January 2021 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] navigating gitlab
by Mark Adams 18 Jan '21

18 Jan '21

> > > --CUDA_CXXFLAGS="-DLANDAU_MAX_Q=4 -DLANDAU_MAX_SPECIES=10 > -DPETSC_HAVE_CUDA_ATOMIC" > Thanks, this got me on the right track. I added these MAX defines to -CUDAFLAGS .

1 0

Re: [petsc-dev] navigating gitlab
by Satish Balay 18 Jan '21

18 Jan '21

On Mon, 18 Jan 2021, Mark Adams wrote: > BTW, how do you get that URL? I searched commits for it and did not find it. Alternative: gitk 7ba7a8179202d83658a309df68e2ebc2037284d4 git show 7ba7a8179202d83658a309df68e2ebc2037284d4 Satish

1 0

Re: [petsc-dev] navigating gitlab
by Matthew Knepley 18 Jan '21

18 Jan '21

On Mon, Jan 18, 2021 at 9:47 AM Mark Adams <mfadams(a)lbl.gov> wrote: > On Mon, Jan 18, 2021 at 8:59 AM Matthew Knepley <knepley(a)gmail.com> wrote: > >> On Mon, Jan 18, 2021 at 8:23 AM Mark Adams <mfadams(a)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(a)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(a)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/>

1 0

Re: [petsc-dev] navigating gitlab
by Matthew Knepley 18 Jan '21

18 Jan '21

On Mon, Jan 18, 2021 at 10:04 AM Jed Brown <jed(a)jedbrown.org> wrote: > Mark Adams <mfadams(a)lbl.gov> writes: > > > BTW, how do you get that URL? I searched commits for it and did not find > it. > > I just type the URL or have Emacs construct it. > Gitlab has had a request to make it work in the Search box for months. I +1'd it. I am not sure where they put their time. Matt -- 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/>

1 0

Re: [petsc-dev] navigating gitlab
by Jed Brown 18 Jan '21

18 Jan '21

Mark Adams <mfadams(a)lbl.gov> writes: > BTW, how do you get that URL? I searched commits for it and did not find it. I just type the URL or have Emacs construct it.

1 0

Re: [petsc-dev] navigating gitlab
by Mark Adams 18 Jan '21

18 Jan '21

On Mon, Jan 18, 2021 at 8:59 AM Matthew Knepley <knepley(a)gmail.com> wrote: > On Mon, Jan 18, 2021 at 8:23 AM Mark Adams <mfadams(a)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. > > 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(a)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(a)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/> >

1 0

Re: [petsc-dev] navigating gitlab
by Matthew Knepley 18 Jan '21

18 Jan '21

On Mon, Jan 18, 2021 at 8:23 AM Mark Adams <mfadams(a)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. 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(a)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(a)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/>

1 0

Re: [petsc-dev] navigating gitlab
by Mark Adams 18 Jan '21

18 Jan '21

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? 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(a)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(a)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 ~$

1 0

Re: [petsc-dev] navigating gitlab
by Mark Adams 18 Jan '21

18 Jan '21

Thanks, I realized this AM that I was getting MRs and commits confused. On Sun, Jan 17, 2021 at 10:53 PM Jed Brown <jed(a)jedbrown.org> wrote: > Mark Adams <mfadams(a)lbl.gov> writes: > > > I have a bug: > > > > Bisecting: 0 revisions left to test after this (roughly 0 steps) > > [c8a6034effdc7533a74cd8b06a631e4a6bd6fb69] Adams/plex noprealloc fix > > > > I am not finding this on gitlab, to see exactly what it did. I looked > under > > "merged" MRs and could not get search to find it and could not find it by > > inspection. > > You can go straight to the commit. Does this help? > > > https://gitlab.com/petsc/petsc/-/commit/c8a6034effdc7533a74cd8b06a631e4a6bd… > > > Could someone that can navigate this interface better than I tell me how > to > > do this. > > > > Thanks, > > Mark >

1 0

Re: [petsc-dev] navigating gitlab
by Jed Brown 17 Jan '21

17 Jan '21

Mark Adams <mfadams(a)lbl.gov> writes: > I have a bug: > > Bisecting: 0 revisions left to test after this (roughly 0 steps) > [c8a6034effdc7533a74cd8b06a631e4a6bd6fb69] Adams/plex noprealloc fix > > I am not finding this on gitlab, to see exactly what it did. I looked under > "merged" MRs and could not get search to find it and could not find it by > inspection. You can go straight to the commit. Does this help? https://gitlab.com/petsc/petsc/-/commit/c8a6034effdc7533a74cd8b06a631e4a6bd… > Could someone that can navigate this interface better than I tell me how to > do this. > > Thanks, > Mark

1 0