I've been trying to dig in more in the code... I think there is a bug
when the ArmTLB is trying to switch to the Ruby memory subsystem.
Is anyone having any trouble with this aswell?
Carlos
On 21/2/20 15:59, Ciro Santilli wrote:
I observe the following behavior on se.py ARM and X86, which depends on
--restore-with-cpu. I have used a minimal userland executable that only
does two things in assembly: m5 checkpoint and m5 exit (total of only 5
instructions in aarch64)
@Jason: can you confirm the difference between se.py --restore-with-cpu
and --cpu-type during a checkpoint restore (and if any of the outcome
below are buggy or not)? At
https://stackoverflow.com/questions/49011096/how-to-switch-cpu-models-in-gem5-after-restoring-a-checkpoint-and-then-observe-t
it was mentioned that --restore-with-cpu
should be the old CPU (defaults to AtomicSimpleCPU), and --cpu-type the
new CPU, is that correct? Asking because I have observed less failures
if I set --restore-with-cpu to equal --cpu-type.
I'll open/update clear bugs with the test binaries for any behavior
which is not correct.
== Default Build
Restore with: --cpu-type DerivO3CPU
Outcome: gem5 crashes: panic: Attempted to execute unknown instruction
(inst 0x00000000)
Restore with: --cpu-type DerivO3CPU --restore-with-cpu DerivO3CPU
Outcome: works
== MESI_Three_Level build
Checkpoint run with: --ruby
Outcome: hangs forever when taking the checkpoint.
Checkpoint run at the commit before
bb94296373dde1d0ce971ee58ad111f4225c425e (which
https://gem5.atlassian.net/projects/GEM5/issues/GEM5-331 claims broke
Ruby checkpointing on MOESI_hammer):
Outcome: "panic: Runtime Error at MESI_Three_Level-L0cache.sm:249:
Invalid RubyRequestType." This is the correct outcome because
MESE_Three_Level does have flush operations which are needed to create
checkpoints.
Checkpoint run: without --ruby
Restore run: --ruby --cpu-type DerivO3CPU
Outcome: gem5 crashes with: build/ARM/sim/port.hh:131: void
Port::takeOverFrom(Port*): Assertion `old->isConnected()' failed.
Checkpoint run: without --ruby
Restore run: --ruby --cpu-type DerivO3CPU --restore-with-cpu DerivO3CPU
Outcome: works
Checkpoint run on a C hello world without m5ops: --fast-forward 1000
--ruby --cpu-type DerivO3CPU --restore-with-cpu DerivO3CPU
Outcome: build/ARM/sim/port.hh:131: void Port::takeOverFrom(Port*):
Assertion `old->isConnected()' failed.
== MOESI_hammer build
Same as MESI_Three_Level, except that the checkpoint with --ruby works
before bb94296373dde1d0ce971ee58ad111f4225c425e as expected.
On 2/12/20 9:45 AM, Carlos Escuin wrote:
Thank you for replying,
Yes, it seems that something is not going well while switching from
AtomicSimpleCPU to DerivO3CPU.
Carlos
On 11/2/20 21:48, Giacomo Travaglini wrote:
Seems like there is an issue when you are switching over the new cpu
model.
Will have a look into that
Giacomo
------------------------------------------------------------------------
*From:* gem5-users <gem5-users-boun...@gem5.org> on behalf of Carlos
Escuin <esc...@unizar.es>
*Sent:* 11 February 2020 16:46
*To:* gem5-users@gem5.org <gem5-users@gem5.org>
*Subject:* [gem5-users] ARM DerivO3CPU assertion failed
Hi all,
I'm trying to execute bzip2 spec 2006 benchmark in ARM, se, DerivO3CPU,
ruby, fast-forward.
Anyone has any idea why I'm getting this assertion failing?
Thank you,
Carlos
OUTPUT:
command line: gem5/build/ARM_MOESI_CMP_directory/gem5.opt -v
--outdir=gem5/m5out/spec2k6/bzip2_base gem5/configs/example/se.py
--num-cpus=1 --cmd=CPU2006_ARM/bzip2_dir/bzip2 --input= --output=
'--options=CPU2006_ARM/bzip2_dir/input.source 280'
--fast-forward=5000000 --maxinsts=12000000 --cpu-type=DerivO3CPU --ruby
--l1d_size=32kB --l1i_size=32kB --l1d_assoc=8 --l1i_assoc=8
--l2_size=1MB --l2_assoc=16 --network=garnet2.0
info: Standard input is not a terminal, disabling listeners.
Global frequency set at 1000000000000 ticks per second
warn: Sockets disabled, not accepting gdb connections
Switch at instruction count:5000000
info: Entering event queue @ 0. Starting simulation...
info: Increasing stack size by one page.
Switched CPUS @ tick 2977722500
switching cpus
warn: ClockedObject: Already in the requested power state, request
ignored
warn: User mode does not have SPSR
warn: User mode does not have SPSR
gem5.opt: build/ARM_MOESI_CMP_directory/sim/port.hh:135: void
Port::takeOverFrom(Port*): Assertion `old->isConnected()' failed.
Program aborted at tick 2977722500
--- BEGIN LIBC BACKTRACE ---
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(_Z15print_backtracev+0x15)[0x1281115]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(_Z12abortHandleri+0x36)[0x128b526]
/lib64/libpthread.so.0(+0xf7e0)[0x2afcc8dfd7e0]
/lib64/libc.so.6(gsignal+0x35)[0x2afcc99ea4f5]
/lib64/libc.so.6(abort+0x175)[0x2afcc99ebcd5]
/lib64/libc.so.6(+0x2b66e)[0x2afcc99e366e]
/lib64/libc.so.6(__assert_perror_fail+0x0)[0x2afcc99e3730]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(_ZN7BaseCPU19updateCycleCountersENS_8CPUStateE+0x0)[0x10557f0]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(_ZN7BaseCPU12takeOverFromEPS_+0x209)[0x10469e9]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(_ZN9FullO3CPUI9O3CPUImplE12takeOverFromEP7BaseCPU+0x11)[0x11da821]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt[0xe4ac96]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt[0xdfb79f]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x74bf)[0x2afcc8ab374f]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x80d)[0x2afcc8ab5a7d]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x6962)[0x2afcc8ab2bf2]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x80d)[0x2afcc8ab5a7d]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x6962)[0x2afcc8ab2bf2]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x80d)[0x2afcc8ab5a7d]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x6962)[0x2afcc8ab2bf2]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x80d)[0x2afcc8ab5a7d]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCode+0x32)[0x2afcc8ab5bb2]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x8f7b)[0x2afcc8ab520b]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x80d)[0x2afcc8ab5a7d]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalFrameEx+0x6962)[0x2afcc8ab2bf2]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCodeEx+0x80d)[0x2afcc8ab5a7d]
python-2.7.15/libpython2.7.so.1.0(PyEval_EvalCode+0x32)[0x2afcc8ab5bb2]
python-2.7.15/libpython2.7.so.1.0(PyRun_StringFlags+0x79)[0x2afcc8ae17e9]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(_Z6m5MainiPPc+0x5f)[0x128a2ef]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt(main+0x38)[0xa9aee8]
/lib64/libc.so.6(__libc_start_main+0x100)[0x2afcc99d6d20]
gem5/build/ARM_MOESI_CMP_directory/gem5.opt[0xa9ad79]
--- END LIBC BACKTRACE ---
./execute_se_arm.sh: line 296: 12602 Aborted (core
dumped) gem5/build/ARM_MOESI_CMP_directory/gem5.opt -v
--outdir=gem5/m5out/spec2k6/bzip2_base gem5/configs/example/se.py
--num-cpus=1 --cmd="CPU2006_ARM/bzip2_dir/bzip2" --input="" --output=""
--options="CPU2006_ARM/bzip2_dir/input.source 280"
--fast-forward=5000000 --maxinsts=12000000 --cpu-type=DerivO3CPU --ruby
--l1d_size=32kB --l1i_size=32kB --l1d_assoc=8 --l1i_assoc=8
--l2_size=1MB --l2_assoc=16 --network=garnet2.0
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
IMPORTANT NOTICE: The contents of this email and any attachments are
confidential and may also be privileged. If you are not the intended
recipient, please notify the sender immediately and do not disclose
the contents to any other person, use it for any purpose, or store or
copy the information in any medium. Thank you.
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
IMPORTANT NOTICE: The contents of this email and any attachments are
confidential and may also be privileged. If you are not the intended
recipient, please notify the sender immediately and do not disclose
the contents to any other person, use it for any purpose, or store or
copy the information in any medium. Thank you.
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
