Elliot Ronaghan

Those times look off to me, especially for nemesis where I would expect the real and user times to be pretty similar since the blocked qthreads shouldn't be using up...

@npe9 https://github.com/chapel-lang/chapel/blob/master/third-party/qthread/Makefile has everything that we set to build qthreads. I think Dylan and/or George also added something to https://github.com/Qthreads/qthreads/blob/master/scripts/build.pl to mimic our config. And by default we run w/o...

And what if you use `` shepherds? For us, nemesis still gets good performance and the real/user times are pretty close, but for distrib we get worse perf and `time...

Ok, we don't have many benchmarks that use a lot of highly contended sync vars, so thread-ring performance isn't really indicative of performance for many other benchmarks. The performance of...

Ah ok. I think it's going to be tricky to compare nemesis and distrib until after #38 is resolved. Prior to using a hybrid spin/condwait approach for nemesis, distrib had...

This is still of mild interest to us, just because it's a somewhat interesting proxy for sync var performance, but it's low priority.

Yeah, that's a good point. We do want to look at using binders, although I haven't had a chance to yet. FWIW using binders with QT_CPUBIND unset has roughly the...

Was looking into this a little bit and it appears that almost all of the startup time is spent in `hwloc_topology_load` I found a couple other threads from HPX-5 and...

FYI hwloc 1.11.6 had some nice speed ups for topology discovery: - "Make the Linux discovery about twice faster, especially on the CPU side, by trying to avoid sysfs file...

This is still of interest to us, though I don't know how much we can do about it. I think the next things for me to try are upgrading to...