bts.gr  -- Solves Ax = b for dense lower triangular matrix A.

The solution x over-writes b.  A is divided into N x N blocks, and b
is divided in N blocks.  Let A<i><j> refer to block i, j of A, and let
b<i> refer to block i (vector) of b. 

The algorithm is.

for j = 0 .. N-1 {
   solve(A<j><j>, b<j>);  -- solves for x<j> using sequential alg.
   parallel for i = j+1 .. N-1 {
     b<i> = b<i> - A<i><j> * b<j>
   }
}

The program generates its own dummy input values that yield a solution
vector that contains all 1.0s.

The problem size may be changed in GetSys's node subroutine.

N - number of blocks.
blk - size of a block.

Thus matrix A is (N*blk) X (N*blk).

