OPM
- made NDEBUG on in release
Benchmark result overview:
| Test | Configuration | Relative |
|---|---|---|
| opm-git | OPM Benchmark: drogon - Threads: 1 | 1.01 |
| opm-git | OPM Benchmark: drogon - Threads: 8 | 1.19 |
| opm-git | OPM Benchmark: smeaheia - Threads: 1 | 1.007 |
| opm-git | OPM Benchmark: smeaheia - Threads: 8 | 1.152 |
| opm-git | OPM Benchmark: spe10_model_1 - Threads: 1 | 1.016 |
| opm-git | OPM Benchmark: spe10_model_1 - Threads: 8 | 1.018 |
| opm-git | OPM Benchmark: flow_mpi_extra - Threads: 1 | 1.015 |
| opm-git | OPM Benchmark: flow_mpi_extra - Threads: 8 | 1.035 |
| opm-git | OPM Benchmark: flow_mpi_norne - Threads: 1 | 1.022 |
| opm-git | OPM Benchmark: flow_mpi_norne - Threads: 8 | 1.048 |
- 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=1986
Benchmark result overview:
| Test | Configuration | Relative |
|---|---|---|
| opm-git | OPM Benchmark: drogon - Threads: 1 | 0.994 |
| opm-git | OPM Benchmark: drogon - Threads: 8 | 1.196 |
| opm-git | OPM Benchmark: smeaheia - Threads: 1 | 0.991 |
| opm-git | OPM Benchmark: smeaheia - Threads: 8 | 1.146 |
| opm-git | OPM Benchmark: spe10_model_1 - Threads: 1 | 1.008 |
| opm-git | OPM Benchmark: spe10_model_1 - Threads: 8 | 1.018 |
| opm-git | OPM Benchmark: flow_mpi_extra - Threads: 1 | 1.015 |
| opm-git | OPM Benchmark: flow_mpi_extra - Threads: 8 | 1.112 |
| opm-git | OPM Benchmark: flow_mpi_norne - Threads: 1 | 1.016 |
| opm-git | OPM Benchmark: flow_mpi_norne - Threads: 8 | 1.032 |
- 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=1986
This is really strange on spe9 the effect i > 25 % on update on my machine single threaded.
I think the time has come for this to be merged.
The only reason to NOT define NDEBUG for release builds is to preserve assert()s in the executable as a last-resort guard against producing wrong results. Such assert()s have now been removed from most parts of the code (although a few remain) and replaced with if (problem) and exceptions which is a better: assert() should never be used for checking things depending on user input, only to ensure the programmer did not do something stupid.
The remaining assert()s combined with not defining NDEBUG can result in users experiencing errors terminating their runs, when it in some cases at least could have been possible to run through after timestep cuts. That is another reason to make this change.