Memory spikes

[outpaddling]

Monitoring daligner with some of our sample data under "top" shows that it uses about 3 gigs of RAM per process about 99% of the time and occasionally spikes to 10 gigs.

I'm wondering if anyone knows what causes the spike and whether it could be eliminated. This would help on systems with a high core/memory ratio, where there isn't enough RAM to support 10G for each 4-thread process. We have many such systems in our HTCondor grid that we might want to use for alignment.

Given that the spikes are so brief, it's unlikely that trading some performance for a less memory-intensive approach in that code segment would have much impact on overall performance.

Thanks,

Jason