zhangyiss/spack
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix GROMACS to require FFTW when using cuda (#19505)

Browse Source 
GROMACS still requires a version of FFTW when compiling it to utilize
NVIDIA GPUs. In fact, the type of calculation that depends on FFTW --
Particle-Mesh Ewald (PME) -- is generally run on the host system's CPUs,
even when GPUs are available.
This commit is contained in:
Marty Kandes 2020-10-23 16:58:13 -07:00 committed by GitHub
parent 978d8ad0af
commit 4cf85ed5f7
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
var/spack/repos/builtin/packages/gromacs/package.py
View File

@ -90,7 +90,7 @@ class Gromacs(CMakePackage):
depends_on('plumed@2.5.0:2.5.9~mpi', when='@2018.6+plumed~mpi')
depends_on('plumed+mpi', when='+plumed+mpi')
depends_on('plumed~mpi', when='+plumed~mpi')
depends_on('fftw-api@3', when='~cuda')
depends_on('fftw-api@3')
depends_on('mkl', when='fft=mkl')
depends_on('cmake@2.8.8:3.99.99', type='build')
depends_on('cmake@3.4.3:3.99.99', type='build', when='@2018:')