OPM
testing removing fucntion invertMatrix
If I recall correctly, we have this here to get better performance. In Dune there was a specialization missing and since DenseMatrix was introduced the optimized invert method was not picked up. I can be missing something.
We definitely will need to benchmark the changes here to be sure that the prerequisites do not hold anymore.
Did not go through the jenkins failures in detail. But it looks like only small fluctuations. Very likely that we can go ahead to remove the MatrixBlock unless more information shows the opposite.
I do not know where to find the benchmark results anymore.
I have notified Michael about the benchmark problems.
In addition we need to do some manual benchmarking with more than 3 phases, too.
In addition we need to do some manual benchmarking with more than 3 phases, too.
For the blackoil cases, StandardWell has 4 primary variables, so the 4X4 inversion might be used. But I do not think it will affect the performance in a noticeable way.
For other cases, maybe the cases with 4 primary variables, like blackoil + polymer might use it, we can check it. Other than that, there is not much 4X4 matrices involved I believe.
Benchmark result overview:
| Test | Configuration | Relative |
|---|---|---|
| opm-git | OPM Benchmark: drogon - Threads: 1 | 1 |
| opm-git | OPM Benchmark: drogon - Threads: 8 | 0.978 |
| opm-git | OPM Benchmark: punqs3 - Threads: 1 | 1.013 |
| opm-git | OPM Benchmark: punqs3 - Threads: 8 | 1.003 |
| opm-git | OPM Benchmark: smeaheia - Threads: 1 | 1.018 |
| opm-git | OPM Benchmark: smeaheia - Threads: 8 | 1.007 |
| opm-git | OPM Benchmark: spe10_model_1 - Threads: 1 | 1.001 |
| opm-git | OPM Benchmark: spe10_model_1 - Threads: 8 | 1 |
| opm-git | OPM Benchmark: flow_mpi_extra - Threads: 1 | 1.009 |
| opm-git | OPM Benchmark: flow_mpi_extra - Threads: 8 | 1.006 |
| opm-git | OPM Benchmark: flow_mpi_norne - Threads: 1 | 0.985 |
| opm-git | OPM Benchmark: flow_mpi_norne - Threads: 8 | 0.986 |
| opm-git | OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 1 | 1.008 |
| opm-git | OPM Benchmark: flow_mpi_norne_4c_msw - Threads: 8 | 0.995 |
- Speed-up = Total time master / Total time pull request. Above 1.0 is an improvement. *
View result details @ https://www.ytelses.com/opm/?page=result&id=2515
