Noah Evans

I tested the benchmark with nemesis and distrib. I got: nemesis real 2m13.800s user 140m33.444s sys 1m52.144s distrib real 2m17.940s user 144m54.988s sys 1m53.671s I started looking that the profiling...

Not yet. What's your exactly configure invocation? (or where can I find it in the chapel source?)

@ronawho Using your environment variable settings (and 1 shepherd to make sure numa isn't screwing anything up) I get the following on a 2 socket 32 core haswell: sherwood real...

Got it to the right timing. Perf was interfering. With distrib I'm getting 15.56s user 0.37s system 99% cpu 15.952 total

Here's a diff of nemesis and distrib running thread ring on 32 cores: > [nevans@shepard-lsm1 benchmarks]$ perf diff perf.data.nemesis perf.data.distrib > # Event cycles > # > # Baseline Delta...

Let me try this in the chapel thread ring proper rather than qthreads benchmark and get back to you.

@ronawho Exactly. I'm just getting a representative environment up. Looking at the performance results in #39 is there a particular benchmark I can spin on optimizing that models a particular...

Gotcha, if I can help with anything here, feel free to ping me.

@ronawho Have you tried out the binders stuff yet?

@eschnett was this answer sufficient?