Bitcoin Forum

I have been trying to design my own SHA256 ASIC for a while now and I'm starting to nearly grasp how the entire hashing algorithm works.
The only thing I seem to have difficulty understanding is how the hash function deals with "extended" strings of data.

That is, its ability to deal with varying lengths of data.
How does it deal with it and does it interfere much with ASIC hashing?

You might have more luck in the technical forum.

https://bitcointalk.org/index.php?board=6.0

Thanks. Is it possible to get this moved?

Yep, you can even do it yourself. Look at the bottom left.

I don't know the first thing about ASIC programming.... but if you're writing software that requires a large iteration count (PBKDF2, etc.) I'd probably take the "easy" way out and just do the first iteration in software, and do the rest with a single hash block and a hard-coded data length in the ASIC (what I lazily do with GPU/OpenCL-based stuff).

Are you trying to design a generic SHA256 ASIC, or a Bitcoin mining ASIC?  There's a fair bit of difference there.

If the latter, I'd suggest looking at some of the public documents for the older FPGA miners, Avalon's chips, the data mining software sends to simple chips (i.e. not miners with their own controller board). etc.
For a quick "ahhh", see: http://crypto.stackexchange.com/questions/1862/how-can-i-calculate-the-sha-256-midstate

1. In the general case, you add padding.
2. In the bitcoin case, the header has a fixed size. In fact, the nonce is in the 2nd block so you can reuse the state after you have done the first block.

Bitcoin mining ASIC. Thanks for the link. I'll look it over  :)

You could start reading Dadda's paper.
"The design of a high speed ASIC unit for the hash function SHA-256 (384, 512)"
 

Great info! Thanks for sharing.

EDIT: Hmm. Probably no place to get this without paywall, right?

Maybe at university libraries?

Google → filetype:pdf "design of a high speed ASIC unit"
( you'll want to "View more results" on the last page - unless you get a different result order )

Or just pay the $31.. if you're designing an ASIC, that should be one of the rather small line items in your expenses :)

Let me know if this is the one. If it is, I'll email you a copy and take this one down.

https://i.imgur.com/0N3Ch78.jpg

That looks like real useful piece of information. Is it possible to get that e-mailed please?

Instructions are hard, I guess:
→ http://www.date-conference.com/proceedings/PAPERS/2004/DATE04/DF_FILES/04D_5.PDF

The instructions were easy. I had difficulty finding the right PDF. Thank you and sorry for the extra work you had to put in to link it.  ;)

Tip me a coffee, I think that's fair recompense ;)
( wait, does that tip bot thing even monitor BCT yet? Dangit. )

No, that tip bot thingie is pure spam and only a centralized service - there would be next to no difference if Changetip was used together with PayPal instead of Bitcoin.

Maybe when I actually have BTC to start with.

Hah - was firmly tongue-in-cheek, though now I almost wish I could tip you ;)

See above - don't want to derail the thread with a discussion on tipping systems :)

I've got a pretty good idea of how the hashing actually woks although I have difficulty wrapping my head around the compression.

From what I've gathered, the data is split into suitable chunks and then hashed. If its less then ideal, it becomes padded although I doubt Bitcoin hashing requires it.

The Avalon data sheets point to a method of sorts for preprocessing where the data is split up outside the chip before being sent to it.

I also remember TheRealSteve telling me that a difference exists between typical SHA256 and the Merkle-Damgard construct...or Merkle Tree, if my vernacular is correct.