ExaGeoStatR is an R-Wrapper for [ExaGeoStat framework]((https://github.com/ecrc/exageostat), a parallel high performance unified software for geostatistics on manycore systems.
- Major changes in the structure of the package to meet CRAN requirements and to facilitate the installation on different platforms.
- Large-scale synthetic Geostatistics data generator.
- Support exact computation of the Maximum Likelihood Estimation (MLE) function using shared-memory, GPUS, or distributed-memory systems
- Support approximate computation (i.e., Diagonal Super-Tile (DST) and Tile Low-Rank (TLR) of the Maximum Likelihood Estimation (MLE) function using shared-memory, GPUS, or distributed-memory systems.
- BLAS/CBLAS/LAPACK/LAPACKE optimized implementation, ex., AMD Core Math Library (ACML), Arm Performance Libraries, ATLAS, Intel Math Kernel Library (MKL), or OpenBLAS.
- Portable Hardware Locality (hwloc).
- NLopt.
- GNU Scientific Library (GSL).
- StarPU.
- Chameleon.
- HiCMA.
- STARS-H.
All these dependencies are automatically installed with the package if not exist (OpenBLAS is the default BLAS library) on the system (ExaGeoStatR v1.0.1).
library("devtools")
Sys.setenv(MKLROOT="/opt/intel/mkl")
install_git(url="https://github.com/ecrc/exageostatR")library("devtools")
Sys.setenv(MKLROOT="/opt/intel/mkl")
Sys.setenv(MAKE="make -j 1")
install_git(url="https://github.com/ecrc/exageostatR")library("devtools")
Sys.setenv(MKLROOT="/opt/intel/mkl")
install_git(url="https://github.com/ecrc/exageostatR", configure.args=C('--enable-cuda'))library("devtools")
Sys.setenv(MKLROOT="/opt/intel/mkl")
install_git(url="https://github.com/ecrc/exageostatR", configure.args=C('--enable-mpi'))- Download exageostat_1.0.1.tar.gz from release)
- Use R to install exageostat_1.0.1.tar.gz)
install.packages(repos=NULL, "exageostat_1.0.1.tar.gz"))Operations:
- Generate synthetic spatial datasets (i.e., locations & environmental measurements).
- Maximum likelihood evaluation using dense matrices.
- Maximum likelihood evaluation using compressed matrices based on Tile Low-Rank(TLR).
- Maximum likelihood evaluation using matrices based on Diagonal Super-Tile(DST).
A more detailed description of the underlying ExaGeoStat software package can be found. here
- Test Generating Z vector using random (x, y) locations with exact MLE computation.
library("exageostatr") #Load ExaGeoStatR lib.
seed = 0 #Initial seed to generate XY locs.
sigma_sq = 1 #Initial variance.
beta = 0.1 #Initial range.
nu = 0.5 #Initial smoothness.
dmetric = "euclidean" #"euclidean", or "great_circle".
n = 1600 #n*n locations grid.
exageostat_init(hardware = list (ncores=2, ngpus=0,
ts=320, pgrid=1, qgrid=1)) #Initiate exageostat instance.
data = simulate_data_exact(sigma_sq, beta, nu,
dmetric, n, seed) #Generate Z observation vector.
result = exact_mle(data, dmetric, optimization = list(clb = c(0.001, 0.001, 0.001),
cub = c(5, 5,5 ), tol = 1e-4, max_iters = 20)) #Estimate MLE parameters (Exact).
exageostat_finalize() #Finalize exageostat instance.- Test Generating Z vector using random (x, y) locations with TLR MLE computation.
library("exageostatr") #Load ExaGeoStatR lib.
seed = 0 #Initial seed to generate XY locs.
sigma_sq = 1 #Initial variance.
beta = 0.03 #Initial range.
nu = 0.5 #Initial smoothness.
dmetric = "euclidean" #"euclidean", or "great_circle".
n = 900 #n*n locations grid.
tlr_acc = 7 #TLR accuracy 10^-(acc).
tlr_maxrank = 450 #TLR Max Rank.
exageostat_init(hardware = list (ncores=2, ngpus=0,
ts=320, lts=600, pgrid=1, qgrid=1)) #Initiate exageostat instance.
data = simulate_data_exact(sigma_sq, beta, nu,
dmetric, n, seed) #Generate Z observation vector.
result = tlr_mle(data, tlr_acc, tlr_maxrank, dmetric, optimization =
list(clb = c(0.001, 0.001, 0.001), cub = c(5, 5,5 ),
tol = 1e-4, max_iters = 20)) #Estimate MLE parameters (TLR).
exageostat_finalize() #Finalize exageostat instance.- Test Generating Z vector using random (x, y) locations with DST MLE computation.
library("exageostatr") #Load ExaGeoStatR lib.
seed = 0 #Initial seed to generate XY locs.
sigma_sq = 1 #Initial variance.
beta = 0.03 #Initial range.
nu = 0.5 #Initial smoothness.
dmetric = "euclidean" #"euclidean", or "great_circle".
n = 900 #n*n locations grid.
dst_band = 3 #Number of diagonal double tiles.
exageostat_init(hardware = list (ncores=4, ngpus=0,
ts=320, lts=0, pgrid=1, qgrid=1)) #Initiate exageostat instance.
data = simulate_data_exact(sigma_sq, beta, nu,
dmetric, n, seed) #Generate Z observation vector.
result = dst_mle(data, dst_band, dmetric, optimization =
list(clb = c(0.001, 0.001, 0.001), cub = c(5, 5,5 ),
tol = 1e-4, max_iters = 20)) #Estimate MLE parameters (DST).
exageostat_finalize() #Finalize exageostat instance.- Test Generating Z vector using given (x, y) locations with exact MLE computation.
library("exageostatr") #Load ExaGeoStatR lib.
sigma_sq = 1 #Initial variance.
beta = 0.1 #Initial range.
nu = 0.5 #Initial smoothness.
dmetric = "euclidean" #"euclidean", or "great_circle",
n = 1600 #n*n locations grid.
x = rnorm(n = 1600, mean = 39.74, sd = 25.09) #x measurements of n locations.
y = rnorm(n = 1600, mean = 80.45, sd = 100.19) #y measurements of n locations.
exageostat_init(hardware = list (ncores=2, ngpus=0,
ts=320, lts=0, pgrid=1, qgrid=1)) #Initiate exageostat instance.
data = simulate_obs_exact( x, y, sigma_sq,
beta, nu, dmetric) #Generate Z observation vector.
result = exact_mle(data, dmetric, optimization =
list(clb = c(0.001, 0.001, 0.001), cub = c(5, 5,5 ), tol = 1e-4, max_iters = 20))
exageostat_finalize() #Finalize exageostat instance.#!/bin/bash
#SBATCH --job-name=job_name
#SBATCH --output=output_file.txt
#SBATCH --partition=XXXX
#SBATCH --nodes=4
#SBATCH --ntasks=4
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=31
#SBATCH --time 00:30:00
# RExample.r includes one of the examples above
srun Rscript RExample.r