swarm-support
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Seg fault after 2 days of simulation


From: Jan Kreft
Subject: Re: Seg fault after 2 days of simulation
Date: Fri, 15 Oct 1999 14:53:13 +0100 (BST)

Tried to send this off yesterday, but didn't get through somehow. Second
try now. Apologies if you got this already.

Dear Marcus,

On 13 Oct 1999, Marcus G. Daniels wrote:

> >>>>> "JK" == Jan Kreft <address@hidden> writes:
> 
> JK> BTW, would be better as an array of BOOL instead of char, does
> JK> anyone know if that's possible?
> 
> You'd have to use bit shift operations and masking to do that, e.g.
> "value |= (1 << offset)", "value &= ~(1 << offset)".  It's usually not
> worth it unless the sequences are really big.  If the sequences have
> *known* top-end size, you can allocate an ivar of sequence-size / 8 or
> use the ": size" bitfield syntax. 
> 
> Btw, I really want to get away from use of raw memory in Swarm.  These
> nasty consequences to unexpected allocation requirements are way too
> nasty.  Even with a lot of experience and good tools, debugging memory
> corruption problems can be a big time sink that ends up teaching you
> nothing (except the benefits of obsessive behavior).  This bug
> wouldn't have occurred if you were using the Java layer, because you'd
> have used automatically sized and garbage collected Strings.

I would love such automatically sized strings. But, and this is a long
but, my sims already run for about 2 days, and I plan to go to 3D next
year, probably slowing them down to a standstill. Or possibly include
hydrodynamics. In other words, soon I'll have to use some sort of
supercomputer to run my models, and that will mean going to pure C rather
than Java from Objective C.  Preemptively, I have made this years
additions to BacSim as close to C as I could within the limits imposed by
the existing code. Not that I'm proud of it, but I see no other way. I
looked around for something like Java for Cray, but all I found was very
experimental support for SMP in Java, and certainly no Java for Cray. HP
Fortran would be even better, I suppose, but I've never used Fortran, so
that's above my league. Given the current focus of Swarm development, I
doubt Swarm on Cray will be ready when I would need it.

Cheers, Jan.





                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]