[PATCH] D39506: Add a script to run various benchmarks and send the result to lnt

[Rafael Avila de Espindola via llvm-commits]

James Henderson via Phabricator <reviews at reviews.llvm.org> writes:

> jhenderson added a comment.
>
> I just tried running this locally without using cset, and I got an error from the '--' part of the command, which presumably is intended for use with cset. Removing it would cause cset to fail, if used as you described, but it looks like it is possible to do "--wrapper sudo,cset,shield,-e,--" and still get the expected result, even though the -e is in the seemingly wrong place. Note: I have not verified that in this case the run is definitely shielded, although it looks likely.

Thanks. I have fixed that in the last version. I also noticed that
.split() removes empty results, but .split(',') does not, so we now
handle that explicitly.

> One feature request, that doesn't need doing in the first iteration, is for the ability to print the summarised results to stdout optionally instead of posting to LNT. At the moment, this script requires LNT to run, which for a casual user who just wants to quickly compare two sets of times may not be the best situation.

For the first version I would like to focus on the bare minimum to get a
bot started. But I agree that would be a nice feature for a followup
patch.

Cheers,
Rafael