matrix element-wise multiplication (MatLab)

tnum = 1e4;
a = 101;
b = 700;
c = 201;
d = 800;
Ez = rand(b+1, d);
Ez2 = Ez;
Hx = rand(b+1, d);
Hx2 = Hx;
A = rand(b-a+1, d-c+1);
B = rand(b-a+1, d-c+1);
C = rand(b-a+1, d-c+1);
D = rand(b-a+1, d-c+1);

% MATLAB
tic
for t=1:tnum
  Ez(a:b,c:d)=A.*Ez(a:b,c:d)+B.*(Hx(a+1:b+1,c:d)-Hx(a:b,c:d));
  Hx(a:b,c:d)=C.*Hx(a:b,c:d)+D.*(Ez(a+1:b+1,c:d)-Ez(a:b,c:d));
end
toc

% MEX with armadillo and MKL BLAS
tic
[Ez3, Hx3] = test_mkl_blas(tnum, a, b, c, d, Ez2, Hx2, A, B, C, D);
toc

% compile commands
MKLROOT = 'C:\Program Files (x86)\Intel\Composer XE 2015\mkl';
ICLLIB = 'C:\Program Files (x86)\Intel\Composer XE 2015\compiler\lib\intel64';
OPTIMFLAGS = 'OPTIMFLAGS= /O3 /DNDEBUG /QxHost /Qparallel /Qopenmp /Qprec-div- /Qipo /Qinline-calloc';
LINKER = 'LINKER=xilink';

mex('-f', 'intel_cpp_15_vs2013.xml', '-v', '-largeArrayDims', ['-L', ICLLIB], ...
  'test_c.cpp', OPTIMFLAGS, LINKER)

mex('-f', 'intel_cpp_15_vs2013.xml', '-v', '-largeArrayDims', '-IC:\armadillo\include', ...
    ['-I', MKLROOT, '\include'], ['-L', MKLROOT, '\lib\intel64'], ['-L', ICLLIB], ...
    'test_mkl_blas.cpp', OPTIMFLAGS, LINKER, '-lmkl_intel_lp64', '-lmkl_intel_thread', ...
    '-lmkl_core', '-llibiomp5md')