Re: RAM-CPU Singularity

Paul E. Black schrieb:
On Tue, 07 Feb 2006 05:07:35 -0800, yaoziyua wrote:

... I have this idea of RAM-CPU singularity:

That is to say, software applications should sometimes treat the RAM and
CPU resources as one single computing resource, using an idle resource to
compensate the urgent needs of another kind of resource.

Based on space-time tradeoff principles in algorithm design theory, we can
redesign object libraries like STL to be able to adapt to any extremely
unbalanced space-time requirements (e.g. a very fast CPU but very limited
RAM, or a very slow CPU but very rich RAM) and able to re-adapt to a new
requirement on-the-fly. Application frameworks like MFC, VCL or GTK should
also be redesigned to serve this purpose. But the details of resource
management should be transparent to the end developer.

An intriguing idea! I don't know if the trade-off could be made at a
fine grain, like page swapping, but it might work at a coarser grain.

Some computations "memoize" intermediate computations. If memory is
tight, it may be better to recompute more things and memoize fewer.

I think aspect-oriented techniques could be an elegant way to achieve this variability of caching. On slow systems, one could deploy an aspect that caches all (or some; perhaps depending on speed/RAM-ratio) return values of calculations.

Another point would be to offer more abstract algorithm and data structures: instead of letting the developer choose the concrete algorithm or structure, he should be able to give hints on memory-usage (size) and usage frequency of the data. The library could then choose an appropriate algorithm.

These trade-off could be made at initial run time. That is, the
program's start-up detects whether the execution environment is
memory-rich or computation-rich, then run the branch/version optimized
for that.

-paul-
In my scenario, the environment would choose an algorithm/data-structure factory and possibly deploy caching strategy aspects at startup.

Regards,
Karl
.

Relevant Pages

  • RAM-CPU Singularity
    ... RAM tomorrow, I have this idea of RAM-CPU singularity: ... and CPU resources as one single computing resource, ... resource to compensate the urgent needs of another kind of resource. ... between different kinds of computing resources, ...
    (comp.theory)
  • Re: [ckrm-tech] [RFC][PATCH] UBC: user resource beancounters
    ... kernel memory and virtual memory. ... Do you have any plans to post a CPU controller? ... the infrastructure (for resource management) we ...
    (Linux-Kernel)
  • Re: Garbage Collection and Library Deployment
    ... However, looking at Task Manager, the RAM is ... Just because you're done with the memory, that doesn't mean .NET releases the OS allocation it made in order to satisfy your need. ... Just because RAM is "a valuable resource for your low-end system", that doesn't mean your program should be explicitly trying to manage it. ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: [ckrm-tech] [RFC][PATCH] UBC: user resource beancounters
    ... kernel memory and virtual memory. ... fair CPU scheduler is not tied to UBC in any regard. ... the infrastructure (for resource management) we ... it is also more logical to have them separate since they ...
    (Linux-Kernel)
  • Re: Minutes from 5/19 CKRM/PAGG discussion
    ... >> One important input the PAGG team could give is some real ... > One example would be the implementation of CPU sets (or pools) a la ... it seems to me that even within the field of resource ...
    (Linux-Kernel)