[mira_talk] Re: Mira says "killed" as last word after being almost done
- From: Bastien Chevreux <bach@xxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Sun, 15 May 2011 22:00:26 +0200
On Sunday 15 May 2011 21:42:30 Adrian Pelin wrote:
> But if kpmf is set at 20 (and the default is 15), why does this still
> happen?
kpmf is a wish. Citing the docs: "it *tries* to keep x percent of the memory
free."
And because sometimes the estimates MIRA makes in earlier passes are slightly
off for later passes, this leads to increased memory needs where the memory
allocator is unable to keep memory needs within the allocated space and then
has to grow higher than "wished for".
Combine this with a bug of the 3.2.0 and 3.2.1 versions ... and memory
sometimes exploded. That bug was fixed in 3.2.1.9, leading to much less
situations with drastic memory increases.
> Indeed the swap space was only 1 GB last time, now I increased it
> to 49 GB and the RAM itself is 48 GB.
I'll never stop to wonder why people set up machines without adequate swap ...
1 GB swap for a 48 GB machine really is not enough :-)
> I am doing a hybrid assembly with
> high coverages so it is to be expected that the program requires lots of
> memory.
You did not write which version of MIRA you are using, but may I whole-
heartedly recommend to use one of the latest MIRA development versions? They
have some really intersting improvements especially for situations you
describe. Version 3.2.1.16 is fresh on SourceForge.
B.
Other related posts:
