Benchmark 85 has a special BC of normal traction, and Benchmark 164 is about random walk.
On Jenkins, loading numdiff downgrades the gcc from 6.2 to 4.8. For the two failed benchmarks, the codes compiled with gcc 4.8 and 6.2, respectively, give results with a slight difference. The reference results were generated by using the code compiled with gcc 6.2, while Jenkin uses gcc 4.8 after running cli.sh. This makes the two benchmarks failed.
The following figures shows the displacement magnitude by the codes compiled with gcc 4.8, and 6.2, respectively:
Awesome Wenqing, thanks for figuring this out! The build is marked as unstable because there are new compiler warnings because of the newer compiler. I will take of this next week. Thanks a lot!
This PR upgrades numdiff and uses gcc 6.2 on jenkins.
There are two benchmarks are marked as failed in the recent PRs
Benchmark 85 has a special BC of normal traction, and Benchmark 164 is about random walk.
On Jenkins, loading numdiff downgrades the gcc from 6.2 to 4.8. For the two failed benchmarks, the codes compiled with gcc 4.8 and 6.2, respectively, give results with a slight difference. The reference results were generated by using the code compiled with gcc 6.2, while Jenkin uses gcc 4.8 after running cli.sh. This makes the two benchmarks failed.
The following figures shows the displacement magnitude by the codes compiled with gcc 4.8, and 6.2, respectively:![comp1](https://user-images.githubusercontent.com/1343839/51400262-4920ea80-1b48-11e9-945c-ea68eeeb2141.png)