I'm getting OutOfMemoryErrors on some machines when calling GC.malloc (or new) for anything large (more than about 1GB), where core.stdc.malloc works fine.

Anyone ever come accross this before? I suspect it's related to some system memory monitor/management as it's only happening to me on supercomputer nodes.

On Tuesday, 24 June 2014 at 10:06:10 UTC, John Colvin wrote:
> I'm getting OutOfMemoryErrors on some machines when calling GC.malloc (or new) for anything large (more than about 1GB), where core.stdc.malloc works fine.
>
> Anyone ever come accross this before? I suspect it's related to some system memory monitor/management as it's only happening to me on supercomputer nodes.

P.S. yes I've checked the return value of core.stdc.malloc. Also, I'm not doing anything silly like calling them one after the other and then wondering why the second one doesn't work (although seeing as I've got ~24GB of RAM to play with that shouldn't be a problem).

John Colvin:

> I'm getting OutOfMemoryErrors on some machines when calling GC.malloc (or new) for anything large (more than about 1GB), where core.stdc.malloc works fine.

I think that when you go in territories where most people have not done much coding in D, like allocating than 1 GB in one shot, you hit some bugs. In general if some part of D is not used a lot, you have to assume it's buggy. Please report in Bugzilla the problems you have located.

Bye,
bearophile

On Tuesday, 24 June 2014 at 10:16:38 UTC, bearophile wrote:
> John Colvin:
>
>> I'm getting OutOfMemoryErrors on some machines when calling GC.malloc (or new) for anything large (more than about 1GB), where core.stdc.malloc works fine.
>
> I think that when you go in territories where most people have not done much coding in D, like allocating than 1 GB in one shot, you hit some bugs. In general if some part of D is not used a lot, you have to assume it's buggy. Please report in Bugzilla the problems you have located.
>
> Bye,
> bearophile

After a little more investigation It seems that I'm getting an OutOfMemoryError at around 1GB memory usage, regardless of the granularity of allocation.

On Wednesday, 25 June 2014 at 14:44:22 UTC, John Colvin wrote:
> On Tuesday, 24 June 2014 at 10:16:38 UTC, bearophile wrote:
>> John Colvin:
>>
>>> I'm getting OutOfMemoryErrors on some machines when calling GC.malloc (or new) for anything large (more than about 1GB), where core.stdc.malloc works fine.
>>
>> I think that when you go in territories where most people have not done much coding in D, like allocating than 1 GB in one shot, you hit some bugs. In general if some part of D is not used a lot, you have to assume it's buggy. Please report in Bugzilla the problems you have located.
>>
>> Bye,
>> bearophile
>
> After a little more investigation It seems that I'm getting an OutOfMemoryError at around 1GB memory usage, regardless of the granularity of allocation.

User error, carry on, nothing to see here...