Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Re: storing n-ary data in complex numbers...

37 views
Skip to first unread message

Chris M. Thomasson

unread,
Oct 28, 2016, 12:24:27 AM10/28/16
to
I forgot to post this to the other groups wrt comp.arch!

Sorry about that non-sense!

;^o

On 10/27/2016 9:13 PM, Chris M. Thomasson wrote:
> FWIW, I have created a little storage mechanism that can store n-ary
> data in complex numbers via the iterates of a formula:
>
> z^b+c
>
> This z is a complex number to the power of the base we are working with,
> hence the n-aryness. The plus c part is the secret key.
>
> The key is mapping n-ary data to the power b in the formula above. I
> have a program that does this.
>
>
> Before I dive into this detail too far, here is some C++ code that
> stores 256 bits in a single 128 bit complex number comprised of two
> 64-bit double parts for real and imaginary components. Well, double is
> 64-bit on my system at least!
>
> Some possible compression for you to possibly take a look at?... ;^)
>
> Here is the code on github:
>
> https://github.com/ChrisMThomasson/fractal_cipher/blob/master/RIFC/cpp/ct_rifc_sample.cpp
>
>
>
> And a link to an online C++ compiler version of the code:
>
> http://cpp.sh/9ihc
>
> Take a look at this content for some further information:
>
> https://groups.google.com/forum/#!original/sci.math/PUkOJVSDzN4/c7zeZDUdAwAJ
>
>
> Can you compile the C++11 code?
>
> Sorry for the brief information, I promise to add more context.
>
> My main concern is that compressing this data is very delicate, and I
> would like to make it more general. Try playing around with storing 32
> symbols or so, you can get to a place where floating rounding errors
> will rear up and bite the living shi% out of everything, such that we
> get a data corrupted error., the load FAILED!... ;^o
>
> Take a look at the C++ code... Try to run it online from:
>
> http://cpp.sh/9ihc
>
> Then run it on your system in C++11... Well, do you get coherent data,
> or corrupted?
>
> I am worried about running this floating point intensive code on
> different arch's... Coherency is key!
>
> I hope this is somewhat on topic!?
>
> ;^/
>

0 new messages