Can not generate omp parallel code for sparse tensors

[MaxwellF1]

I'm using the taco C++ API (built with OpenMP ON) to compute the contraction between two sparse tensors using this code:

Format csf({Sparse, Sparse, Sparse, Sparse});
Tensor<double> X = read("x.tns", csf);
Tensor<double> Y = read("y.tns", csf);
Tensor<double> Z({X.getDimension(0),X.getDimension(1),Y.getDimension(0),Y.getDimension(1)}, csf);

IndexVar i, j, k, l, m, n;
Z(i,j,m,n) = X(i,j,k,l) * Y(m,n,k,l); 

Z.compile();
Z.printComputeIR(std::cout);
Z.assemble();
Z.compute();

And I found the printed code is not omp parallel in the loop. However, I test the spmv computation and it is generated with omp parallel.

[rohany]

There won't be any parallel loops generated by default with this setup as there is no outer dense loop to parallelize over. Perhaps if you made a different format {Dense, Sparse, Sparse, Sparse} you would see parallel loops generated.