RFR: Enable TLHS by default (again)

[Roman Kennke]

> On 11/14/2018 04:56 PM, Roman Kennke wrote:
>> Do we want to finally enable TLHS for good? Or do we considere too risky
>> compared to (almost zero immediate) usefulness? It'd allow some more
>> concurrent processing going forward.
> 
> I think it is too risky without apparent benefit, at this point.

Ok, right. Fair enough.

> Adding the simple regression test
> with TLHS specifically enabled would be nice, though.

The problem with that is that TLHS did work for most code from day 1,
the only program that I remember being problematic is serial benchmark,
and then only with traversal mode. That hardly qualifies as 'simple
regression test' that we could include in our jtreg tests. We could,
however, add something to CI that runs this specific setup nightly? In
the future, should we have have an actual use-case for TLHS, we could
look at those CI jobs and have some confidence that it's now safe to
enable TLHS?

WDYT?

Roman