====== Matlab with XeonPhi ======
Using automtaic-offload to XeonPhi. The key is to tell Matlab to use **MKL** subroutine for **blas** and **lapack**
**Example**
version -blas
version -lapack
getenv('MKL_MIC_ENABLE')
% generate random n x n matrix
A = rand(25000);
% Matrix Multiply
tic
B = A * A;
toc
% LU Matrix Factorization
tic
B = lu(A);
toc
**SGE script**
#!/bin/bash
#$ -q xphi.q
#S -l h_vmem=10G
#$ -l excl ##request exclusive access
# load Matlab and Intel environment
module load intel/mic
module load matlab
## Tell matlab to use MKL
export BLAS_VERSION=$MKLROOT/lib/intel64/libmkl_rt.so
export LAPACK_VERSION=$MKLROOT/lib/intel64/libmkl_rt.so
export MKL_MIC_MAX_MEMORY=16G
# This single line allows MKL to decide whether to offload
# certain matrix operations to one or both MICs
export MKL_MIC_ENABLE=1
# This next line reports what was
# offloaded to the MIC
export OFFLOAD_REPORT=2
# Run program
matlab -nodisplay -nosplash < matrix.m