1998-11-03 - hardware sharing question

Header Data

From: Jyri Poldre <jp@pld.ttu.ee>
To: Cypherpunks <cypherpunks@toad.com>
Message Hash: 81ca57aa82df905be148425b0b39de4c3c2c039ccf72a60c443ad8e870cca434
Message ID: <Pine.SV4.3.96.981103173307.19127E-100000@pitsa.pldsise>
Reply To: <199811020422.XAA11460@cti06.citenet.net>
UTC Datetime: 1998-11-03 17:07:56 UTC
Raw Date: Wed, 4 Nov 1998 01:07:56 +0800

Raw message

From: Jyri Poldre <jp@pld.ttu.ee>
Date: Wed, 4 Nov 1998 01:07:56 +0800
To: Cypherpunks <cypherpunks@toad.com>
Subject: hardware sharing question
In-Reply-To: <199811020422.XAA11460@cti06.citenet.net>
Message-ID: <Pine.SV4.3.96.981103173307.19127E-100000@pitsa.pldsise>
MIME-Version: 1.0
Content-Type: text/plain




Dear C'punks,

There are many different hardware accelerator devices available for
speeding up the  modmul operation - The nFast and FastMap to name  a few.
I am about to build a new one. This unit should be: 

1. Expandable
    A. No harm if exponent is larger than N bits - just slower
    B. As much parralel calculations as possible - add ALU and 
       get faster results 
2. Low cost- keep only the heavy iron in the card - all other is ok 
   to do using host CPU.

This should also enable to build the device for several applications -
After all I want only to replace a call to some mathemathical subroutine.


To realize it the Residue Number System, with some additional memory seems
like a good choiche. The system will eventually consist of a card with
slots for addtional ALU and RAM modules. You can do away with only one ALU
- tha base kit, but it will consume bus bandwidth. Add memory and you are
free from that limitation. Slow? Add ALU units. 

But interestingly my main problem is not about cryptography :) I would
like to know, if there are possible other places, where integer
arithmetics could be used. Maybe A world would be a better place with fast
matrix multiplication?

So people. I would very much appreciate, if you could tell me about the
ways we could use this integer calculator for other applications as well.

I will be using RNS, what gives us a lot of multiplications/additions in
parralel. 

Thank you,

Jyri Poldre,
Tallinn Technical University.

 





Thread