Re: AES and dynamic table generation
From: Brian Gladman (brg_at_nowhere.at.all)
Date: 01/12/04
- Next message: Simon Johnson: "Re: New formula unlocks base 10 encryption????"
- Previous message: TomJefries: "Re: Visual Cryptanalysis paper"
- In reply to: Tom St Denis: "Re: AES and dynamic table generation"
- Next in thread: Simon G Best: "Re: AES and dynamic table generation"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Mon, 12 Jan 2004 14:45:59 -0000
"Tom St Denis" <tomstdenis@iahu.ca> wrote in message
news:0fxMb.110570$AAe1.47673@news01.bloor.is.net.cable.rogers.com...
>
> "Simon G Best" <s.g.best@btopenworld.com> wrote in message
> news:40029237.800@btopenworld.com...
> > Tom St Denis wrote:
> > > "Brian Gladman" <brg@nowhere.at.all> wrote in message
> > > news:nFWKb.19364$qx2.2163603@stones.force9.net...
> > >
> > >>Its very easy to make dynamic tables thread safe and they do have
> > >>significant advantages in some environments.
> > >
> > > In a portable fashion?
> >
> > The OP didn't limit the context of the question he asked to just
> > multithreaded, portable applications.
> >
> > Consider a microcontroller with limited nonvolatile memory, but plenty
> > of RAM. If that's going to be used in an application where the speed of
> > lookup tables is required, dynamic table generation might be the only
> > option!
>
> A microcontroller with little ROM and plenty of RAM? Where do these come
> from? 8051 has 256 bytes of ram [that you'd want to use] and upto 64KB of
> ROM... The 6805 series is similar. RAM is hella more expensive in
fielded
> designs then ROM is.
Such situations do exist. I have even been asked to support situations
where the only memory available is stack based RAM.
> The only real "benefit" of dynamic tables is I can now send you a shorter
> executable. But since the program only has to be stored once [this is AES
> afterall] there's no point in wasting the 10KB [or so] of RAM on tables
that
> could just as easily fit in ROM.
This is not the only benefit since in some situations ROM is much slower
than RAM. Hence it can sometimes pay in speed terms to have the tables in
dynamic memory.
Designing code that can be used in both ordinary and in embedded
environments in harder than you might expect since the latter are much less
predictable in terms of the constraints that they impose.
And, of course, inability to cope with needs that are frequently encountered
in embedded systems is a real constraint on portability :-)
Brian Gladman
- Next message: Simon Johnson: "Re: New formula unlocks base 10 encryption????"
- Previous message: TomJefries: "Re: Visual Cryptanalysis paper"
- In reply to: Tom St Denis: "Re: AES and dynamic table generation"
- Next in thread: Simon G Best: "Re: AES and dynamic table generation"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|