I am trying to use Moose to do run graph algorithms on data with about
300,000 nodes. This data is map data downloaded from OpenStreetMap,
and I am trying to analyze it with the Moose-Analysis-Graphs package.
This is cool :)
Once you fix the problem
--memory 1024m
do not forget to tell us about your experience.
I do not remember having a so large graph in Moose (we have model with
more than 600000 entities but we do not feed them to the graph algo
directly).
However, it seems completely unable to handle this, partially due to
unoptimized algorithms, but mostly due to memory limits of the VM or
image itself.
I am using the vm and image of the moose one-click image for mac:
http://www.moosetechnology.org/#install . I believe this is a pharo vm
and image, I don’t know if pharo and squeak are still different these
days, or if it matters. Been out of squeak for a few years
Welcome back. Pharo is moving fast :). We hope to have the Spur release
soon.
This platform is seemingly unable to use more than about 500Mb of
memory. I watch it in mac’s activity monitor. Once it reaches about
450Mb as measured by activity monitor, it is spending nearly all it’s
time in the garbage collector, (as measured by activity monitor’s
Sample Process tool). The most I’ve seen it eventually reach is 556Mb,
after about an hour, then it does a hard VM crash (Pharo has quit
unexpectedly).
Is there a way to remedy this? I tried running the image in 4
alternative VM’s:
- The latest pharo mac vm from http://files.pharo.org/vm/pharo/mac/ .
Same issue as above. likely the same vm
- Cocoa Cog vm from http://squeakvm.org/mac/ : crashed at startup.
could not load image
- Cocoa Stack vm from http://squeakvm.org/mac/ : crashed at startup;
could not load image
- Carbon Stack vm from http://squeakvm.org/mac/ : refused to start
Is it possible to allocate more than 500Mb of memory to squeak?
