Hi Marc,
One way to achieve this is to factor out the initialization of the "di" variable into a non-inlined helper function. This will make sure it is not pulled into the loop. See below for the comparison:
function [xyz, di] = foo
xyz = zeros(2048,3);
for i = 1:size(xyz,1)
xyz(i,:)= 123;
end
di=nan(2048,1);
end
% void foo(double xyz[6144], double di[2048])
% {
% int i;
% for (i = 0; i < 2048; i++) {
% xyz[i] = 123.0;
% xyz[i + 2048] = 123.0;
% xyz[i + 4096] = 123.0;
% di[i] = rtNaN;
% }
% }
Compared to an example with the helper function:
function [xyz, di] = foo_noninline
xyz = zeros(2048,3);
for i = 1:size(xyz,1)
xyz(i,:)= 123;
end
di=initDI();
end
function di = initDI()
coder.inline('never');
di = nan(2048,1);
end
% static void initDI(double di[2048])
% {
% int i;
% for (i = 0; i < 2048; i++) {
% di[i] = rtNaN;
% }
% }
%
% /*
% * Arguments : double xyz[6144]
% * double di[2048]
% * Return Type : void
% */
% void foo_noninline(double xyz[6144], double di[2048])
% {
% int i;
% for (i = 0; i < 2048; i++) {
% xyz[i] = 123.0;
% xyz[i + 2048] = 123.0;
% xyz[i + 4096] = 123.0;
% }
%
% initDI(di);
% }
Hopefully this helps!
Matan
