64-bit version: Segmentation fault with Intel MPI64 位版本：使用 Intel MPI 时出现段错误

Compiling the 64-bit version of DALTON 2020.1 with Intel's OneAPI toolchain using a setup command like
使用 Intel 的 OneAPI 工具链编译 64 位版本的 DALTON 2020.1，采用类似设置命令的方式。

$ ./setup --fc mpiifx --cc mpiicx --cxx mpiicpx --mkl parallel --mpi --int64

for OneAPI 2024 leads to a segmentation fault after the SYMGRP output when running calculations with more than 1 MPI process. This happens on all OneAPI versions, using both the classic (ifort etc.) and LLVM (ifx etc.) compilers.
对于 OneAPI 2024，在使用超过 1 个 MPI 进程进行计算时，执行 SYMGRP 输出后会导致段错误。此问题在所有 OneAPI 版本中均出现，无论是使用经典编译器（如 ifort 等）还是 LLVM 编译器（如 ifx 等）。

The segfault occurs in gen1int/gen1int_host.F90:82 in a call to MPI_Bcast; I have attached a sample calculation with debug flags and backtrace:
段错误发生在 gen1int/gen1int_host.F90:82 中，在调用 MPI_Bcast 时发生；我附上了一个带有调试标志和回溯的示例计算：

dalton.tar.gz

I am fairly certain this issue is due to the use of the mpif.h header in DALTON, which defines MPI_INTEGER_KIND to be 4 bytes:
我相当确定这个问题是由于在 DALTON 中使用了 mpif.h 头文件，该头文件定义 MPI_INTEGER_KIND 为 4 字节：

$ grep -n MPI_INTEGER_KIND $I_MPI_ROOT/include/mpif.h
406:       INTEGER MPI_INTEGER_KIND
407:       PARAMETER (MPI_INTEGER_KIND=4)

Manually changing all #include "mpif.h" statements to use mpi statements before the implicit line makes the segfault disappear and calculations run as expected; however, I've seen some comments in the code saying that use mpi causes problems with MPICH in return. Therefore, seeing as there is already a USE_MPI_MOD_F90 preprocessor flag, would it be possible to automate the usage of the header or module through something like
在 implicit 行之前手动将所有 #include "mpif.h" 语句更改为 use mpi 语句，可以使段错误消失，计算也能按预期运行；然而，我在代码中看到一些评论提到 use mpi 会导致 MPICH 出现问题。因此，鉴于已经有一个 USE_MPI_MOD_F90 预处理器标志，是否可以通过类似的方式自动化使用头文件或模块？

#include "mpi_mod.h"
#include "implicit.h"
#include "mpi_header.h"

with 与

! mpi_mod.h
#if defined(VAR_MPI) && defined(USE_MPI_MOD_F90)
      use mpi
#endif

and 和

! mpi_header.h
#if defined(VAR_MPI) && !defined(USE_MPI_MOD_F90)
#include "mpif.h"
#endif

Having to remember to include two headers is certainly not ideal, but since use mpi has to come before the implicit statement and the #include "mpif.h" has to come after (?), I don't really see another way.
必须记住包含两个头文件当然不理想，但由于 use mpi 必须在 implicit 语句之前，而 #include "mpif.h" 必须在之后（？），我实在想不出其他办法。

Let me know if you need any additional info. Also, I'd be happy to submit a PR if you're fine with the proposed solution.
如有需要其他信息，请告知。此外，如果您对提议的解决方案没问题，我很乐意提交一个 PR。

Best, Yannick Lemke 最好的，扬尼克·伦克

Edit: I just noticed this is not the actual issue tracker, should I move this to https://gitlab.com/dalton/dalton/-/issues?
编辑：我刚注意到这不是实际的问题跟踪器，我应该将其移至 https://gitlab.com/dalton/dalton/-/issues 吗？

changed the description

moved from user-support#80 (已移动)

Thanks for this. I have moved it.
谢谢。我已经移动了它。

@pro7on if you already implemented the changes then it'd be great if you would create a merge request with your changes.
@pro7on 如果你已经实施了这些更改，那么如果你能用你的更改创建一个合并请求，那就太好了。

@hjaaj can you take a look at this?
@hjaaj 你能看一下这个吗？