[Mono-dev] SIGSEGV with large nursery size

[Burkhard Linke]

Hello Vlad,


On 11/02/2016 04:28 PM, Vlad Brezae wrote:
> Hello Burkhard,
>
>           This should be fixed on master by [1].
>
>            Is there any reason to use such big nursery sizes ? Is this a workaround for the stack overflow problem that you submitted earlier ?
It was one attempt to reduce the garbage collection pressure and memory 
fragmentation. The application is creating millions of small objects and 
discards them shortly afterwards. GDB stack traces contained 'LOS 
overflow' as reason for the garbage collection resulting in the stack 
overflow, so using a larger nursery might have prevented object 
promotion. It reduces the number of mmap'ed segments as a second benefit.
>
>            You could also consider using the minor=split params, in order to delay promotion to the major heap.
Thanks, I'll have a look into this.

Regards,
Burkhard