FPGA re-use options

[iambaboon]

I know it's still some way until FPGAs become obsolete, but we have to plan in advance for this.

This https://bitcointalk.org/index.php?topic=136743.0 talks about something like SETI@home.

What I would like here is to centralize ideas on how FPGAs might be reused in the future. Basically, which of these would be most suited for FPGAs.
http://en.wikipedia.org/wiki/List_of_distributed_computing_projects. Or other ideas on re-using them @home.

Normally they should be re-programmable to do anything, isn't it ? What do you guys think ?

I can maintain a list below, that can go later in the bitcoin wiki.
 * alternative coin/blockchain with different algorithm

[rupy]

I wish to build a new alt-currency with a firmware for spartan-6 that contains a new hashing algo (that should be evolvable thus eliminating the possibility to ASIC it). And build a really lightweight client with merkle root thin chain in java.

[nathanrees19]

I wish to build a new alt-currency with a firmware for spartan-6 that contains a new hashing algo (that should be evolvable thus eliminating the possibility to ASIC it). And build a really lightweight client with merkle root thin chain in java.

Technically this just increases the complexity of a possible ASIC.

[rupy]

Nope, not if the firmware is loaded by the miner and it can change algorithm on the fly. For example from SHA-512 to SHA-3, you cant make a ASIC for something you dont know what it is, but you can progam a FPGA to anything, as long as it fits the FPGA you are using...

[nelisky]

So, I don't really want to get into an 'alt-chain' discussion, but what if multiple hashing algos were available, each having a distinct difficulty attached to it, all usable on the same blockchain? This way ASICS would bring sha256(sha256()) difficulty up a lot, but GPU users could still use scrypt and FPGAs sha3. All in the same client, all in the same blockchain...

[BR0KK]

And all three of them solve the same block chain?

[nelisky]

Yeah, I didn't really think this through, of course, so this might be total bs, but it is possible to change the validator so that the hashing algo is  something set on the hash, say sha256(sha256()):BLOCKHASHGOESHERE or scrypt:BLOCKHASHGOESHERE, and this way bitcoind could still validate all blocks irrespective of the hash. Surely the list of available algos needs to be fixed, though that would in theory be upgradeable by new versions of the software.

Not really knowing the validator code I can't say how complex it would be to manage multiple difficulty points, one for each algo, but all in all this would further future proof the system. If sha256 gets broken we can invalidate it using the majority vote for not accepting it on new blocks...

[ldrgn]

Yeah, I didn't really think this through, of course, so this might be total bs, but it is possible to change the validator so that the hashing algo is  something set on the hash, say sha256(sha256()):BLOCKHASHGOESHERE or scrypt:BLOCKHASHGOESHERE, and this way bitcoind could still validate all blocks irrespective of the hash. Surely the list of available algos needs to be fixed, though that would in theory be upgradeable by new versions of the software.

It sounds like you're proposing the creation of a market for mining on different hashing algorithms.  This is a pretty cool idea and I think warrants further discussion in the technical discussion subforum.

[nelisky]

Yeah, I'm liking this idea more and more... I'll start a discussion in the proper location.

(edit) -> https://bitcointalk.org/index.php?topic=138664.0

[Littleshop]

If BFL unlocks the FPGA singles (they may be unlocked already, I do not know), they can be used for other purposes such as password cracking, genetic research and many other things.   The chip in the FPGA single is quite powerful and in a finished box $600 is not an unusual price for this kind of power. 

[iambaboon]

Is there any software/project that is already using, thinks or had planned on using FPGAs for password cracking & genetic research?

[hardcore-fs]

I wish to build a new alt-currency with a firmware for spartan-6 that contains a new hashing algo (that should be evolvable thus eliminating the possibility to ASIC it). And build a really lightweight client with merkle root thin chain in java.

....Just add a memory requirment......