I just noticed that the time for one sweep is 255 ms in Table 3 but more than 300 ms in Figure 13. Any thoughts about why the difference?

hmm, so I have a comment in the tex file too about this. that time per sweep is calculated by multiplying per update time with 10000. Because the smaller ones are too slow and we were running out of time. Because of this we are not doing updating of array after every update, which might be the reason.

So for the entries in Table 3 that are less than 1000 ms, I think it’s worth re-running to get matching numbers. That should only take a few seconds, right?

Also we should look into the Haskell-Maple connection problem. Debugging with strace etc might be in order.

For the Haskell-Maple thing, what about we get a very clear reproducible case? Things were too hectic, and I never quite got what the symptoms were.

The symptom was just that hk-maple hangs (after hk-maple —debug prints “sending to Maple…”).

It’s reproducible on some machines, it seems, but not others. Maybe it’s about Canadian citizenship?