[extropy-chat] Optimal computer configurations

Samantha Atkins sjatkins at mac.com
Thu May 4 17:54:49 UTC 2006


I have written storage management and caching systems.  There are a  
small set of truly different types of storage needs across all common  
types of software applications when you factor the space.  It is  
possible to find a common supporting general memory/cache model with  
two primary subtypes of storage and a small set of storage/cache  
management primitives that can be mixed and matched to cover  
different needs.  The storage subsystem can be designed to be  
adaptive to actual usage patterns.   In this way different memory  
needs can be met at a much higher and more flexible level and with  
far greater performance and efficiency than attempting to satisfy all  
these types of needs using clib and equivalents.

- samantha

On May 3, 2006, at 2:47 PM, Robert Bradbury wrote:

>
> On 5/3/06, Samantha Atkins <sjatkins at mac.com> wrote:
> Allocate large blocks using the underlying system/libc calls.   
> Write alloc, free, etc. substitutes and any more specialized calls  
> you might desire  that
> suballocate from these large blocks with sufficient additional  
> bookkeeping, tagging and such to more efficiently clean up garbage  
> and compact the heap.   This sort of thing has been written many  
> times.  Its main drawback is the larger memory increment when the  
> current large blocks cannot satisfy a request.   But with  
> compacting of the heap large blocks themselves may be released if  
> memory needs go down significantly enough.
>
> I understand that Samantha.
>
> It is a somewhat non-trivial problem because each type of memory  
> (say history records vs. large images vs. garbage collected memory  
> [Javascript???]) requires different memory management strategies.   
> There seems to be 32+ types of *alloc() calls (png_malloc(),  
> GC_malloc(), JS_malloc(), etc.) in the source for Mozilla. I'm not  
> sure what subset is used by Firefox.  What I am seeing suggests  
> that they knew this would be useful but may not have yet taken  
> steps to optimize it.
>
> Robert
>
>
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20060504/9db1992e/attachment.html>


More information about the extropy-chat mailing list