Bouncner

> @Bouncner - Happy to incorporate any changes to make this better/easier. > Saw this tweet - https://twitter.com/hyrise_db/status/1410207307024187405?s=20 Sorry, that took a while. :(

> While obviously not elegant, the resulting behavior is not far from what happened anyway: only the first operator of the subquery is really scheduled and simple operator chains are...

(leaving the approval even though benchmarks are still missing)

Might it make sense to split the two topics of this PR? I was setting up the DELab experiment but wasn't actually quite sure what I would test here.

> We could, but then we have to first do the `benchmark_all.sh` part. Why? However, okay from my side. Do you open the PR? I'll start the benchmarks today.

> > Why? > > Because we need the MT benchmarks to see the effect of the changed subquery scheduling. > > > Do you open the PR? I'll start...

## Intel (nemea) **System** nemea - click to expand | property | value | | -- | -- | | Hostname | nemea | | CPU | Intel(R) Xeon(R) Platinum...

Hmm, I don't know if we can trust these nemea numbers. On AMD, everything is looking fine. ### AMD **System** cx21 - click to expand | property | value |...

So both nemea runs show a clear performance regression on TPC-DS ST. Any ideas?

### DELab - Intel **System** cx01 - click to expand | property | value | | -- | -- | | Hostname | cx01 | | CPU | Intel(R) Xeon(R)...