MATLAB: Sparse half-precision matrices

classhalf

I'm experimenting with solving linear systems in half precision using matlab. I'm using the somewhat recently introduced 'half' data type. I can convert full matrices to half precision no problem. But if I then try to call sparse(M) I get an error saying "Undefined function 'sparse' for input arguments of type 'half'".

Is there a way I can define this function myself? I don't know too much about MATLAB so I'm unsure of where to start.

Best Answer

The sparse format in MATLAB only supports double and logical data types. To use any other data type you would have to write all of the underlying code yourself (a very big undertaking).

Related Solutions

MATLAB: Split a matrix into smaller sub matricies

Using SEPBLOCKFUN ( Download )

   result = sepblockfun(yourMatrix, [10,10], 'sum');

MATLAB: Is half-precision slower than double-precision

This is true.

n = 100;
t1= zeros(1,n);
t2= zeros(1,n);
for i = 1:n
a = ones(10,10,100,100);
b = zeros(10,10,100,100);
a2 = half(a);
b2 = half(b);
tic
temp = plus(a,b);
c = sum(temp(:));
t1(i) = toc;
tic
temp2 = plus(a2,b2);
c2 = sum(temp2(:));
t2(i) = toc;
end
sum(t1)
sum(t2)
ans = 
0.1456
ans=
0.2737

Computations with half-precision data types is slower than those with double precision data types. This is because, unlike double, half-precision is not a native data type in MATLAB and hence, requires additional tweak to do the computations.

Best Answer

Related Solutions

MATLAB: Split a matrix into smaller sub matricies

MATLAB: Is half-precision slower than double-precision

Related Question