Hi Andrew,

I looked at that thread again, thanks.  However, I'm not sure how it is
relevant to your issue -- that thread says they are using ROCm 1.6.4, but
you said you are using 1.6.1.  And yes, you need to compile with gfx801,
but that seems secondary.  Perhaps I'm missing something?

Thanks,
Matt

On Sun, Nov 24, 2019 at 6:27 PM Andrew Chu <chuchih...@gmail.com> wrote:

> Hi Matt,
>
> Thanks for your suggestions. Refer to this thread
> https://www.mail-archive.com/gem5-users@gem5.org/msg16276.html,
> with "--amdgpu-target=gfx801" to the link flag, both CoMD and lulesh can
> run to complete
>
> Matt Sinclair <mattdsincl...@gmail.com> 於 2019年11月22日 週五 下午11:48寫道:
>
>> Hi Andrew,
>>
>> Normally this error indicates that something with your ROCm setup is
>> wrong -- you can look through the mailing list archive and see a few
>> examples of people with the same types of errors along with the suggestions
>> (e.g., what kind of traces) to get to start debugging the problem.  Why did
>> you install ROCm 1.6.1?  If you look at the wiki, AMD suggests using 1.6.x
>> instead: http://gem5.org/GPU_Models#ROCm_tool_chain_and_software_stack.
>> I can't promise that changing from 1.6.1 to 1.6.x will solve your problem,
>> but it would be a good start/thing to check.  Also, I'm not familiar with
>> this sample program -- does it assume the device is a discrete GPU?  If so,
>> the gem5 AMD GPU model assumes the GPU is tightly coupled and cache
>> coherent.  So you would need to remove those copies (although it appears
>> your failure is happening sooner than that).
>>
>> Thanks,
>> Matt
>>
>> On Fri, Nov 15, 2019 at 2:17 AM Andrew Chu <chuchih...@gmail.com> wrote:
>>
>>> I got a GEM5 fatal error which syscall mincore was not implemented. Does
>>> anyone have the same experience about this?
>>>
>>> =========================
>>>
>>> Environment setting:
>>> Host OS: Ubuntu 16.04.6 LTS
>>> GEM5: git clone https://gem5.googlesource.com/amd/gem5
>>> -b agutierr/master-gcn3-staging
>>> ROCM: http://repo.radeon.com/rocm/archive/apt_1.6.1.tar.bz2
>>> SE test program:
>>> https://github.com/ROCm-Developer-Tools/HIP-Examples/tree/master/HIP-Examples-Applications/HelloWorld
>>> with HIP_PLATFORM=hcc
>>> CMD: ./build/GCN3_X86/gem5.opt --debug-flag=GPUALL
>>> configs/example/apu_se.py -n 2 -c tests/test-progs/hip-hello/HelloWorld
>>> Result:
>>> .
>>> .
>>> .
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> warn: ignoring syscall sched_yield(...)
>>> fatal: syscall mincore (#27) unimplemented.
>>>
>>> ===========================
>>>
>>> Once I ignore this system call, the test program will have run time error
>>>
>>> warn: ignoring syscall mincore(...)
>>> warn: ignoring syscall mincore(...)
>>>
>>> Backtrace:
>>> 0x000000000040d941:     <unknown function> + 0x5
>>> 0x000000000040dc50:     <unknown function> + 0x5
>>> 0x000000000040db09:     <unknown function> + 0x5
>>> 0x0000000000426bdd:     <unknown function> + 0x5
>>> 0x00007ffff5f517bf:     <unknown function> + 0x5
>>> 0x000000000041b509:     <unknown function> + 0x5
>>>
>>> ### HCC RUNTIME ERROR: Fail to find compatible kernel at
>>> file:/home/jenkins/jenkins-root/workspace/compute-rocm-rel-1.6/external/hcc-tot/lib/mcwamp.cpp
>>> line:346
>>> _______________________________________________
>>> 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
>
> _______________________________________________
> 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

Reply via email to