I implement an integer(>0) non square matrix multiplication for two dense matrix,A[2000][100],B[100][100], and I use the submatrix method with use of local memory, then I set the work group size to (8,8), and then I got a wrong result, only first row is correct, and other row is all wrong, with lots of zeros and even negetive integer. I think non square matrix should do nothing to the result, but why I got wrong answer?