Updated documentation for FFT_KOKKOS_ flags and CMake variable selection

doc/src/Build_settings.rst

@@ -51,14 +51,18 @@ LAMMPS can use them if they are available on your system.
       .. code-block:: bash
 
          -D FFT=value              # FFTW3 or MKL or KISS, default is FFTW3 if found, else KISS
+         -D FFT_KOKKOS=value       # FFTW3 or MKL or KISS or CUFFT or HIPFFT, default is KISS
          -D FFT_SINGLE=value       # yes or no (default), no = double precision
          -D FFT_PACK=value         # array (default) or pointer or memcpy
 
       .. note::
 
-         The values for the FFT variable must be in upper-case.  This is
-         an exception to the rule that all CMake variables can be specified
-         with lower-case values.
+         When the Kokkos variant of a package is compiled and selected at run time,
+         the FFT library selected by the FFT_KOKKOS variable applies. Otherwise,
+         the FFT library selected by the FFT variable applies.
+         The same FFT settings apply to both. FFT_KOKKOS must be compatible with the
+         Kokkos backend - for example, when using the CUDA backend of Kokkos,
+         you must use either CUFFT or KISS.
 
       Usually these settings are all that is needed.  If FFTW3 is
       selected, then CMake will try to detect, if threaded FFTW
@@ -87,6 +91,8 @@ LAMMPS can use them if they are available on your system.
 
          FFT_INC = -DFFT_FFTW3         # -DFFT_FFTW3, -DFFT_FFTW (same as -DFFT_FFTW3), -DFFT_MKL, or -DFFT_KISS
                                        # default is KISS if not specified
+         FFT_INC = -DFFT_KOKKOS_CUFFT  # -DFFT_KOKKOS_{FFTW,FFTW3,MKL,CUFFT,HIPFFT,KISS}
+                                       # default is KISS if not specified
          FFT_INC = -DFFT_SINGLE        # do not specify for double precision
          FFT_INC = -DFFT_FFTW_THREADS  # enable using threaded FFTW3 libraries
          FFT_INC = -DFFT_MKL_THREADS   # enable using threaded FFTs with MKL libraries
@@ -97,6 +103,8 @@ LAMMPS can use them if they are available on your system.
 
          FFT_INC =       -I/usr/local/include
          FFT_PATH =      -L/usr/local/lib
+         FFT_LIB =       -lhipfft            # hipFFT either precision
+         FFT_LIB =       -lcufft             # cuFFT either precision
          FFT_LIB =       -lfftw3             # FFTW3 double precision
          FFT_LIB =       -lfftw3 -lfftw3_omp # FFTW3 double precision with threads (needs -DFFT_FFTW_THREADS)
          FFT_LIB =       -lfftw3 -lfftw3f    # FFTW3 single precision
@@ -141,6 +149,10 @@ The Intel MKL math library is part of the Intel compiler suite.  It
 can be used with the Intel or GNU compiler (see the ``FFT_LIB`` setting
 above).
 
+The CUFFT and HIPFFT FFT libraries are packaged with NVIDIA's CUDA and AMD's
+HIP installations, respectively. These FFT libraries require the Kokkos acceleration
+package to be enabled and the Kokkos backend to be GPU-resident (ie, HIP or CUDA).
+
 Performing 3d FFTs in parallel can be time-consuming due to data access
 and required communication.  This cost can be reduced by performing
 single-precision FFTs instead of double precision.  Single precision

doc/src/Howto_cmake.rst

@@ -349,6 +349,8 @@ Some common LAMMPS specific variables
      - when set to ``name`` the LAMMPS executable and library will be called ``lmp_name`` and ``liblammps_name.a``
    * - ``FFT``
      - select which FFT library to use: ``FFTW3``, ``MKL``, ``KISS`` (default, unless FFTW3 is found)
+   * - ``FFT_KOKKOS``
+     - select which FFT library to use in Kokkos-enabled styles: ``FFTW3``, ``MKL``, ``HIPFFT``, ``CUFFT``, ``KISS`` (default)
    * - ``FFT_SINGLE``
      - select whether to use single precision FFTs (default: ``off``)
    * - ``WITH_JPEG``

doc/src/kspace_style.rst

@@ -450,7 +450,10 @@ relative RMS error.
   For the KOKKOS package, the *pppm/kk* style performs charge
   assignment and force interpolation calculations, along with the FFTs
   themselves, on the GPU or (optionally) threaded on the CPU when
-  using OpenMP and FFTW3.
+  using OpenMP and FFTW3. The specific FFT library is selected using
+  the FFT_KOKKOS CMake parameter. See the 
+  :doc:`Build settings <Build_settings>` doc page for how to select a
+  3rd-party FFT library.
 
 ----------