Add instructions for swapfile, and let Node use it

[alexrudd2]

Closes #118

[alexrudd2]

@osresearch

[alexrudd2]

Damn it Windows. Probably worth using cross-env since I imagine the other envvars are broken as well?

[alexrudd2]

@lewi0622 I noticed on your blog (cool!) that you were working around Out-of-Memory issues. Can you try this and see if it helps?

[lewi0622]

@lewi0622 I noticed on your blog (cool!) that you were working around Out-of-Memory issues. Can you try this and see if it helps?

I have a 5.1 megabyte file that reliably makes Saxi run out of memory when I plot all three layers. It handles plotting two layers fine.

I guess I'm a little confused exactly what to do. So my RPi3 already has a swap file created (not my explicit doing). Do I need to make another one? And do I need to do anything to allow Node to use it?

The below memory stats are when Saxi isn't running (when it is running, but not plotting, the used Mem goes up to 75M):
lewi0622@axipi:~ $ free -h --si
total used free shared buff/cache available
Mem: 921M 54M 521M 0.0K 345M 802M
Swap: 99M 15M 84M

[alexrudd2]

My (imperfect) understanding is there are/were two problems with memory. First, RPi Zeros only have 512MB RAM and can benefit from a swapfile. Since your RPi3 (1) has more memory and (2) already have a swapfile, this is of no use.

Second, this PR also uses a flag to give Node.js more memory than the default 512MB. See https://stackoverflow.com/a/62650303. Namely, --max-old-space-size=1024 (1GB). Can you try running saxi with that flag and see if that changes anything?

[lewi0622]

I did try with --max-old-space-size=1024 as well as values with 2048, and 3072 (4096 is too much for my RPi3, and fails the build step). I don't know how to tell if it's even trying to use the swap file.

With the attached file, even just running on the main branch, plotting two of the three layers is fine (approx. 14 hrs of plot time), but doing all three layers is too much (approx. 22 hrs of plot time) even with dedicating 3072 (3GB). Each layer of the file is of similar size/complexity.

So it's hard to say for sure, but increasing the --max-old-space-size doesn't seem to be helping in this case?

[alexrudd2]

Damn. Well, it was worth a shot. :)

In the coming weeks I'm bringing in a colleague to hack on this a bit more. Perhaps we can try your test file.