Adding Additional Internal Timers for SLATE Routines

[Treece-Burgess]

This PR will address adding additional internal timers for the following SLATE routines:

• gbsv
• hesv
• gesv_mixed
• gesv_mixed_gmres
• posv_mixed
• posv_mixed_gmres
• gels_cholqr
• hegv

See issue #128.

[Treece-Burgess]

@mgates3 I have started adding additional internal timers for the routines outlined in issue #128. I have currently pushed up four, but only three I believe are ready for review which are: hegv, gbsv, and hesv. I am unsure how you want to name a few of the timers within gesv_mixed, I will make comments at certain lines of code to distinguish where I would like input from you if possible.

[Treece-Burgess]

At line 242 within gesv_mixed.cc, the following code exists:

if (! converged) {
        if (info == 0) {
            // If we performed iter = itermax iterations and never satisfied
            // the stopping criterion, set up the iter flag accordingly.
            iter = -itermax - 1;
        }
        // Fall back to double precision factor and solve.
        // Compute the LU factorization of A.
        info = getrf( A, pivots, opts );
        // Solve the system A * X = B.
        if (info == 0) {
            slate::copy( B, X, opts );
            getrs( A, pivots, X, opts );
        }
    }
    if (target == Target::Devices) {
        // clear instead of release due to previous hold
        A.clearWorkspace();
        B.clearWorkspace();
        X.clearWorkspace();
    }
    return info;
}

We run into calls again for getrf and getrs. Since for example there is already a timer at line 175 for getrf are we wanting to name these to be say getrf_A and getrs_B?

[mgates3]

Great first step. See changes in posv_mixed_gmres. Similar changes need to apply to posv_mixed, gesv_mixed, gesv_mixed_gmres.

This was my first pass. I might have other suggestions later on.

[mgates3]

Few more changes. Make sure same changes are applied to both gesv and posv. Sometimes I use meld to compare files; may or may not be useful.

    meld src/posv_mixed.cc src/gesv_mixed.cc

[Treece-Burgess]

I will resolve the branch conflicts tomorrow, but I just want to comment that I do indeed see them.

As well, as do another pass to make sure everything is correctly in order.