIPS, and POWER
for the long/nightly suite were
"tests/gem5/multi_isa/test_multi_isa.py" and
"gem5/stdlib/test_requires.py". As compilation of gem5 is quite costly,
it'd best we simply remove these tests. They are minor and not very
important.
Compilation of these ISAs will continue to be tested via the compilation
tests.
Change-Id: I98b33eec5d0adb144109d32851033380f1641ad4
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65193
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
The long/nightly tests are failing due to timeout (e.g.,:
https://jenkins.gem5.org/job/nightly/398/). We must therefore be more
careful about what we test on a nightly basis.
Each of these X86 boot tests takes an hour and, generally, are largely
the same (just with different CPU cores, cache hierarchy, and memory
system). Given this is largely redundant, some of these tests have been
remove dto save on testing time.
Change-Id: I761fca1aa5e111a03183f83d4e326aaea1bdbc3a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65192
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
The long/nightly tests are failing due to timeout (e.g.:
https://jenkins.gem5.org/job/nightly/398/). We must therefore be more
careful about what we test on a nightly basis.
Each of these ARM boot tests takes an hour and, generally, are largely
the same (just with different CPU cores, cache hierarchy, and memory
system). Given this is largely redundant, some of these tests have been
removed to save on testing time.
Change-Id: I8d80d3e0869aca67aa7279a164bdce85d20f3682
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65191
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The "config/example/gem5_library/arm-ubuntu-boot-exit.py" script is
renamed to "config/example/gem5_library/arm-ubuntu-run.py". This makes
it more consistent with similar scripts in the
"config/example/gem5_library" directory: "x86-ubuntu-run.py" and
"riscv-ubuntu-run.py".
Change-Id: I9d96fd68e122f2841573b1717b0969cd44972771
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65132
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Since moving `_connect_things` to a pre-init step, the ArmBoard can now
be refactored to set up things in a more logical manner. In particular,
this patch moves activity out of the `_add_disk_to_board` function and
into the `_pre_initialization` function.
Change-Id: I5d40267f28ae87cd483a0396739c09b8b2b46383
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65052
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Through working with the gem5 stdlib there have been instances where
connecting the memory, processor, and cache hierarchy to the board (via
the AbstractBoard's `_connect_things` function) at the point of the
AbstractBoard's construction is problematic as the memory, processor,
and cache hierarchy may require information to connect correctly that is
only known to the AbstractBoard after construction. In particular this
can occur when a Workload contains information needed to configure
correctly.
To resolve this problem the `_connect_things` function has been moved to
run as a pre-initialization step. That is, run immediately before
`m5.instantiate`. This is done in the Simulator module.
This will break cases where a user utilizes the stdlib AbstractBoard but
does not use the stdlib Simulator module. As such, an Exception is
raised in these cases explaining the fix to the user. This is done via a
hack where the boards' `createCCObject` function (inheritted
from SimObject) is overriden with a check to ensure `_connect_things`
has been run. To fix the `_pre_instantiate` function must be executed
prior to `m5.instantiate` in the Python configuration script. Test and
config scripts in the gem5 repo have been updated accordingly.
Change-Id: Ibaef36eb7433ce104b861b1da80fc600f08f715a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65051
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
It is possible that the board has more than just a "main" memory. For
instance, the ArmBoard has a boot memory which is separate from the
`get_memory` function.
This moves the `get_mem_ports` function to the board so that the board
can optionally override it.
Change-Id: I05e388cc93e691e9a4fa674023f158af447349f9
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/64631
Reviewed-by: Bobby Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
The KVM CPU hangs if there are not multiple event queues when more than
one CPU is created. Since GPUFS primarily relies on the KVM CPU, support
for multiple event queues is needed. Some GPU libraries, such as AMD
Research's ATMI library, assume more than one CPU.
This changeset adds support for multiple CPUs and was tested for up to
four CPUs.
Change-Id: Ia354e02209d0fa18195f3ad44f4fb1d58e93b5ca
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65131
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
The PM4 release_mem packet is used as a DMA fence in the driver. It
specifies which queue the interrupt came from by encoding the me, pipe,
and queue fields from the map_queue packet into the interrupt ring ID.
Currently these fields are incorrect because (1) the order in the
bitfield is backwards, (2) the queue constructor assigns a pointer to
the PM4MapQueue packet containing this data to the dmaBuffer which gets
deleted in short order, and (3) the order of the encoding of ring ID is
incorrect.
This change fixes these issues by (1) placing the struct vales in
correct order, (2) creating a const copy of the dmaBuffer on
construction, and (3) using the ring ID encoding expected by the driver:
https://github.com/RadeonOpenCompute/ROCK-Kernel-Driver/blob/roc-4.3.x/
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c#L5989
Change-Id: I72c382980e57573f8a8a6879912c4139c7e2f505
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65095
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
The PM4 NOP header is used to insert spaces in the PM4 ring and can
therefore be any size. This includes zero. A size of zero is denoted by
a value of 0x3fff in the NOP packet header. Currently we assume this
means the remainder of the PM4 queue up to the wptr is empty/NOPs. This
is not always true.
This changeset reworks the PM4 NOP packet to handle the value of 0x3fff
as a special value and advances the rptr by 0 bytes. This fixes issues
where there were additional packets in the queue which were being
skipped over by fast forwarding. Since those packets could be anything,
that leads to undefined behavior afterwards.
Change-Id: I3f5c3f4b7dd50f93ba503fea97454a9d41771e30
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65094
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
SDMA traps are used in the driver as a DMA fence. To pass a fence, the
SDMA sends the driver the interrupt context from a trap packet and the
ring ID which specifies which queue in the SDMA engine is passing a
fence. Currently the interrupt context is using the wrong value in the
packet and the ring ID is hard-coded to always be the gfx queue.
This changeset uses the correct interrupt context from the SDMA packet
and sets the ring ID to either 0 if the gfx queue is currently being
processed or 3 if the page queue is being processed.
The relevant interrupt service routine in the driver can be found at:
https://github.com/RadeonOpenCompute/ROCK-Kernel-Driver/blob/roc-4.3.x/
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c#L2129
Change-Id: Ie4a4a9d6ab1d3bf83bf76bb57a02a91100217b51
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65093
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The interrupt handler's base address is sent via MMIO and must be
shifted by 8 bits to convert to a byte address. The current code is
shifting the MMIO dword first then assigning, resulting in the top 8
bits being shifted out.
This changeset fixes the issue by assigning the dword to the 64-bit
address first then shifting after. Similarly, the upper dword is cast to
a 64-bit value first before shifting.
This fixes some "fence fallback timeout" errors in the m5term output.
These timeouts become a problem because the driver will reset after a
few hundred of them, killing any running GPU applications as part of the
process.
Change-Id: I0beec313f533765c94063bcf4de8c65aacf2986b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65092
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
The GART table is a legacy 1-level page table primarily used for
supervisor mode accesses to GPUs. The PTE size is 64-bits, not 32-bit.
This causes memory sizes >3GB (in X86) to fail loading amdgpu driver.
This changeset fixes the issue by setting the GART mappings to the
correct data type.
Change-Id: Ibfba2443675fe28316d26afa5f1a14885fdce40c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65091
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
The current implementation of SDMA copy calls the GPU memory manager's
read/write method one time passing a physical address as the
source/destination. This implicitly assumes the physical addresses are
contiguous which is generally not true for large allocations. This
results in reading from/writing to the wrong address.
This changeset fixes the problem by copying large copies in chunks of
the minimum possible page size on the GPU (4kB). Each page is translated
seperately to ensure the correct physical address. The final copy "done"
callback is only used for the last transfer. The transfers should
complete in order so the copy command will not complete until all chunks
have been copied. Tested and verified on an application with a large
allocation (~5GB).
Change-Id: I27018a963da7133f5e49dec13b0475c3637c8765
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/64752
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Vega allows for any integer multiple of 4kB pages. However, the current
implementation is designed for 4kB page primarily. In order to support
variable page sizes, the physical address calculation needs to be
updated to add the virtual page offset to the base physical address
rather than bitwise-OR. Bitwise-OR assumes physical pages are at
aligned to the page size which is generally not the case for very
large pages (1GB+).
This changeset changes all of the physical address computations to add
the virtual offset to the physical page address. This fixes many GPUFS
applications which use larger pages. The support was tested by
hipMalloc'ing ~5GB to induce a large page being created. The test
application now passes verification with this change.
Change-Id: Ic8d1475e001def443f3e4ab609449bca0c40b638
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/64751
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
