Add triangular solvers to benchmark

[upsj]

For @lksriemer's experiments, this PR adds the triangular solvers to our benchmarks.

Usage:

echo '[{"filename":"upper-triangular.mtx", "optimal": {"spmv":"csr"}}]' > input.json
echo '[{"filename":"upper-triangular.mtx", "rhs": "rhs.mtx", "optimal": {"spmv":"csr"}}]' > input2.json
benchmark/solver/solver -solvers upper_trs -executor cuda < input.json
benchmark/solver/solver -solvers upper_trs -executor cuda < input2.json

Output:

[
    {
        "filename": "upper-triangular.mtx",
        "optimal": {
            "spmv": "csr"
        },
        "solver": {
            "upper_trs": {
                "recurrent_residuals": [],
                "true_residuals": [],
                "iteration_timestamps": [],
                "rhs_norm": 6.0,
                "generate": {
                    "components": {
                        "upper_trs::generate#3": 100,
                        "upper_trs::init_struct#1": 700
                    },
                    "time": 700.0
                },
                "apply": {
                    "components": {
                        "upper_trs::should_perform_transpose#1": 200,
                        "upper_trs::solve#6": 400
                    },
                    "iterations": 0,
                    "time": 700.0
                },
                "residual_norm": 7.12901703041511e-16,
                "completed": true
            }
        }
    }
]

By default, the right-hand side consists of all ones or the specified vector in the rhs json parameter, it can be set to random values using the -random_rhs flag

The specified runtimes are in nanoseconds.

If you want to extract the lower triangular factor of a MatrixMarket file, you can use

awk '{ if ($1 >= $2 || $1 ~ /^%/) print }' input.mtx > lower_triangular.mtx
awk '{ if ($1 <= $2 || $1 ~ /^%/) print }' input.mtx > upper_triangular.mtx

you only need to modify the number of non-zeros in the header line afterwards

Finally, you can specify different strategies by using different csr formats in optimal, i.e. csrc, csrm, csri.

[tcojean]

LGTM

[lksriemer]

Just wanted to say thanks a lot! That'll be a great help to me.

[codecov]

Codecov Report

Merging #664 (f25a8d7) into develop (471108e) will decrease coverage by 0.02%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           develop     #664      +/-   ##
===========================================
- Coverage    92.87%   92.84%   -0.03%     
===========================================
  Files          325      325              
  Lines        23296    23296              
===========================================
- Hits         21635    21630       -5     
- Misses        1661     1666       +5     

Impacted Files	Coverage Δ
omp/reorder/rcm_kernels.cpp	94.44% <0.00%> (-3.09%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 471108e...f25a8d7. Read the comment docs.

[pratikvn]

LGTM!

[pratikvn]

Just a question, we dont use a template type here ? Maybe it is better to add ValueType to all template parameters ?

[upsj]

That's a valid point! However, I believe we should address this in another PR

[thoasm]

LGTM!

[Slaedr]

This would be useful. I just had a small nitpick about the documentation.

[fritzgoebel]

LGTM

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities (and 0 Security Hotspots to review)
0 Code Smells

0.0% Coverage
0.0% Duplication

[hartwiganzt]

Not sure whether this is necessary, but should we mention that lowe/upper_trs expect a triangular matrix?

[upsj]

Too late sorry, the PR is already merged 🤷‍♂️

[Slaedr]

Essentially, I guess we should define what are 'lower_trs' and 'upper_trs', in case people are not familiar. From an applications point of view, or indeed from most common points of view, these two are really not in the same category as bicgstab, gmres, etc. We should address this in some other PR.

[tcojean]

I don't think we have to explain everything about the TRS here necessarily, because this here is only the benchmarks description. If you want information from a user POV, the class description etc should give you all you need. On the other hand, someone who reads this should know what to benchmark, and only wants to know how.