(ch_fortran)= # PETSc for Fortran Users Make sure the suffix of your Fortran files is .F90, not .f or .f90. ## Fortran and MPI By default PETSc uses the MPI Fortran module `mpi`. To use `mpi_f08` run `./configure` with `--with-mpi-ftn-module=mpi_f08`. We do not recommend it but it is possible to write Fortran code that works with both `use mpi` or `use mpi_f08`. You must declare MPI objects using `MPIU_XXX` (for example `MPIU_Comm`, which the PETSc include files map to either `integer4` or `type(MPI_XXX)`. In addition, you must handle `MPIU_Status` declarations and access to entries using the Fortran preprocessor. For example, ```fortran #if defined(PETSC_USE_MPI_F08) MPIU_Status status #else MPIU_Status status(MPI_STATUS_SIZE) #endif ``` and then ```fortran #if defined(PETSC_USE_MPI_F08) tag = status%MPI_TAG #else tag = status(MPI_TAG) #endif ``` ## Numerical Constants Since Fortran compilers do not automatically change the length of numerical constant arguments (integer and real) in subroutine calls to the expected length PETSc provides parameters that indicate the constants' kind. ```fortran interface subroutine SampleSubroutine(real, complex, integer, MPIinteger) PetscReal real ! real(PETSC_REAL_KIND) PetscComplex complex ! real(PETSC_REAL_KIND) or complex(PETSC_REAL_KIND) PetscInt integer ! integer(PETSC_INT_KIND) PetscMPIInt MPIinteger ! integer(PETSC_MPIINT_KIND) end subroutine end interface ... ! Fortran () automatically sets the complex KIND to correspond to the KIND of constant arguments call SampleSubroutine(real = 1.0_PETSC_REAL_KIND, complex = (0.1_PETSC_REAL_KIND, 0.0_PETSC_REAL_KIND), integer = 1_PETSC_INT_KIND, MPIinteger = 1_PETSC_MPIINT_KIND) ! For variable arguments one must set the complex kind explicitly or it defaults to single precision PetscReal a = 0.1, b = 0.0 call SampleSubroutine(real = 1.0_PETSC_REAL_KIND, complex = cmplx(a, b, PETSC_REAL_KIND), integer = 1_PETSC_INT_KIND, MPIinteger = 1_PETSC_MPIINT_KIND) ``` ## Basic Fortran API Differences (sec_fortran_includes)= ### Modules and Include Files You must use both PETSc include files and modules. At the beginning of every function and module definition you need something like ```fortran #include "petsc/finclude/petscXXX.h" use petscXXX ``` The Fortran include files for PETSc are located in the directory `$PETSC_DIR/$PETSC_ARCH/include/petsc/finclude` and the module files are located in `$PETSC_DIR/$PETSC_ARCH/include` The include files are nested, that is, for example, `petsc/finclude/petscmat.h` automatically includes `petsc/finclude/petscvec.h` and so on. The modules are also nested. One can use ```c use petsc ``` to include all of them. ### Declaring PETSc Object Variables You can declare PETSc object variables using either of the following: ```fortran XXX variablename ``` ```fortran type(tXXX) variablename ``` For example, ```fortran #include "petsc/finclude/petscvec.h" use petscvec Vec b type(tVec) x ``` PETSc types like `PetscInt` and `PetscReal` are simply aliases for basic Fortran types and cannot be written as `type(tPetscInt)` PETSc objects are always automatically initialized when declared so you do not need to (and should not) do ```fortran type(tXXX) x = PETSC_NULL_XXX XXX x = PETSC_NULL_XXX ``` To make a variable no longer point to its previously assigned PETSc object use, for example, ```fortran Vec x, y PetscInt one = 1 PetscCallA(VecCreateMPI(PETSC_COMM_WORLD, one, PETSC_DETERMINE, x, ierr)) y = x PetscCallA(VecDestroy(x, ierr)) PetscObjectNullify(y) ``` Otherwise `y` will be a dangling pointer whose access will cause a crash. ### Calling Sequences The calling sequences for the Fortran version are in most cases identical to the C version, except for the error checking variable discussed in {any}`sec_fortran_errors`. The key differences in handling arguments when calling PETSc functions from Fortran are - One cannot pass a scalar variable to a function expecting an array, {any}`sec_passarray`. - One must use type specific `PETSC_NULL` arguments, such as `PETSC_NULL_INTEGER`, {any}`sec_nullptr`. - One must pass pointers to arrays for arguments that output an array, for example `PetscScalar, pointer \:\: a(\:)`, {any}`sec_fortranarrays`. - `PETSC_DECIDE` and friends need to match the argument type, for example `PETSC_DECIDE_INTEGER`. When passing floating point numbers into PETSc Fortran subroutines, always make sure you have them marked as double precision (e.g., pass in `10.d0` instead of `10.0` or declare them as PETSc variables, e.g. `PetscScalar one = 1.0`). Otherwise, the compiler interprets the input as a single precision number, which can cause crashes or other mysterious problems. We **highly** recommend using the `implicit none` option at the beginning of each Fortran subroutine and declaring all variables. (sec_fortran_errors)= ### Error Checking In the Fortran version, each PETSc routine has as its final argument an integer error variable. The error code is nonzero if an error has been detected; otherwise, it is zero. For example, the Fortran and C variants of `KSPSolve()` are given, respectively, below, where `ierr` denotes the `PetscErrorCode` error variable: ```fortran call KSPSolve(ksp, b, x, ierr) ! Fortran ierr = KSPSolve(ksp, b, x); // C ``` For proper error handling one should not use the above syntax instead one should use ```fortran PetscCall(KSPSolve(ksp, b, x, ierr)) ! Fortran subroutines PetscCallA(KSPSolve(ksp, b, x, ierr)) ! Fortran main program PetscCall(KSPSolve(ksp, b, x)) // C ``` (sec_passarray)= ### Passing Arrays To PETSc Functions Many PETSc functions take arrays as arguments; in Fortran they must be passed as arrays even if the "array" is of length one (unlike Fortran 77 where one can pass scalars to functions expecting arrays). When passing a single value one can use the Fortran [] notation to pass the scalar as an array, for example ```fortran PetscCall(VecSetValues(v, one, [i], [val], ierr)) ``` This trick can only be used for arrays used to pass data into a PETSc routine, it cannot be used for arrays used to receive data from a PETSc routine. For example, ```fortran PetscCall(VecGetValues(v, one, idx, [val], ierr)) ``` is invalid and will not set `val` with the correct value. (sec_nullptr)= ### Passing null pointers to PETSc functions Many PETSc C functions have the option of passing a `NULL` argument (for example, the fifth argument of `MatCreateSeqAIJ()`). From Fortran, users *must* pass `PETSC_NULL_XXX` to indicate a null argument (where `XXX` is `INTEGER`, `DOUBLE`, `CHARACTER`, `SCALAR`, `VEC`, `MAT`, etc depending on the argument type). For example, when no options prefix is desired in the routine `PetscOptionsGetInt()`, one must use the following command in Fortran: ```fortran PetscCall(PetscOptionsGetInt(PETSC_NULL_OPTIONS, PETSC_NULL_CHARACTER, PETSC_NULL_CHARACTER, '-name', N, flg, ierr)) ``` Where the code expects an array, then use `PETSC_NULL_XXX_ARRAY`. For example: ```fortran PetscCall(MatCreateSeqDense(comm, m, n, PETSC_NULL_SCALAR_ARRAY, A)) ``` When a PETSc function returns multiple arrays, such as `DMDAGetOwnershipRanges()` and the user does not need certain arrays they must pass `PETSC_NULL_XXX_POINTER` as the argument. For example, ```fortran PetscInt, pointer :: lx(:), ly(:) PetscCallA(DMDAGetOwnershipRanges(dm, lx, ly, PETSC_NULL_INTEGER_POINTER, ierr)) PetscCallA(DMDARestoreOwnershipRanges(dm, lx, ly, PETSC_NULL_INTEGER_POINTER, ierr)) ``` Arguments that are fully defined Fortran derived types (C structs), such as `MatFactorInfo` or `PetscSFNode`, cannot be passed as null from Fortran. A properly defined variable must be passed in for those arguments. Finally when a subroutine returns a `PetscObject` through an argument, to check if it is `NULL` you must use: ```fortran if (PetscObjectIsNull(dm)) then if (.not. PetscObjectIsNull(dm)) then ``` you cannot use ```fortran if (dm .eq. PETSC_NULL_DM) then ``` Note that ```fortran if (PetscObjectIsNull(PETSC_NULL_VEC)) then ``` will always return true, for any PETSc object. These specializations with `NULL` types are required because of Fortran's strict type checking system and lack of a concept of `NULL`, the Fortran compiler will often warn you if the wrong `NULL` type is passed. (sec_fortranarrays)= ### Output Arrays from PETSc functions For PETSc routine arguments that return an array of `PetscInt`, `PetscScalar`, `PetscReal` or of PETSc objects, one passes in a pointer to an array and the PETSc routine returns an array containing the values. For example, ```c PetscScalar *a; Vec v; VecGetArray(v, &a); ``` is in Fortran, ```fortran PetscScalar, pointer :: a(:) Vec, v VecGetArray(v, a, ierr) ``` For PETSc routine arguments that return a character string (array), e.g. `const char *str[]` pass a string long enough to hold the result. For example, ```fortran character*(80) str PetscCall(KSPGetType(ksp,str,ierr)) ``` The result is copied into `str`. Similarly, for PETSc routines where the user provides a character array (to be filled) followed by the array's length, e.g. `char name[], size_t nlen`. In Fortran pass a string long enough to hold the result, but not the separate length argument. For example, ```fortran character*(80) str PetscCall(PetscGetHostName(name,ierr)) ``` ### Matrix, Vector and IS Indices All matrices, vectors and `IS` in PETSc use zero-based indexing in the PETSc API regardless of whether C or Fortran is being used. For example, `MatSetValues()` and `VecSetValues()` always use zero indexing. See {any}`sec_matoptions` for further details. Indexing into Fortran arrays, for example obtained with `VecGetArray()`, uses the Fortran convention and generally begin with 1 except for special routines such as `DMDAVecGetArray()` which uses the ranges provided by `DMDAGetCorners()`. (sec_fortran_context)= ### Setting Routines and Contexts Some PETSc functions take as arguments user-functions and contexts for the function. For example ```fortran external func SNESSetFunction(snes, r, func, ctx, ierr) SNES snes Vec r PetscErrorCode ierr ``` where `func` has the calling sequence ```fortran subroutine func(snes, x, f, ctx, ierr) SNES snes Vec x,f PetscErrorCode ierr ``` and `ctx` can be almost anything (represented as `void *` in C). In Fortran, it has to be a derived type as in ```fortran subroutine func(snes, x, f, ctx, ierr) SNES snes Vec x,f type (AppCtx) ctx PetscErrorCode ierr ... external func SNESSetFunction(snes, r, func, ctx, ierr) SNES snes Vec r PetscErrorCode ierr type (AppCtx) ctx ``` or a PETSc object ```fortran subroutine func(snes, x, f, ctx, ierr) SNES snes Vec x,f Vec ctx PetscErrorCode ierr ... external func SNESSetFunction(snes, r, func, ctx, ierr) SNES snes Vec r PetscErrorCode ierr Vec ctx ``` or nothing ```fortran subroutine func(snes, x, f, dummy, ierr) SNES snes Vec x,f integer dummy(*) PetscErrorCode ierr ... external func SNESSetFunction(snes, r, func, 0, ierr) SNES snes Vec r PetscErrorCode ierr ``` Certain PETSc functions return a context in an argument, for example, `SNESGetApplicationContext()`. In C, they are handled as a `void *` pointer so one can write code such as ```c AppCtx *ctx; SNESGetApplicationContext(snes, &ctx); ``` In Fortran, they must be declared as a pointer with, for example, ```fortran type(AppCtx), pointer :: ctx call SNESGetApplicationContext(snes, ctx) ``` But sadly, this alone will not work. One must also specifically tell the Fortran compiler in an interface definition that `SNESGetApplicationContext()` expects the `ctx` argument to be a pointer to `type(AppCtx)` since Fortran forbids pointers to unknown types. PETSc provides macros to provide this information easily using, for example, ```fortran Interface_SNESGetApplicationContext(AppCtx) end interface ``` One must insert these lines into the Fortran source code where one inserts interface definitions, see, for example, src/snes/tests/ex590.F90. For those interested, the source code of these macros may be found in the generated Fortran include files located at `$PETSC_DIR/$PETSC_ARCH/include/petsc/finclude/*.h`. When a function pointer (declared as external in Fortran) is passed as an argument to a PETSc function, it is assumed that this function references a routine written in the same language as the PETSc interface function that was called. For instance, if `SNESSetFunction()` is called from C, the function must be a C function. Likewise, if it is called from Fortran, the function must be (a subroutine) written in Fortran. If you are using Fortran classes that have bound functions (methods) as in src/snes/tests/ex18f90.F90, the context cannot be passed to function pointer setting routines, such as `SNESSetFunction()`. Instead, one must use `SNESSetFunctionNoInterface()`, and define the interface directly in the user code, see ex18f90.F90 for a full demonstration. (sec_fortcompile)= ### Compiling and Linking Fortran Programs See {any}`sec_writing_application_codes`. (sec_fortran_examples)= ## Sample Fortran Programs Sample programs that illustrate the PETSc interface for Fortran are given below, corresponding to Vec Test ex19f, Vec Tutorial ex4f, Draw Test ex5f, and SNES Tutorial ex1f, respectively. We also refer Fortran programmers to the C examples listed throughout the manual, since PETSc usage within the two languages differs only slightly. :::{admonition} Listing: `src/vec/vec/tests/ex19f.F90` :name: vec-test-ex19f ```{literalinclude} /../src/vec/vec/tests/ex19f.F90 :end-at: end :language: fortran ``` ::: (listing_vec_ex4f)= :::{admonition} Listing: `src/vec/vec/tutorials/ex4f.F90` :name: vec-ex4f ```{literalinclude} /../src/vec/vec/tutorials/ex4f.F90 :end-before: '!/*TEST' :language: fortran ``` ::: :::{admonition} Listing: `src/sys/classes/draw/tests/ex5f.F90` :name: draw-test-ex5f ```{literalinclude} /../src/sys/classes/draw/tests/ex5f.F90 :end-at: end :language: fortran ``` ::: :::{admonition} Listing: `src/snes/tutorials/ex1f.F90` :name: snes-ex1f ```{literalinclude} /../src/snes/tutorials/ex1f.F90 :end-before: '!/*TEST' :language: fortran ``` ::: ### Calling Fortran Routines from C (and C Routines from Fortran) The information here applies only if you plan to call your **own** C functions from Fortran or Fortran functions from C. Different compilers have different methods of naming Fortran routines called from C (or C routines called from Fortran). Most Fortran compilers change the capital letters in Fortran routines to all lowercase. With some compilers, the Fortran compiler appends an underscore to the end of each Fortran routine name; for example, the Fortran routine `Dabsc()` would be called from C with `dabsc_()`. Other compilers change all the letters in Fortran routine names to capitals. PETSc provides two macros (defined in C/C++) to help write portable code that mixes C/C++ and Fortran. They are `PETSC_HAVE_FORTRAN_UNDERSCORE` and `PETSC_HAVE_FORTRAN_CAPS` , which will be defined in the file `$PETSC_DIR/$PETSC_ARCH/include/petscconf.h` based on the compilers conventions. The macros are used, for example, as follows: ```fortran #if defined(PETSC_HAVE_FORTRAN_CAPS) #define dabsc_ DABSC #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE) #define dabsc_ dabsc #endif ..... dabsc_( &n,x,y); /* call the Fortran function */ ```