[petsc-dev] navigating gitlab

[Matthew Knepley]

On Mon, Jan 18, 2021 at 9:47 AM Mark Adams <mfadams at lbl.gov> wrote:

> On Mon, Jan 18, 2021 at 8:59 AM Matthew Knepley <knepley at gmail.com> wrote:
>
>> On Mon, Jan 18, 2021 at 8:23 AM Mark Adams <mfadams at lbl.gov> wrote:
>>
>>> BTW, how do you get that URL? I searched commits for it and did not find
>>> it.
>>>
>>> Anyway, that bisect failed because I had an error in my verification
>>> test (that has since been fixed). I now get this.
>>>
>>> Barry: This commit seems to break my cuda Landau code. I'm guessing that
>>> I need to change my configure script attached. Any ideas?
>>>
>>
>> Send the configure.log. Barry put output there that will let us see what
>> it is doing.
>>
>
Okay, this is in the report at the end:

cuda:

  Version:  10.1

  Includes: -I/sw/summit/cuda/10.1.243/include

  Library:  -Wl,-rpath,/sw/summit/cuda/10.1.243/lib64
-L/sw/summit/cuda/10.1.243/lib64 -lcufft -lcublas -lcudart -lcusparse
-lcusolver

  CUDA underlying compiler: CUDA_CXX gcc

  CUDA underlying compiler flags: CUDA_CXXFLAGS

It looks like you might need to add

--CUDA_CXXFLAGS="-DLANDAU_MAX_Q=4 -DLANDAU_MAX_SPECIES=10
-DPETSC_HAVE_CUDA_ATOMIC"

  Thanks,

      Matt

  Thanks,
>>
>>      Matt
>>
>>
>>> Thanks,
>>> Mark
>>>
>>> 08:06 (7ba7a81...)|BISECTING /gpfs/alpine/csc314/scratch/adams/petsc$
>>> git bisect bad
>>> 7ba7a8179202d83658a309df68e2ebc2037284d4 is the first bad commit
>>> commit 7ba7a8179202d83658a309df68e2ebc2037284d4
>>> Author: Barry Smith <bsmith at mcs.anl.gov>
>>> Date:   Thu Dec 31 00:28:23 2020 -0500
>>>
>>>     The previous code passed options for the PETSc C, C preprocessor and
>>> CXX compiler to the
>>>     ccbin compiler used by nvcc even if it was a completely different
>>> compiler and the arguments
>>>     caused the compiler to error with unknown arguments.
>>>
>>>     This has been an outstanding problem for years, this is the first
>>> crude attempt to fix it by
>>>     not passing the arguments unless the compiler is identical to the
>>> PETSc compiler
>>>
>>>     TODO: have a better check that the compilers are the same
>>>     TODO: provide a systematic way for the user to provide the compiler
>>> used by nvcc, perhaps default to the compiler used by PETSc?
>>>
>>>     /spend 3.5h
>>>     Thanks-to: Jose E. Roman <jroman at dsic.upv.es>
>>>
>>>     Fix typo in list of c99 flags
>>>
>>>     Add error message regarding Intel compilers breaking Gnu compilers
>>> when c99 test fails with gcc.
>>>
>>>     Eventually we may want more systematic checks of Intel compilers
>>> breaking Gnu compilers
>>>
>>>     /spend 15m
>>>
>>> :040000 040000 b4d44a220b09ceebb8020979eebf1b98da41e3ab
>>> f24913d3614f54de6dfa49a789325f2c71c9983f M config
>>> :100644 100644 9e53e76f9a4960cd943304f56bd1a8a10a3ae516
>>> 5309077a5a107b3f553b7875ad2a19c17741655b M gmakefile.test
>>> :040000 040000 717c9a395a53442fe6917c0aa033b61d3427fbb2
>>> aa365576e1efa7f29d8817c69d03e1983612caa1 M lib(base) 08:16  ~$
>>>
>>>
>>
>> --
>> 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/>
>>
>

-- 
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/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20210118/b28b7685/attachment.html>