<div dir="ltr"><div dir="ltr">On Thu, Oct 19, 2023 at 1:46 PM Enrico <<a href="mailto:degregori@dkrz.de">degregori@dkrz.de</a>> wrote:<br></div><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Sorry but I don't want another partition, Petsc internally is changing <br>
the partition. I would like to have the same partition that I have in <br>
the application. Is the example not clear?<br></blockquote><div><br></div><div>"Petsc internally is changing the partition" This is not correct. PETSc does</div><div>not prescribe partitions. I refer to the documentation for the creation of Vec:</div><div><br></div><div>  <a href="https://petsc.org/main/manualpages/Vec/VecCreateMPI/">https://petsc.org/main/manualpages/Vec/VecCreateMPI/</a></div><div><br></div><div>Here the user sets the local and global sizes. Since data is contiguous, these</div><div>completely define the vector, and are under user control.</div><div><br></div><div>  Thanks,</div><div><br></div><div>     Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
On 19/10/2023 19:43, Matthew Knepley wrote:<br>
> On Thu, Oct 19, 2023 at 1:00 PM Enrico <<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <br>
> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>> wrote:<br>
> <br>
>     Here is a very very simple reproducer of my problem. It is a fortran<br>
>     program and it has to run with 2 processes.<br>
> <br>
> <br>
> You seem to be saying that you start with one partition of your data, <br>
> but you would like<br>
> another partition. For this, you have to initially communicate. For this <br>
> I would use VecScatter.<br>
> However, since most data is generated, I would consider not generating <br>
> my data in that initial<br>
> distribution.<br>
> <br>
> There are many examples in the repository. In the discretization of a <br>
> PDE, we first divide the domain,<br>
> then number up each piece, then assemble the linear algebra objects.<br>
> <br>
>    Thanks,<br>
> <br>
>        Matt<br>
> <br>
>     The output is:<br>
> <br>
>        process             0 : xx_v(            1 ) =     0.000000000000000<br>
>        process             0 : xx_v(            2 ) =     1.000000000000000<br>
>        process             0 : xx_v(            3 ) =     2.000000000000000<br>
>        process             1 : xx_v(            1 ) =     3.000000000000000<br>
>        process             1 : xx_v(            2 ) =     4.000000000000000<br>
>        process             1 : xx_v(            3 ) =     5.000000000000000<br>
> <br>
>     and I would like to have:<br>
> <br>
>        process             0 : xx_v(            1 ) =     2.000000000000000<br>
>        process             0 : xx_v(            2 ) =     3.000000000000000<br>
>        process             0 : xx_v(            3 ) =     4.000000000000000<br>
>        process             1 : xx_v(            1 ) =     0.000000000000000<br>
>        process             1 : xx_v(            2 ) =     1.000000000000000<br>
>        process             1 : xx_v(            3 ) =     5.000000000000000<br>
> <br>
>     How can I do that?<br>
> <br>
>     program main<br>
>     #include <petsc/finclude/petscksp.h><br>
>           use petscksp<br>
>           implicit none<br>
> <br>
>           PetscErrorCode ierr<br>
>           PetscInt  :: Psize = 6<br>
>           integer  :: Lsize<br>
>           PetscInt  :: work_size<br>
>           PetscInt  :: work_rank<br>
>           Vec :: b<br>
>           integer, allocatable, dimension(:) :: glb_index<br>
>           double precision, allocatable, dimension(:) :: array<br>
>           PetscScalar, pointer :: xx_v(:)<br>
>           integer :: i<br>
>           PetscCount :: csize<br>
> <br>
>           CALL PetscInitialize(ierr)<br>
> <br>
>           Lsize = 3<br>
>           csize = Lsize<br>
> <br>
>           allocate(glb_index(0:Lsize-1), array(0:Lsize-1))<br>
> <br>
>           CALL MPI_Comm_size(PETSC_COMM_WORLD, work_size, ierr);<br>
>           CALL MPI_Comm_rank(PETSC_COMM_WORLD, work_rank, ierr);<br>
>           if (work_rank == 0) then<br>
>             glb_index(0) = 2<br>
>             glb_index(1) = 3<br>
>             glb_index(2) = 4<br>
>             array(0) = 2<br>
>             array(1) = 3<br>
>             array(2) = 4<br>
>           else if (work_rank == 1) then<br>
>             glb_index(0) = 0<br>
>             glb_index(1) = 1<br>
>             glb_index(2) = 5<br>
>             array(0) = 0<br>
>             array(1) = 1<br>
>             array(2) = 5<br>
>           end if<br>
> <br>
>           ! Create and fill rhs vector<br>
>           CALL VecCreate(PETSC_COMM_WORLD, b, ierr);<br>
>           CALL VecSetSizes(b, Lsize, Psize, ierr);<br>
>           CALL VecSetType(b, VECMPI, ierr);<br>
> <br>
>           CALL VecSetPreallocationCOO(b, csize, glb_index, ierr)<br>
>           CALL VecSetValuesCOO(b, array, INSERT_VALUES, ierr)<br>
> <br>
>           CALL VecGetArrayReadF90(b, xx_v, ierr)<br>
> <br>
>           do i=1,Lsize<br>
>             write(*,*) 'process ', work_rank, ': xx_v(',i,') = ', xx_v(i)<br>
>           end do<br>
> <br>
>           CALL VecRestoreArrayReadF90(b, xx_v, ierr)<br>
> <br>
>           deallocate(glb_index, array)<br>
>           CALL VecDestroy(b,ierr)<br>
> <br>
>           CALL PetscFinalize(ierr)<br>
> <br>
>     end program main<br>
> <br>
> <br>
>     On 19/10/2023 17:36, Matthew Knepley wrote:<br>
>      > On Thu, Oct 19, 2023 at 11:33 AM Enrico <<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      > <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>> wrote:<br>
>      ><br>
>      >     The layout is not poor, just the global indices are not<br>
>     contiguous,this<br>
>      >     has nothing to do with the local memory layout which is extremely<br>
>      >     optimized for different architectures. I can not change the<br>
>     layout<br>
>      >     anyway because it's a climate model with a million lines of code.<br>
>      ><br>
>      >     I don't understand why Petsc is doing all this MPI<br>
>     communication under<br>
>      >     the hood.<br>
>      ><br>
>      ><br>
>      > I don't think we are communicating under the hood.<br>
>      ><br>
>      >     I mean, it is changing the layout of the application and doing<br>
>      >     a lot of communication.<br>
>      ><br>
>      ><br>
>      > We do not create the layout. The user creates the data layout<br>
>     when they<br>
>      > create a vector or matrix.<br>
>      ><br>
>      >     Is there no way to force the same layout and<br>
>      >     provide info about how to do the halo exchange? In this way I<br>
>     can have<br>
>      >     the same memory layout and there is no communication when I<br>
>     fill or<br>
>      >     fetch the vectors and the matrix.<br>
>      ><br>
>      ><br>
>      > Yes, you tell the vector/matrix your data layout when you create it.<br>
>      ><br>
>      >    Thanks,<br>
>      ><br>
>      >        Matt<br>
>      ><br>
>      >     Cheers,<br>
>      >     Enrico<br>
>      ><br>
>      >     On 19/10/2023 17:21, Matthew Knepley wrote:<br>
>      >      > On Thu, Oct 19, 2023 at 10:51 AM Enrico <<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >      > <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>> wrote:<br>
>      >      ><br>
>      >      >     In the application the storage is contiguous but the<br>
>     global<br>
>      >     indexing is<br>
>      >      >     not. I would like to use AO as a translation layer but<br>
>     I don't<br>
>      >      >     understand it.<br>
>      >      ><br>
>      >      ><br>
>      >      > Why would you choose to index differently from your storage?<br>
>      >      ><br>
>      >      >     My case is actually simple even if it is in a large<br>
>      >     application, I have<br>
>      >      ><br>
>      >      >     Mat A, Vec b and Vec x<br>
>      >      ><br>
>      >      >     After calling KSPSolve, I use VecGetArrayReadF90 to get a<br>
>      >     pointer to<br>
>      >      >     the<br>
>      >      >     data and they are in the wrong ordering, so for<br>
>     example the first<br>
>      >      >     element of the solution array on process 0 belongs to<br>
>     process<br>
>      >     1 in the<br>
>      >      >     application.<br>
>      >      ><br>
>      >      ><br>
>      >      > Again, this seems to be a poor choice of layout. What we<br>
>      >     typically do is<br>
>      >      > to partition<br>
>      >      > the data into chunks owned by each process first.<br>
>      >      ><br>
>      >      >     Is it at this point that I should use the AO translation<br>
>      >     layer? This<br>
>      >      >     would be quite bad, it means to build Mat A and Vec b<br>
>     there<br>
>      >     is MPI<br>
>      >      >     communication and also to get the data of Vec x back<br>
>     in the<br>
>      >     application.<br>
>      >      ><br>
>      >      ><br>
>      >      > If you want to store data that process i updates on process j,<br>
>      >     this will<br>
>      >      > need communication.<br>
>      >      ><br>
>      >      >     Anyway, I've tried to use<br>
>     AOPetscToApplicationPermuteReal on the<br>
>      >      >     solution array but it doesn't work as I would like. Is<br>
>     this<br>
>      >     function<br>
>      >      >     suppose to do MPI communication between processes and<br>
>     fetch<br>
>      >     the values<br>
>      >      >     of the application ordering?<br>
>      >      ><br>
>      >      ><br>
>      >      > There is no communication here. That function call just<br>
>     changes one<br>
>      >      > integer into another.<br>
>      >      > If you want to update values on another process, we<br>
>     recommend using<br>
>      >      > VecScatter() or<br>
>      >      > MatSetValues(), both of which take global indices and do<br>
>      >     communication<br>
>      >      > if necessary.<br>
>      >      ><br>
>      >      >    Thanks,<br>
>      >      ><br>
>      >      >      Matt<br>
>      >      ><br>
>      >      >     Cheers,<br>
>      >      >     Enrico<br>
>      >      ><br>
>      >      >     On 19/10/2023 15:25, Matthew Knepley wrote:<br>
>      >      >      > On Thu, Oct 19, 2023 at 8:57 AM Enrico<br>
>     <<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >      >      > <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>>> wrote:<br>
>      >      >      ><br>
>      >      >      >     Maybe I wasn't clear enough. I would like to<br>
>      >     completely get<br>
>      >      >     rid of<br>
>      >      >      >     Petsc<br>
>      >      >      >     ordering because I don't want extra<br>
>     communication between<br>
>      >      >     processes to<br>
>      >      >      >     construct the vector and the matrix (since I<br>
>     have to fill<br>
>      >      >     them every<br>
>      >      >      >     time step because I'm just using the linear solver<br>
>      >     with a Mat<br>
>      >      >     and a Vec<br>
>      >      >      >     data structure). I don't understand how I can<br>
>     do that.<br>
>      >      >      ><br>
>      >      >      ><br>
>      >      >      > Any program you write to do linear algebra will have<br>
>      >     contiguous<br>
>      >      >     storage<br>
>      >      >      > because it<br>
>      >      >      > is so much faster. Contiguous indexing makes sense for<br>
>      >     contiguous<br>
>      >      >      > storage. If you<br>
>      >      >      > want to use non-contiguous indexing for contiguous<br>
>      >     storage, you<br>
>      >      >     would<br>
>      >      >      > need some<br>
>      >      >      > translation layer. The AO is such a translation,<br>
>     but you<br>
>      >     could do<br>
>      >      >     this<br>
>      >      >      > any way you want.<br>
>      >      >      ><br>
>      >      >      >    Thanks,<br>
>      >      >      ><br>
>      >      >      >       Matt<br>
>      >      >      ><br>
>      >      >      >     My initial idea was to create another global index<br>
>      >     ordering<br>
>      >      >     within my<br>
>      >      >      >     application to use only for the Petsc interface<br>
>     but then I<br>
>      >      >     think that<br>
>      >      >      >     the ghost cells are wrong.<br>
>      >      >      ><br>
>      >      >      >     On 19/10/2023 14:50, Matthew Knepley wrote:<br>
>      >      >      >      > On Thu, Oct 19, 2023 at 6:51 AM Enrico<br>
>      >     <<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>><br>
>      >      >      >      > <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>>>> wrote:<br>
>      >      >      >      ><br>
>      >      >      >      >     Hello,<br>
>      >      >      >      ><br>
>      >      >      >      >     if I create an application ordering using<br>
>      >      >     AOCreateBasic, should I<br>
>      >      >      >      >     provide the same array for const PetscInt<br>
>      >     myapp[] and<br>
>      >      >     const<br>
>      >      >      >     PetscInt<br>
>      >      >      >      >     mypetsc[] in order to get the same<br>
>     ordering of the<br>
>      >      >     application<br>
>      >      >      >      >     within PETSC?<br>
>      >      >      >      ><br>
>      >      >      >      ><br>
>      >      >      >      > Are you asking if the identity permutation<br>
>     can be<br>
>      >     constructed<br>
>      >      >      >     using the<br>
>      >      >      >      > same array twice? Yes.<br>
>      >      >      >      ><br>
>      >      >      >      >     And once I define the ordering so that<br>
>     the local<br>
>      >      >     vector and<br>
>      >      >      >     matrix are<br>
>      >      >      >      >     defined in PETSC as in my application,<br>
>     how can<br>
>      >     I use it to<br>
>      >      >      >     create the<br>
>      >      >      >      >     actual vector and matrix?<br>
>      >      >      >      ><br>
>      >      >      >      ><br>
>      >      >      >      > The vectors and matrices do not change. The<br>
>     AO is a<br>
>      >      >     permutation.<br>
>      >      >      >     You can<br>
>      >      >      >      > use it to permute<br>
>      >      >      >      > a vector into another order, or to convert<br>
>     on index to<br>
>      >      >     another.<br>
>      >      >      >      ><br>
>      >      >      >      >    Thanks,<br>
>      >      >      >      ><br>
>      >      >      >      >        Matt<br>
>      >      >      >      ><br>
>      >      >      >      >     Thanks in advance for the help.<br>
>      >      >      >      ><br>
>      >      >      >      >     Cheers,<br>
>      >      >      >      >     Enrico<br>
>      >      >      >      ><br>
>      >      >      >      >     On 18/10/2023 13:39, Matthew Knepley wrote:<br>
>      >      >      >      >      > On Wed, Oct 18, 2023 at 5:55 AM Enrico<br>
>      >      >     <<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>><br>
>      >      >      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>>><br>
>      >      >      >      >      > <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>><br>
>      >      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>><br>
>      >      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>><br>
>      >     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>><br>
>     <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a> <mailto:<a href="mailto:degregori@dkrz.de" target="_blank">degregori@dkrz.de</a>>>>>>>> wrote:<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     Hello,<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     I'm trying to use Petsc to solve<br>
>     a linear<br>
>      >      >     system in an<br>
>      >      >      >      >     application. I'm<br>
>      >      >      >      >      >     using the coordinate format to<br>
>     define the<br>
>      >      >     matrix and the<br>
>      >      >      >      >     vector (it<br>
>      >      >      >      >      >     should work better on GPU but at<br>
>     the moment<br>
>      >      >     every test<br>
>      >      >      >     is on<br>
>      >      >      >      >     CPU).<br>
>      >      >      >      >      >     After<br>
>      >      >      >      >      >     the call to VecSetValuesCOO, I've<br>
>      >     noticed that the<br>
>      >      >      >     vector is<br>
>      >      >      >      >     storing<br>
>      >      >      >      >      >     the<br>
>      >      >      >      >      >     data in a different way from my<br>
>      >     application. For<br>
>      >      >      >     example with two<br>
>      >      >      >      >      >     processes in the application<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     process 0 owns cells 2, 3, 4<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     process 1 owns cells 0, 1, 5<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     But in the vector data structure<br>
>     of Petsc<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     process 0 owns cells 0, 1, 2<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     process 1 owns cells 3, 4, 5<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     This is in principle not a big issue,<br>
>      >     but after<br>
>      >      >      >     solving the<br>
>      >      >      >      >     linear<br>
>      >      >      >      >      >     system I get the solution vector<br>
>     x and I<br>
>      >     want<br>
>      >      >     to get the<br>
>      >      >      >      >     values in the<br>
>      >      >      >      >      >     correct processes. Is there a way<br>
>     to get<br>
>      >     vector<br>
>      >      >     values<br>
>      >      >      >     from other<br>
>      >      >      >      >      >     processes or to get a mapping so<br>
>     that I<br>
>      >     can do<br>
>      >      >     it myself?<br>
>      >      >      >      >      ><br>
>      >      >      >      >      ><br>
>      >      >      >      >      > By definition, PETSc vectors and<br>
>     matrices own<br>
>      >      >     contiguous row<br>
>      >      >      >      >     blocks. If<br>
>      >      >      >      >      > you want to have another,<br>
>      >      >      >      >      > global ordering, we support that with<br>
>      >      >      >      >      ><br>
>     <a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>><br>
>      >      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>>><br>
>      >      >      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>><br>
>      >      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>>>><br>
>      >      >      >      >      ><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>><br>
>      >      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>>><br>
>      >      >      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>><br>
>      >      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>><br>
>      >      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>><br>
>      >     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a><br>
>     <<a href="https://petsc.org/main/manualpages/AO/" rel="noreferrer" target="_blank">https://petsc.org/main/manualpages/AO/</a>>>>>>><br>
>      >      >      >      >      ><br>
>      >      >      >      >      >    Thanks,<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >       Matt<br>
>      >      >      >      >      ><br>
>      >      >      >      >      >     Cheers,<br>
>      >      >      >      >      >     Enrico Degregori<br>
>      >      >      >      >      ><br>
>      >      >      >      >      ><br>
>      >      >      >      >      ><br>
>      >      >      >      >      > --<br>
>      >      >      >      >      > What most experimenters take for<br>
>     granted before<br>
>      >      >     they begin<br>
>      >      >      >     their<br>
>      >      >      >      >      > experiments is infinitely more<br>
>     interesting<br>
>      >     than any<br>
>      >      >      >     results to which<br>
>      >      >      >      >      > their experiments lead.<br>
>      >      >      >      >      > -- Norbert Wiener<br>
>      >      >      >      >      ><br>
>      >      >      >      >      > <a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>>><br>
>      >      >      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>>>><br>
>      >      >      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>>><br>
>      >      >      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>>>>><br>
>      >      >      >      ><br>
>      >      >      >      ><br>
>      >      >      >      ><br>
>      >      >      >      > --<br>
>      >      >      >      > What most experimenters take for granted before<br>
>      >     they begin<br>
>      >      >     their<br>
>      >      >      >      > experiments is infinitely more interesting<br>
>     than any<br>
>      >      >     results to which<br>
>      >      >      >      > their experiments lead.<br>
>      >      >      >      > -- Norbert Wiener<br>
>      >      >      >      ><br>
>      >      >      >      > <a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>>><br>
>      >      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>>>><br>
>      >      >      ><br>
>      >      >      ><br>
>      >      >      ><br>
>      >      >      > --<br>
>      >      >      > What most experimenters take for granted before<br>
>     they begin<br>
>      >     their<br>
>      >      >      > experiments is infinitely more interesting than any<br>
>      >     results to which<br>
>      >      >      > their experiments lead.<br>
>      >      >      > -- Norbert Wiener<br>
>      >      >      ><br>
>      >      >      > <a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>>><br>
>      >      ><br>
>      >      ><br>
>      >      ><br>
>      >      > --<br>
>      >      > What most experimenters take for granted before they begin<br>
>     their<br>
>      >      > experiments is infinitely more interesting than any<br>
>     results to which<br>
>      >      > their experiments lead.<br>
>      >      > -- Norbert Wiener<br>
>      >      ><br>
>      >      > <a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
>      >     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>>><br>
>      ><br>
>      ><br>
>      ><br>
>      > --<br>
>      > What most experimenters take for granted before they begin their<br>
>      > experiments is infinitely more interesting than any results to which<br>
>      > their experiments lead.<br>
>      > -- Norbert Wiener<br>
>      ><br>
>      > <a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a>><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a><br>
>     <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>>><br>
> <br>
> <br>
> <br>
> -- <br>
> What most experimenters take for granted before they begin their <br>
> experiments is infinitely more interesting than any results to which <br>
> their experiments lead.<br>
> -- Norbert Wiener<br>
> <br>
> <a href="https://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">https://www.cse.buffalo.edu/~knepley/</a> <<a href="http://www.cse.buffalo.edu/~knepley/" rel="noreferrer" target="_blank">http://www.cse.buffalo.edu/~knepley/</a>><br>
</blockquote></div><br clear="all"><div><br></div><span class="gmail_signature_prefix">-- </span><br><div dir="ltr" class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div>What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>-- Norbert Wiener</div><div><br></div><div><a href="http://www.cse.buffalo.edu/~knepley/" target="_blank">https://www.cse.buffalo.edu/~knepley/</a><br></div></div></div></div></div></div></div></div>