Before checking the performance regression, we need to check that with the current flags the benchmarks still verify their output and we are not leading to compilation errors (at least as far as the NPB are concerned).
This is simply done by greping for "UNSUCCESSFUL" in the output during the runner pipeline.
We finally expand the runner so that it dumps the output error in case a command fails.
Before checking the performance regression, we need to check that with the current flags the benchmarks still verify their output and we are not leading to compilation errors (at least as far as the NPB are concerned).
This is simply done by greping for "UNSUCCESSFUL" in the output during the runner pipeline.
We finally expand the runner so that it dumps the output error in case a command fails.