I am on the petsc-main
commit 1390d3a27d88add7d79c9b38bf1a895ae5e67af6
Merge: 96c919c d5f3255
Author: Satish Balay <ba...@mcs.anl.gov>
Date: Wed Jan 26 10:28:32 2022 -0600
Merge remote-tracking branch 'origin/release'
It is still broken.
Thanks,
Fande
On Wed, Jan 26, 2022 at 7:40 AM Junchao Zhang <junchao.zh...@gmail.com>
wrote:
> The good uses the compiler's default library/header path. The bad
> searches from cuda toolkit path and uses rpath linking.
> Though the paths look the same on the login node, they could have
> different behavior on a compute node depending on its environment.
> I think we fixed the issue in cuda.py (i.e., first try the compiler's
> default, then toolkit). That's why I wanted Fande to use petsc/main.
>
> --Junchao Zhang
>
>
> On Tue, Jan 25, 2022 at 11:59 PM Barry Smith <bsm...@petsc.dev> wrote:
>
>>
>> bad has extra
>>
>> -L/apps/local/spack/software/gcc-7.5.0/cuda-10.1.243-v4ymjqcrr7f72qfiuzsstuy5jiajbuey/lib64/stubs
>> -lcuda
>>
>> good does not.
>>
>> Try removing the stubs directory and -lcuda from the bad
>> $PETSC_ARCH/lib/petsc/conf/variables and likely the bad will start working.
>>
>> Barry
>>
>> I never liked the stubs stuff.
>>
>> On Jan 25, 2022, at 11:29 PM, Fande Kong <fdkong...@gmail.com> wrote:
>>
>> Hi Junchao,
>>
>> I attached a "bad" configure log and a "good" configure log.
>>
>> The "bad" one was on produced at 246ba74192519a5f34fb6e227d1c64364e19ce2c
>>
>> and the "good" one at 384645a00975869a1aacbd3169de62ba40cad683
>>
>> This good hash is the last good hash that is just the right before the
>> bad one.
>>
>> I think you could do a comparison between these two logs, and check what
>> the differences were.
>>
>> Thanks,
>>
>> Fande
>>
>> On Tue, Jan 25, 2022 at 8:21 PM Junchao Zhang <junchao.zh...@gmail.com>
>> wrote:
>>
>>> Fande, could you send the configure.log that works (i.e., before this
>>> offending commit)?
>>> --Junchao Zhang
>>>
>>>
>>> On Tue, Jan 25, 2022 at 8:21 PM Fande Kong <fdkong...@gmail.com> wrote:
>>>
>>>> Not sure if this is helpful. I did "git bisect", and here was the
>>>> result:
>>>>
>>>> [kongf@sawtooth2 petsc]$ git bisect bad
>>>> 246ba74192519a5f34fb6e227d1c64364e19ce2c is the first bad commit
>>>> commit 246ba74192519a5f34fb6e227d1c64364e19ce2c
>>>> Author: Junchao Zhang <jczh...@mcs.anl.gov>
>>>> Date: Wed Oct 13 05:32:43 2021 +0000
>>>>
>>>> Config: fix CUDA library and header dirs
>>>>
>>>> :040000 040000 187c86055adb80f53c1d0565a8888704fec43a96
>>>> ea1efd7f594fd5e8df54170bc1bc7b00f35e4d5f M config
>>>>
>>>>
>>>> Started from this commit, and GPU did not work for me on our HPC
>>>>
>>>> Thanks,
>>>> Fande
>>>>
>>>> On Tue, Jan 25, 2022 at 7:18 PM Fande Kong <fdkong...@gmail.com> wrote:
>>>>
>>>>>
>>>>>
>>>>> On Tue, Jan 25, 2022 at 9:04 AM Jacob Faibussowitsch <
>>>>> jacob....@gmail.com> wrote:
>>>>>
>>>>>> Configure should not have an impact here I think. The reason I had
>>>>>> you run `cudaGetDeviceCount()` is because this is the CUDA call (and in
>>>>>> fact the only CUDA call) in the initialization sequence that returns the
>>>>>> error code. There should be no prior CUDA calls. Maybe this is a problem
>>>>>> with oversubscribing GPU’s? In the runs that crash, how many ranks are
>>>>>> using any given GPU at once? Maybe MPS is required.
>>>>>>
>>>>>
>>>>> I used one MPI rank.
>>>>>
>>>>> Fande
>>>>>
>>>>>
>>>>>
>>>>>>
>>>>>> Best regards,
>>>>>>
>>>>>> Jacob Faibussowitsch
>>>>>> (Jacob Fai - booss - oh - vitch)
>>>>>>
>>>>>> On Jan 21, 2022, at 12:01, Fande Kong <fdkong...@gmail.com> wrote:
>>>>>>
>>>>>> Thanks Jacob,
>>>>>>
>>>>>> On Thu, Jan 20, 2022 at 6:25 PM Jacob Faibussowitsch <
>>>>>> jacob....@gmail.com> wrote:
>>>>>>
>>>>>>> Segfault is caused by the following check at
>>>>>>> src/sys/objects/device/impls/cupm/cupmdevice.cxx:349 being a
>>>>>>> PetscUnlikelyDebug() rather than just PetscUnlikely():
>>>>>>>
>>>>>>> ```
>>>>>>> if (PetscUnlikelyDebug(_defaultDevice < 0)) { // _defaultDevice is
>>>>>>> in fact < 0 here and uncaught
>>>>>>> ```
>>>>>>>
>>>>>>> To clarify:
>>>>>>>
>>>>>>> “lazy” initialization is not that lazy after all, it still does some
>>>>>>> 50% of the initialization that “eager” initialization does. It stops
>>>>>>> short
>>>>>>> initializing the CUDA runtime, checking CUDA aware MPI, gathering device
>>>>>>> data, and initializing cublas and friends. Lazy also importantly
>>>>>>> swallows
>>>>>>> any errors that crop up during initialization, storing the resulting
>>>>>>> error
>>>>>>> code for later (specifically _defaultDevice = -init_error_value;).
>>>>>>>
>>>>>>> So whether you initialize lazily or eagerly makes no difference
>>>>>>> here, as _defaultDevice will always contain -35.
>>>>>>>
>>>>>>> The bigger question is why cudaGetDeviceCount() is returning
>>>>>>> cudaErrorInsufficientDriver. Can you compile and run
>>>>>>>
>>>>>>> ```
>>>>>>> #include <cuda_runtime.h>
>>>>>>>
>>>>>>> int main()
>>>>>>> {
>>>>>>> int ndev;
>>>>>>> return cudaGetDeviceCount(&ndev):
>>>>>>> }
>>>>>>> ```
>>>>>>>
>>>>>>> Then show the value of "echo $?”?
>>>>>>>
>>>>>>
>>>>>> Modify your code a little to get more information.
>>>>>>
>>>>>> #include <cuda_runtime.h>
>>>>>> #include <cstdio>
>>>>>>
>>>>>> int main()
>>>>>> {
>>>>>> int ndev;
>>>>>> int error = cudaGetDeviceCount(&ndev);
>>>>>> printf("ndev %d \n", ndev);
>>>>>> printf("error %d \n", error);
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> Results:
>>>>>>
>>>>>> $ ./a.out
>>>>>> ndev 4
>>>>>> error 0
>>>>>>
>>>>>>
>>>>>> I have not read the PETSc cuda initialization code yet. If I need to
>>>>>> guess at what was happening. I will naively think that PETSc did not get
>>>>>> correct GPU information in the configuration because the compiler node
>>>>>> does
>>>>>> not have GPUs, and there was no way to get any GPU device information.
>>>>>>
>>>>>>
>>>>>> During the runtime on GPU nodes, PETSc might have incorrect
>>>>>> information grabbed during configuration and had this kind of false error
>>>>>> message.
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Fande
>>>>>>
>>>>>>
>>>>>>
>>>>>>>
>>>>>>> Best regards,
>>>>>>>
>>>>>>> Jacob Faibussowitsch
>>>>>>> (Jacob Fai - booss - oh - vitch)
>>>>>>>
>>>>>>> On Jan 20, 2022, at 17:47, Matthew Knepley <knep...@gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>> On Thu, Jan 20, 2022 at 6:44 PM Fande Kong <fdkong...@gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Thanks, Jed
>>>>>>>>
>>>>>>>> On Thu, Jan 20, 2022 at 4:34 PM Jed Brown <j...@jedbrown.org> wrote:
>>>>>>>>
>>>>>>>>> You can't create CUDA or Kokkos Vecs if you're running on a node
>>>>>>>>> without a GPU.
>>>>>>>>
>>>>>>>>
>>>>>>>> I am running the code on compute nodes that do have GPUs.
>>>>>>>>
>>>>>>>
>>>>>>> If you are actually running on GPUs, why would you need lazy
>>>>>>> initialization? It would not break with GPUs present.
>>>>>>>
>>>>>>> Matt
>>>>>>>
>>>>>>>
>>>>>>>> With PETSc-3.16.1, I got good speedup by running GAMG on GPUs.
>>>>>>>> That might be a bug of PETSc-main.
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>>
>>>>>>>> Fande
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> KSPSetUp 13 1.0 6.4400e-01 1.0 2.02e+09 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 5 0 0 0 0 5 0 0 0 3140 64630 15
>>>>>>>> 1.05e+02 5 3.49e+01 100
>>>>>>>> KSPSolve 1 1.0 1.0109e+00 1.0 3.49e+10 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 87 0 0 0 0 87 0 0 0 34522 69556 4
>>>>>>>> 4.35e-03 1 2.38e-03 100
>>>>>>>> KSPGMRESOrthog 142 1.0 1.2674e-01 1.0 1.06e+10 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 27 0 0 0 0 27 0 0 0 83755 87801 0
>>>>>>>> 0.00e+00 0 0.00e+00 100
>>>>>>>> SNESSolve 1 1.0 4.4402e+01 1.0 4.00e+10 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 21100 0 0 0 21100 0 0 0 901 51365 57
>>>>>>>> 1.10e+03 52 8.78e+02 100
>>>>>>>> SNESSetUp 1 1.0 3.9101e-05 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> SNESFunctionEval 2 1.0 1.7097e+01 1.0 1.60e+07 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 8 0 0 0 0 8 0 0 0 0 1 0 0
>>>>>>>> 0.00e+00 6 1.92e+02 0
>>>>>>>> SNESJacobianEval 1 1.0 1.6213e+01 1.0 2.80e+07 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 8 0 0 0 0 8 0 0 0 0 2 0 0
>>>>>>>> 0.00e+00 1 3.20e+01 0
>>>>>>>> SNESLineSearch 1 1.0 8.5582e+00 1.0 1.24e+08 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 4 0 0 0 0 4 0 0 0 0 14 64153 1
>>>>>>>> 3.20e+01 3 9.61e+01 94
>>>>>>>> PCGAMGGraph_AGG 5 1.0 3.0509e+00 1.0 8.19e+07 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 1 0 0 0 0 1 0 0 0 0 27 0 5
>>>>>>>> 3.49e+01 9 7.43e+01 0
>>>>>>>> PCGAMGCoarse_AGG 5 1.0 3.8711e+00 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 2 0 0 0 0 2 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> PCGAMGProl_AGG 5 1.0 7.0748e-01 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> PCGAMGPOpt_AGG 5 1.0 1.2904e+00 1.0 2.14e+09 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 1 5 0 0 0 1 5 0 0 0 1661 29807 26
>>>>>>>> 7.15e+02 20 2.90e+02 99
>>>>>>>> GAMG: createProl 5 1.0 8.9489e+00 1.0 2.22e+09 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 4 6 0 0 0 4 6 0 0 0 249 29666 31
>>>>>>>> 7.50e+02 29 3.64e+02 96
>>>>>>>> Graph 10 1.0 3.0478e+00 1.0 8.19e+07 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 1 0 0 0 0 1 0 0 0 0 27 0 5
>>>>>>>> 3.49e+01 9 7.43e+01 0
>>>>>>>> MIS/Agg 5 1.0 4.1290e-01 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> SA: col data 5 1.0 1.9127e-02 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> SA: frmProl0 5 1.0 6.2662e-01 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> SA: smooth 5 1.0 4.9595e-01 1.0 1.21e+08 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 244 2709 15
>>>>>>>> 1.97e+02 15 2.55e+02 90
>>>>>>>> GAMG: partLevel 5 1.0 4.7330e-01 1.0 6.98e+08 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 2 0 0 0 0 2 0 0 0 1475 4120 5
>>>>>>>> 1.78e+02 10 2.55e+02 100
>>>>>>>> PCGAMG Squ l00 1 1.0 2.6027e+00 1.0 0.00e+00 0.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 1 0 0 0 0 1 0 0 0 0 0 0 0
>>>>>>>> 0.00e+00 0 0.00e+00 0
>>>>>>>> PCGAMG Gal l00 1 1.0 3.8406e-01 1.0 5.48e+08 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 1 0 0 0 0 1 0 0 0 1426 4270 1
>>>>>>>> 1.48e+02 2 2.11e+02 100
>>>>>>>> PCGAMG Opt l00 1 1.0 2.4932e-01 1.0 7.20e+07 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 289 2653 1
>>>>>>>> 6.41e+01 1 1.13e+02 100
>>>>>>>> PCGAMG Gal l01 1 1.0 6.6279e-02 1.0 1.09e+08 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 1645 3851 1
>>>>>>>> 2.40e+01 2 3.64e+01 100
>>>>>>>> PCGAMG Opt l01 1 1.0 2.9544e-02 1.0 7.15e+06 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 242 1671 1
>>>>>>>> 4.84e+00 1 1.23e+01 100
>>>>>>>> PCGAMG Gal l02 1 1.0 1.8874e-02 1.0 3.72e+07 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 1974 3636 1
>>>>>>>> 5.04e+00 2 6.58e+00 100
>>>>>>>> PCGAMG Opt l02 1 1.0 7.4353e-03 1.0 2.40e+06 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 323 1457 1
>>>>>>>> 7.71e-01 1 2.30e+00 100
>>>>>>>> PCGAMG Gal l03 1 1.0 2.8479e-03 1.0 4.10e+06 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 1440 2266 1
>>>>>>>> 4.44e-01 2 5.51e-01 100
>>>>>>>> PCGAMG Opt l03 1 1.0 8.2684e-04 1.0 2.80e+05 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 339 1667 1
>>>>>>>> 6.72e-02 1 2.03e-01 100
>>>>>>>> PCGAMG Gal l04 1 1.0 1.2238e-03 1.0 2.09e+05 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 170 244 1
>>>>>>>> 2.05e-02 2 2.53e-02 100
>>>>>>>> PCGAMG Opt l04 1 1.0 4.1008e-04 1.0 1.77e+04 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 0 0 0 0 0 0 0 0 0 0 43 165 1
>>>>>>>> 4.49e-03 1 1.19e-02 100
>>>>>>>> PCSetUp 2 1.0 9.9632e+00 1.0 4.95e+09 1.0 0.0e+00
>>>>>>>> 0.0e+00 0.0e+00 5 12 0 0 0 5 12 0 0 0 496 17826 55
>>>>>>>> 1.03e+03 45 6.54e+02 98
>>>>>>>> PCSetUpOnBlocks 44 1.0 9.9087e-04 1.0 2.88e+03 1.0
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> The point of lazy initialization is to make it possible to run a
>>>>>>>>> solve that doesn't use a GPU in PETSC_ARCH that supports GPUs,
>>>>>>>>> regardless
>>>>>>>>> of whether a GPU is actually present.
>>>>>>>>>
>>>>>>>>> Fande Kong <fdkong...@gmail.com> writes:
>>>>>>>>>
>>>>>>>>> > I spoke too soon. It seems that we have trouble creating
>>>>>>>>> cuda/kokkos vecs
>>>>>>>>> > now. Got Segmentation fault.
>>>>>>>>> >
>>>>>>>>> > Thanks,
>>>>>>>>> >
>>>>>>>>> > Fande
>>>>>>>>> >
>>>>>>>>> > Program received signal SIGSEGV, Segmentation fault.
>>>>>>>>> > 0x00002aaab5558b11 in
>>>>>>>>> >
>>>>>>>>> Petsc::CUPMDevice<(Petsc::CUPMDeviceType)0>::CUPMDeviceInternal::initialize
>>>>>>>>> > (this=0x1) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/sys/objects/device/impls/cupm/cupmdevice.cxx:54
>>>>>>>>> > 54 PetscErrorCode
>>>>>>>>> CUPMDevice<T>::CUPMDeviceInternal::initialize() noexcept
>>>>>>>>> > Missing separate debuginfos, use: debuginfo-install
>>>>>>>>> > bzip2-libs-1.0.6-13.el7.x86_64 elfutils-libelf-0.176-5.el7.x86_64
>>>>>>>>> > elfutils-libs-0.176-5.el7.x86_64 glibc-2.17-325.el7_9.x86_64
>>>>>>>>> > libX11-1.6.7-4.el7_9.x86_64 libXau-1.0.8-2.1.el7.x86_64
>>>>>>>>> > libattr-2.4.46-13.el7.x86_64 libcap-2.22-11.el7.x86_64
>>>>>>>>> > libibmad-5.4.0.MLNX20190423.1d917ae-0.1.49224.x86_64
>>>>>>>>> > libibumad-43.1.1.MLNX20200211.078947f-0.1.49224.x86_64
>>>>>>>>> > libibverbs-41mlnx1-OFED.4.9.0.0.7.49224.x86_64
>>>>>>>>> > libmlx4-41mlnx1-OFED.4.7.3.0.3.49224.x86_64
>>>>>>>>> > libmlx5-41mlnx1-OFED.4.9.0.1.2.49224.x86_64
>>>>>>>>> libnl3-3.2.28-4.el7.x86_64
>>>>>>>>> > librdmacm-41mlnx1-OFED.4.7.3.0.6.49224.x86_64
>>>>>>>>> > librxe-41mlnx1-OFED.4.4.2.4.6.49224.x86_64
>>>>>>>>> libxcb-1.13-1.el7.x86_64
>>>>>>>>> > libxml2-2.9.1-6.el7_9.6.x86_64 numactl-libs-2.0.12-5.el7.x86_64
>>>>>>>>> > systemd-libs-219-78.el7_9.3.x86_64 xz-libs-5.2.2-1.el7.x86_64
>>>>>>>>> > zlib-1.2.7-19.el7_9.x86_64
>>>>>>>>> > (gdb) bt
>>>>>>>>> > #0 0x00002aaab5558b11 in
>>>>>>>>> >
>>>>>>>>> Petsc::CUPMDevice<(Petsc::CUPMDeviceType)0>::CUPMDeviceInternal::initialize
>>>>>>>>> > (this=0x1) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/sys/objects/device/impls/cupm/cupmdevice.cxx:54
>>>>>>>>> > #1 0x00002aaab5558db7 in
>>>>>>>>> > Petsc::CUPMDevice<(Petsc::CUPMDeviceType)0>::getDevice
>>>>>>>>> > (this=this@entry=0x2aaab7f37b70
>>>>>>>>> > <CUDADevice>, device=0x115da00, id=-35, id@entry=-1) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/sys/objects/device/impls/cupm/cupmdevice.cxx:344
>>>>>>>>> > #2 0x00002aaab55577de in PetscDeviceCreate (type=type@entry
>>>>>>>>> =PETSC_DEVICE_CUDA,
>>>>>>>>> > devid=devid@entry=-1, device=device@entry=0x2aaab7f37b48
>>>>>>>>> > <defaultDevices+8>) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/sys/objects/device/interface/device.cxx:107
>>>>>>>>> > #3 0x00002aaab5557b3a in
>>>>>>>>> PetscDeviceInitializeDefaultDevice_Internal
>>>>>>>>> > (type=type@entry=PETSC_DEVICE_CUDA,
>>>>>>>>> defaultDeviceId=defaultDeviceId@entry=-1)
>>>>>>>>> > at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/sys/objects/device/interface/device.cxx:273
>>>>>>>>> > #4 0x00002aaab5557bf6 in PetscDeviceInitialize
>>>>>>>>> > (type=type@entry=PETSC_DEVICE_CUDA)
>>>>>>>>> > at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/sys/objects/device/interface/device.cxx:234
>>>>>>>>> > #5 0x00002aaab5661fcd in VecCreate_SeqCUDA (V=0x115d150) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/vec/vec/impls/seq/seqcuda/veccuda.c:244
>>>>>>>>> > #6 0x00002aaab5649b40 in VecSetType (vec=vec@entry=0x115d150,
>>>>>>>>> > method=method@entry=0x2aaab70b45b8 "seqcuda") at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/vec/vec/interface/vecreg.c:93
>>>>>>>>> > #7 0x00002aaab579c33f in VecCreate_CUDA (v=0x115d150) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/vec/vec/impls/mpi/mpicuda/
>>>>>>>>> > mpicuda.cu:214
>>>>>>>>> > #8 0x00002aaab5649b40 in VecSetType (vec=vec@entry=0x115d150,
>>>>>>>>> > method=method@entry=0x7fffffff9260 "cuda") at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/vec/vec/interface/vecreg.c:93
>>>>>>>>> > #9 0x00002aaab5648bf1 in VecSetTypeFromOptions_Private
>>>>>>>>> (vec=0x115d150,
>>>>>>>>> > PetscOptionsObject=0x7fffffff9210) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/vec/vec/interface/vector.c:1263
>>>>>>>>> > #10 VecSetFromOptions (vec=0x115d150) at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/petsc/src/vec/vec/interface/vector.c:1297
>>>>>>>>> > #11 0x00002aaab02ef227 in libMesh::PetscVector<double>::init
>>>>>>>>> > (this=0x11cd1a0, n=441, n_local=441, fast=false,
>>>>>>>>> ptype=libMesh::PARALLEL)
>>>>>>>>> > at
>>>>>>>>> >
>>>>>>>>> /home/kongf/workhome/sawtooth/moosegpu/scripts/../libmesh/installed/include/libmesh/petsc_vector.h:693
>>>>>>>>> >
>>>>>>>>> > On Thu, Jan 20, 2022 at 1:09 PM Fande Kong <fdkong...@gmail.com>
>>>>>>>>> wrote:
>>>>>>>>> >
>>>>>>>>> >> Thanks, Jed,
>>>>>>>>> >>
>>>>>>>>> >> This worked!
>>>>>>>>> >>
>>>>>>>>> >> Fande
>>>>>>>>> >>
>>>>>>>>> >> On Wed, Jan 19, 2022 at 11:03 PM Jed Brown <j...@jedbrown.org>
>>>>>>>>> wrote:
>>>>>>>>> >>
>>>>>>>>> >>> Fande Kong <fdkong...@gmail.com> writes:
>>>>>>>>> >>>
>>>>>>>>> >>> > On Wed, Jan 19, 2022 at 11:39 AM Jacob Faibussowitsch <
>>>>>>>>> >>> jacob....@gmail.com>
>>>>>>>>> >>> > wrote:
>>>>>>>>> >>> >
>>>>>>>>> >>> >> Are you running on login nodes or compute nodes (I can’t
>>>>>>>>> seem to tell
>>>>>>>>> >>> from
>>>>>>>>> >>> >> the configure.log)?
>>>>>>>>> >>> >>
>>>>>>>>> >>> >
>>>>>>>>> >>> > I was compiling codes on login nodes, and running codes on
>>>>>>>>> compute
>>>>>>>>> >>> nodes.
>>>>>>>>> >>> > Login nodes do not have GPUs, but compute nodes do have GPUs.
>>>>>>>>> >>> >
>>>>>>>>> >>> > Just to be clear, the same thing (code, machine) with
>>>>>>>>> PETSc-3.16.1
>>>>>>>>> >>> worked
>>>>>>>>> >>> > perfectly. I have this trouble with PETSc-main.
>>>>>>>>> >>>
>>>>>>>>> >>> I assume you can
>>>>>>>>> >>>
>>>>>>>>> >>> export PETSC_OPTIONS='-device_enable lazy'
>>>>>>>>> >>>
>>>>>>>>> >>> and it'll work.
>>>>>>>>> >>>
>>>>>>>>> >>> I think this should be the default. The main complaint is that
>>>>>>>>> timing the
>>>>>>>>> >>> first GPU-using event isn't accurate if it includes
>>>>>>>>> initialization, but I
>>>>>>>>> >>> think this is mostly hypothetical because you can't trust any
>>>>>>>>> timing that
>>>>>>>>> >>> doesn't preload in some form and the first GPU-using event
>>>>>>>>> will almost
>>>>>>>>> >>> always be something uninteresting so I think it will rarely
>>>>>>>>> lead to
>>>>>>>>> >>> confusion. Meanwhile, eager initialization is viscerally
>>>>>>>>> disruptive for
>>>>>>>>> >>> lots of people.
>>>>>>>>> >>>
>>>>>>>>> >>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> What most experimenters take for granted before they begin their
>>>>>>> experiments is infinitely more interesting than any results to which
>>>>>>> their
>>>>>>> experiments lead.
>>>>>>> -- Norbert Wiener
>>>>>>>
>>>>>>> https://www.cse.buffalo.edu/~knepley/
>>>>>>> <http://www.cse.buffalo.edu/~knepley/>
>>>>>>>
>>>>>>>
>>>>>> <configure_bad.log><configure_good.log>
>>
>>
>>
