Re: MP problems?
Havard Eidnes <[email protected]> Sat, 18 Sep 2021 22:03:03 +0200 (CEST)
| Newsgroups | gmane.os.netbsd.ports.macppc,gmane.os.netbsd.ports.powerpc |
|---|---|
| Message-ID | <[email protected]> |
Hi, I have some good news on this: > Is it starting to swap? I ran into issues like that with too many > cc1plus instances ( as in, building clang with less than 1GB per > compiler instance ), and rustc is an infamous memory pig. First this: nope. 10M swap is used, and doesn't vary. rustc mostly appears to be in the 250M - 400M virtual size range which in this day and age isn't excessive. Well, I should not have spoken so soon -- it looks like one of them grows to 8-900M, and on 32-bit ARM I've seen one of the processes expand up to nearly 1.7G, challenging the max user-land VM space on that port. However, what I neglected to tell (and remember) was that I still had 9.1 user-land installed, after the earlier instaibility I experienced when upgrading the kernel. I've now installed the matching 9.99.88 user-land, and the wip/rust build is now progressing as normal with MAKE_JOBS set to 2, and has arrived at the LLVM build stage. Regards, - HÃ¥vard