[petsc-users] petsc-users Digest, Vol 200, Issue 3

[kiran hegde]

Dear Sanjay,

Both the construction works, no error reported. I need to test the code
further to see if it really passes the address. There are a few more errors
reported afterwards. Eg. preconditioner (PC) is now included in
petsc/finclude/petscpc.h whereas in older versions it is available from
petsc/finclude/petscpc.h.  I will try to fix issues and update to you later
:

call VecCreateSeqWithArray(PETSC_COMM_WORLD, nvar, nvar*g%nvl, &
                                [s%res(1,1)] , p%v_res, ierr); CHKERRQ(ierr)
   OR
call VecCreateSeqWithArray(PETSC_COMM_WORLD, nvar, nvar*g%nvl, &
                                (/ s%res(1,1) /), p%v_res, ierr);
CHKERRQ(ierr)


On Mon, Aug 11, 2025 at 9:51 PM Barry Smith <bsmith at petsc.dev> wrote:

>
>   I do not know. It depends on how the Fortran compiler decides to
> interprete the [] and pass the address down. In the end the address of the
> actual array in p has to sent down since it will be used later so if the
> compiler makes a new array with [] or does something fancy it may or may
> not work.
>
> On Aug 11, 2025, at 12:08 PM, Sanjay Govindjee <s_g at berkeley.edu> wrote:
>
>
> Would a construct like
> call VecCreateSeqWithArray(PETSC_COMM_WORLD, nvar, nvar*g%nvl, &
>                                 [s%res(1,1)] , p%v_res, ierr);
> CHKERRQ(ierr)
> or
> call VecCreateSeqWithArray(PETSC_COMM_WORLD, nvar, nvar*g%nvl, &
>                                 (/ s%res(1,1) /), p%v_res, ierr);
> CHKERRQ(ierr)
> be sufficient in this case?
>
> -sanjay
>
> On Mon, Aug 11, 2025 at 7:54 AM Barry Smith <bsmith at petsc.dev> wrote:
>
>>
>>   PETSc now provides Fortran interface definitions in the Fortran modules
>> for most PETSc functions. This means the argument types and array
>> dimensions of the function must match the expected type and dimension.
>>
>>    In this case, the argument s%res(1,1) is likely a scalar while
>> VecCreateSeqWithArray is expected to be a one-dimensional array.  You can
>> change your code to declare res(1,1) to be a one-dimensional array, or
>> possibly use reshape to have the compiler treat it as a one-dimensional
>> array when passing it in as an argument.
>>
>>
>>
>> On Aug 10, 2025, at 4:12 AM, kiran hegde <kiranhegde75 at gmail.com> wrote:
>>
>> Hi,
>> The error highlighted is gone, the path to the include file was wrong.
>> The code was developed during  petsc-3.9 and it looks like i have to fix
>> many of the petsc call functions.
>> There are big  changes in  post petsc-3.9.
>>
>> Currently, the error shifted to another location in the code:
>> for the below line i am getting error:
>> call VecCreateSeqWithArray(PETSC_COMM_WORLD, nvar, nvar*g%nvl, &
>>                                  s%res(1,1), p%v_res, ierr); CHKERRQ(ierr)
>>
>> Error: There is no specific subroutine for the generic
>> 'veccreateseqwitharray' at (1)
>>
>> Regards,
>> Kiran
>>
>>
>>
>>
>> On Fri, Aug 8, 2025 at 6:39 AM Barry Smith <bsmith at petsc.dev> wrote:
>>
>>>
>>>  Please send (cut and paste) all the output of
>>>
>>>  make ex5f  (should succeed)
>>>
>>> and
>>>
>>> make "your code"
>>>
>>> we need to see the exact calls for both codes and the exact error
>>> message.
>>>
>>>
>>>
>>>
>>> On Aug 5, 2025, at 12:44 AM, kiran hegde <kiranhegde75 at gmail.com> wrote:
>>>
>>> Hi,
>>>
>>> Sorry, the tutorial case run succeeded. No errors reported. When i used
>>> the makefile file flags in my code, my code is still reported error.
>>>
>>> Regards,
>>>
>>> Kiran Hegde
>>> Mobile:+91 9972600235
>>>
>>> On Tue, Aug 5, 2025, 6:11 AM Barry Smith <bsmith at petsc.dev> wrote:
>>>
>>>>
>>>>    "I have tried the options ex5f, but getting the same error."
>>>>
>>>>    So you did
>>>>
>>>>     cd src/snes/tutorials/
>>>>     make ex5f
>>>>
>>>>     and got some errors? Please cut and paste all the output from make
>>>> ex5f and sent it here.
>>>>
>>>>
>>>>
>>>> On Aug 4, 2025, at 11:15 AM, kiran hegde <kiranhegde75 at gmail.com>
>>>> wrote:
>>>>
>>>> Hi Satish,
>>>>
>>>> I have tried the options ex5f, but getting the same error.
>>>> The code was running without any issues long back and I was using
>>>> petsc ver 3.9 using gcc 12.X.X.
>>>> My makefile flags are :
>>>>
>>>> FFLAGS = -march=native -O3 -pipe -ffree-form -cpp
>>>> -ffree-line-length-none \
>>>>                         -Wall -Wextra -Wtabs -Wintrinsics-std
>>>> -Wsurprising -Waliasing \
>>>>                         -Wno-unused-parameter
>>>> -Wno-unused-dummy-argument \
>>>>                         -D$(MODEL) \
>>>>                         -DVERSION=\"$(GIT_VERSION)\" \
>>>>                         -I$(PETSC_DIR)/include
>>>> FPPFLAGS = -cpp -E -dM -C -P -D$(MODEL)
>>>> LDFLAGS = -lpetsc -L$(PETSC_DIR)/lib
>>>> LDFLAGS += -Wl,-rpath=$(PETSC_DIR)/lib
>>>>
>>>> Have i missed anything?
>>>>
>>>>
>>>> Regards,
>>>> Kiran
>>>>
>>>>
>>>>
>>>>
>>>> On Sun, Aug 3, 2025 at 9:56 PM Satish Balay <balay.anl at fastmail.org>
>>>> wrote:
>>>>
>>>>> On Sun, 3 Aug 2025, kiran hegde wrote:
>>>>>
>>>>> > Hi,
>>>>> >
>>>>> > Please find the attached configure.log file.
>>>>> > PETSC installation successfully completes with make check, without
>>>>> any
>>>>> > error.
>>>>> > But, when compiling fortan codes with   petsc  library+include in the
>>>>> > makefile i get this error.
>>>>>
>>>>> Then likely your compile command is missing some options. you can:
>>>>>
>>>>> - compile a petsc example - say:
>>>>> cd src/snes/tutorials
>>>>> make ex5f
>>>>>
>>>>> - now compile your application
>>>>>
>>>>> - And compare the compile commands and see if there are any
>>>>> differences (missing options, wrong order of include files etc..)
>>>>>
>>>>> Satish
>>>>>
>>>>
>>>>
>>>> --
>>>> Regards,
>>>>
>>>> Kiran Hegde
>>>> Mobile:+91 9972600235
>>>>
>>>>
>>>>
>>>
>>
>> --
>> Regards,
>>
>> Kiran Hegde
>> Mobile:+91 9972600235
>>
>>
>>
>

-- 
Regards,

Kiran Hegde
Mobile:+91 9972600235
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20250813/a313917b/attachment.html>