Problems in DGEMV_TD

[joseemoreira]

Andy, you had the following definition:

const double epsilon = pow(1, -9);

This evaluates to 1 (1^(-9) = 1/(1^9) = 1) and is too big for epsilon. I changed the declaration to

const double epsilon = 1e-9;

Also, dscal cannot be used with negative increments. Therefore, I had to change

dscal(n, beta, y, incy);

to

dscal(n, beta, y, abs(incy));

I have uploaded those changes. Please review and let me know if you disagree.

[atcheng2]

Ah yes, thanks for catching these mistakes!