in reply to Re: Perl garbage collector and swap (5 vs 6)
in thread Perl garbage collector and swap

This might strike quite painfully in web server programming where you need to run mass quantities of web servers. So this seems to be the first real argument against Perl 6. I hope a better threading could be used that to reduce the impact of this problem.

By the way I guess it's a Parrot thing. So all other languages run on Parrot will suffer as well?

  • Comment on Re: Re: Perl garbage collector and swap (5 vs 6)

Replies are listed 'Best First'.
Re: Re: Re: Perl garbage collector and swap (5 vs 6)
by TimToady (Parson) on Mar 18, 2004 at 08:51 UTC
    Yes, it's a Parrot thing.

    It's possible that something like a generational GC can "cycle" slowly enough that the pages to be swapped in can be controlled and even anticipated. Also, a compacting GC will potentially run in a smaller footprint than Perl 5. (Though I think it was decided recently that a compacting GC was incompatible with threading in Parrot, if I recall.)

    But I agree with Abigail. If you're running a server, real memory almost always outperforms virtual memory. And with 64-bit computing coming online, there's really no limit to the amount of memory you can add now, apart from arbitrary limitations of particular hardware architectures that chintz out on giving you all your address lines...

[reply]
Re: Re: Re: Perl garbage collector and swap (5 vs 6)
by Elian (Parson) on Mar 18, 2004 at 16:24 UTC
    By the way I guess it's a Parrot thing. So all other languages run on Parrot will suffer as well?
    No, it's not a parrot thing, and tye's wrong on this. A sweep for dead objects only touches the object header arenas and those areas of the heap that belong to active objects and can contain pointers to other object headers. Those memory areas are generally segregated from the areas of the heap that contain non-pointer data such as strings.
[reply]
      I thought the discussion was about GC in general. Dead object destruction is only a part of that, surely...
[reply]

      So Java gets this wrong?

      And even based on just what you've said, it isn't so much that this isn't an added issue with Parrot vs. old Perl, but that you think the impact will be limited (not zero).

      - tye        

[reply]

In Section Seekers of Perl Wisdom