mit-scheme-devel
[Top][All Lists]
Advanced

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

Re: [MIT-Scheme-devel] tag bits


From: Chris Hanson
Subject: Re: [MIT-Scheme-devel] tag bits
Date: Sat, 21 Aug 2010 02:18:41 -0700

It hardly seems worth the trouble, especially since the compiler won't
handle !HEAP_IN_LOW_MEMORY.  It's already the case that the limiting
factor on some systems is the inability to find a sufficiently large
chunk of memory in the low address space.  A similar change will be
needed to reclaim those low bits, so it might be worth putting in that
effort first.

Or you could wait until you get a 64-bit machine, surely not too far
in the future.

On Tue, Aug 10, 2010 at 3:01 PM, Taylor R Campbell <address@hidden> wrote:
> Currently, I believe the only objects whose addresses are not aligned
> on word boundaries are compiled entries.  If we forced them to be
> aligned, then the log_2 (bits per word) low-order bits of every object
> address would be zero, and we could shift the address by that many
> bits when storing it as an object's datum.
>
> This would let us quadruple the usable virtual address space on 32-bit
> systems, or octuple it if we got rid of the scode type tags and some
> unused type tags too (using records instead or something) and reduced
> the tags to five bits.
>
> Probably more trouble than it's worth, but it's a cute thought
> nevertheless which wouldn't require rewriting half of MIT Scheme.
>
> (I still occasionally think about this because while some of you lucky
> folks may be on super-snappy 64-bit machines, I'm still using a six-
> year-old 32-bit laptop.)
>
> _______________________________________________
> MIT-Scheme-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/mit-scheme-devel
>



reply via email to

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