Skip to content

Commit

Permalink
CUDA experiment for LLNL-Sierra-IBM-power9-V100-Infiniband
Browse files Browse the repository at this point in the history
  • Loading branch information
Riyaz Haque committed Feb 10, 2024
1 parent 4347564 commit 0738ba5
Show file tree
Hide file tree
Showing 6 changed files with 59 additions and 17 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -65,3 +65,21 @@ compilers:
modules: []
environment: {}
extra_rpaths: []
- compiler:
spec: [email protected]
paths:
cc: /usr/tce/packages/clang/clang-16.0.6-cuda-11.8.0-gcc-11.2.1/bin/clang-gpu
cxx: /usr/tce/packages/clang/clang-16.0.6-cuda-11.8.0-gcc-11.2.1/bin/clang++-gpu
f77: /usr/tce/packages/gcc/gcc-11.2.1/bin/gfortran
fc: /usr/tce/packages/gcc/gcc-11.2.1/bin/gfortran
flags:
cflags: -g -O2
cxxflags: -g -O2 -std=c++17
fflags: ''
operating_system: rhel7
target: ppc64le
modules: []
environment:
set:
GXX_PREFIX: "/usr/tce/packages/gcc/gcc-11.2.1"
extra_rpaths: []
Original file line number Diff line number Diff line change
Expand Up @@ -48,26 +48,45 @@ packages:
- spec: [email protected]
prefix: /usr/tce/packages/cuda/cuda-11.8.0
buildable: false
blas:
cublas:
externals:
- spec: [email protected]
prefix: /usr/tce/packages/cuda/cuda-10.1.243
- spec: [email protected]
prefix: /usr/tce/packages/cuda/cuda-11.8.0
buildable: false
cusolver:
externals:
- spec: [email protected]
prefix: /usr/tce/packages/cuda/cuda-11.8.0
buildable: false
blas:
externals:
- spec: [email protected]
prefix: /usr/tcetmp/packages/blas/blas-3.6.0-gfortran-4.8.5
buildable: false
fftw:
externals:
- spec: [email protected]
prefix: /usr/tcetmp/packages/fftw/fftw-3.3.10-xl-2023.06.28
buildable: false
lapack:
externals:
- spec: lapack-xl@3.9.0
prefix: /usr/tcetmp/packages/lapack/lapack-3.9.0-xl-2020.03.18
- spec: lapack-xl@3.6.0
prefix: /usr/tcetmp/packages/lapack/lapack-3.6.0-gfortran-4.8.5
buildable: false
mpi:
externals:
- spec: [email protected]cuda-11.8.0
- spec: [email protected]cuda11.8.0
prefix: /usr/tce/packages/spectrum-mpi/spectrum-mpi-rolling-release-xl-2022.08.19-cuda-11.8.0
extra_attributes:
ldflags: "-lmpiprofilesupport -lmpi_ibm_usempi -lmpi_ibm_mpifh -lmpi_ibm"
- spec: [email protected]cuda-10.1.243
- spec: [email protected]cuda10.1.243
prefix: /usr/tce/packages/spectrum-mpi/spectrum-mpi-rolling-release-xl-2022.08.19-cuda-10.1.243
extra_attributes:
ldflags: "-lmpiprofilesupport -lmpi_ibm_usempi -lmpi_ibm_mpifh -lmpi_ibm"
- spec: [email protected]
prefix: /usr/tce/packages/spectrum-mpi/spectrum-mpi-rolling-release-clang-16.0.6-cuda-11.8.0-gcc-11.2.1
extra_attributes:
ldflags: "-lmpiprofilesupport -lmpi_ibm_usempi -lmpi_ibm_mpifh -lmpi_ibm"
buildable: false
12 changes: 8 additions & 4 deletions configs/LLNL-Sierra-IBM-power9-V100-Infiniband/spack.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -6,12 +6,16 @@
spack:
packages:
default-compiler:
spack_spec: xl@16.1.1-2022.08.19-cuda{default_cuda_version}
spack_spec: clang@16.0.6-cuda{default_cuda_version}
default-mpi:
spack_spec: [email protected]
spack_spec: [email protected]-clang16.0.6-cuda{default_cuda_version}
compiler-gcc:
spack_spec: [email protected]
blas:
spack_spec: [email protected]
cublas-cuda:
spack_spec: cublas@{default_cuda_version}
spack_spec: cublas@{default_cuda_version}
lapack:
spack_spec: [email protected]
spack_spec: [email protected]
fftw:
spack_spec: [email protected]
2 changes: 0 additions & 2 deletions experiments/gromacs/cuda/execute_experiment.tpl
Original file line number Diff line number Diff line change
Expand Up @@ -11,8 +11,6 @@

cd {experiment_run_dir}

{spack_setup}

{experiment_setup}

{command}
10 changes: 4 additions & 6 deletions experiments/gromacs/cuda/ramble.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -22,8 +22,8 @@ ramble:
variables:
experiment_setup: ''
n_ranks: '{processes_per_node} * {n_nodes}'
processes_per_node: '4'
n_nodes: '2'
processes_per_node: '2'
n_nodes: '4'
omp_num_threads: '22'
target: 'gpu'
experiments:
Expand All @@ -44,15 +44,13 @@ ramble:
spack:
concretized: true
packages:
fftw:
spack_spec: [email protected] +mpi+openmp
compiler: default-compiler
gromacs:
spack_spec: [email protected] +mpi+openmp+cuda~hwloc cuda_arch=={cuda_arch} ^cuda@{default_cuda_version}
spack_spec: [email protected] +mpi+openmp+cuda~hwloc~double cuda_arch=={cuda_arch} ^cuda@{default_cuda_version}
compiler: default-compiler
environments:
gromacs:
packages:
- blas
- lapack
- default-mpi
- fftw
Expand Down
5 changes: 5 additions & 0 deletions repo/gromacs/package.py
Original file line number Diff line number Diff line change
Expand Up @@ -494,6 +494,9 @@ def cmake_args(self):
else:
options.append("-DGMX_GPLUSPLUS_PATH=%s/g++" % self.spec["gcc"].prefix.bin)

if target.family == "ppc64le":
options.append("-DGMX_GPLUSPLUS_PATH=%s/g++" % '/usr/tce/packages/gcc/gcc-11.2.1/bin')

if "+double" in self.spec:
options.append("-DGMX_DOUBLE:BOOL=ON")

Expand Down Expand Up @@ -585,6 +588,8 @@ def cmake_args(self):
elif target >= "bulldozer":
# AMD Family 15h
options.append("-DGMX_SIMD=AVX_128_FMA")
elif target.family == "ppc64le":
options.append("-DGMX_SIMD=None")
elif "vsx" in target:
# IBM Power 7 and beyond
if self.spec.satisfies("%nvhpc"):
Expand Down

0 comments on commit 0738ba5

Please sign in to comment.