unrecognised option at configure

Question

unrecognised option at configure

DGCaprace opened this issue 5 years ago · comments

Denis-Gabriel Caprace commented 5 years ago

Hi,

I would like to use the recommended --enable-oned --enable-stride1 options, as advertised in the documentation, when configuring the install of the library. However, I get the following warning when executing the configure script:

configure: WARNING: unrecognized options: --enable-oned, --enable-stride1

Are these options ignored? deprecated? or have they been made default and one doesn't need to specify them?

Dmitry Pekurovsky · Answer 1 · Fri Nov 08 2019 02:53:18 GMT+0800 (China Standard Time)

Hi, these options are from P3DFFT (version 2.x) and are deprecated in P3DFFT++ (version 3.x). Which documentation were you looking at?

…

On Thu, Nov 7, 2019 at 6:57 AM DeeGee94 ***@***.***> wrote: Hi, I would like to use the recommended --enable-oned --enable-stride1 options, as advertised in the documentation, when configuring the install of the library. However, I get the following warning when executing the configure script: configure: WARNING: unrecognized options: --enable-oned, --enable-stride1 Are these options ignored? deprecated? or have they been made default and one doesn't need to specify them? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=ACXPORU54V3BRIJZBJ6KGV3QSQUF7A5CNFSM4JKH5Q3KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HXUFMLQ>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXPORQMO5Z3PFJRE3NQXT3QSQUF7ANCNFSM4JKH5Q3A> .

Denis-Gabriel Caprace · Answer 2 · Fri Nov 08 2019 04:10:12 GMT+0800 (China Standard Time)

Hi Dr. Pekurovski,

Thanks for the quick reply: I was referring to the table of arguments on the install page of the online readthedoc of P3DFFT++ which still has these options, to date. Then, could you please confirm that P3DFFT v3 always use stride-1 accesses in the calls to FFTW? Same question for --enable-useeven: are you always padding in order to call MPI_Alltoall instead of MPI_Alltoallv?

Dmitry Pekurovsky · Answer 3 · Fri Nov 08 2019 04:27:25 GMT+0800 (China Standard Time)

Thanks for your comments. We will adjust the documentation. To answer your question: P3DFFT++ (v. 3) uses a more sophisticated interface that gives the user power to define the data layout, such as using stride-1 or not. Within the library all calls to FFTW are done with stride-1. Currently we don't use padding for mpi_alltoallv, as this was relevant only for Cray XT systems, as far as I know; but I'm open for more information about this.

…

On Thu, Nov 7, 2019 at 12:10 PM DeeGee94 ***@***.***> wrote: Hi Dr. Pekurovski, Thanks for the quick reply: I was referring to the table of arguments on the install page of the online readthedoc of P3DFFT++ <https://p3dfft.readthedocs.io/en/latest/install_p3dfft++.html#how-to-compile-p3dfft> which still has these options, to date. Then, could you please confirm that P3DFFT v3 always use stride-1 accesses in the calls to FFTW? Same question for --enable-useeven: are you always padding in order to call MPI_Alltoall instead of MPI_Alltoallv? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=ACXPORS4FI7LMQYJE2GZDZ3QSRY2LA5CNFSM4JKH5Q3KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDNU43I#issuecomment-551243373>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXPORWSC4FYWW3BQSOFAFTQSRY2LANCNFSM4JKH5Q3A> .