problem with shared heap?

david wallin <>
Thu Mar 6 11:47:04 CET 2003


On Thu, 2003-03-06 at 09:14, Bjorn Gustavsson wrote:
> Shared heap is still experimental. The garbage collector is
> in particular needs more attention (it is basically the same garbage
> collector as in the separate-heap emulator).
That makes sense since my code probably creates a lot of 'stuff' that
never will be used again.

> Shared heap can be better if you have an application with a lot
> of concurrency and message passing. Message passing will be faster
> (because messages are not copied) and the emulator will use less memory
> (because processes can share terms instead of having their own separate
> copies).

Yes, but when shared heap is finished, will there be cases when running
a shared heap will actually perform worse than without it, or is it just
now in its experimental stages that these things can occur?


> /Bjorn
> david wallin <> writes:
> > When trying the new R9B-1 I decided to compile with --enable-shared-heap
> > to see how much faster my code would run, to my surprise the result was
> > quite alot slower.
> > 
> > My system is a Pentium4 running Linux 2.4.18 and R9B-1 with HiPE enabled
> > (actually: not disabled) in both cases. 
> > 
> > Is this a known problem (and why shared heap is still experimental), or
> > is there any rule of thumb when to use or not to use shared heap ?
> > 
> [...]
david wallin <>

More information about the erlang-questions mailing list