This EIP modifies ethash in order to break ASIC miners specialized for the current ethash mining algorithm.
## Abstract
This EIP pursue "obsolete current ASIC miners" by modifying PoW algorithm in a very low risk manner and update to latest hash algorithm from deprecated FNV Hash algorithms.
Following TEthashV1 algorithm suggests safe transition of PoW algorithms and secure the FNV Algorithm in MIX Parts.
## Motivation
Provide original Ethash proof of work verification with minimal set of changes by updating FNV0 algorithm
## Specification
#### 1. Reference materials on ETHASH FNV0
#### Where FNV Applied on ETHASH
- In [ETHASH](https://github.com/ethereum/wiki/wiki/Ethash) , FNV Hash is used on
* 1) On data aggregation function, MIX parts.
* Ethash Algorithm
```
Header + Nonce
|
Keccak
|
**[MIX 0]** --> **[DAG Page]**
| |
Mixing <--|
...
|
**[Mix 63]**
|
|-----> Mix64 [Process] ---> Mix Digest [32B]
```
* FNV used in DAG Generation
and Mixing for random access or DAG Page.
#### 2. Current applied Ethash FNV hash implementation is deprecated now.
When analysed FNV-0 , there's very weak [avalanche effect](https://simple.wikipedia.org/wiki/Avalanche_effect), when hash input changes on 1~2bits. refer [FNV-Analysis reference section](https://github.com/tao-foundation/FNV-Analysis#how-to-test-and-analysis-reference-test-code)
We need to research and apply newer FNV hash or short message hash algorithm.
There's a implementation that looks like "Missing Offset Bias" at **FNV1A**.
Quotation of [original algorithm FNV1A](https://en.wikipedia.org/wiki/Fowler%E2%80%93Noll%E2%80%93Vo_hash_function#FNV-1a_hash)
```
use hash offset
FNV-1a hash
The FNV-1a hash differs from the FNV-1 hash by only the order in which the multiply and XOR is performed:[8][10]
hash = FNV_offset_basis
for each byte_of_data to be hashed
hash = hash XOR byte_of_data
hash = hash × FNV_prime
return hash
```
FNV_offset_basis and computation order change of xor and multiplication Makes one more xor and multiply computation, but more secure hash effects than FNV0.
and make dispersion boundary condition (0, even number, ..) by using of Prime Number.
#### 4. Real Implementation for FNV1A
Consider real computation resources, in TEthashV1 uses hash byte_of_data to 4bytes aligned data.
In TETHashV1, Adapts fully follow the FNV1A implementation.
`VEC(0, 1, ffffffff, 0)` : is `fnv_reduce` input vector (a,b,c,d)
`FNV :00000001, DF=00000001(01) DS(00000001)` :
*`FNV(00,01)` result is 00000001 ,
*`DF` : is changed bitcounts, compared with previous outputs, in this case prev[00,00] current[00,01] input is 1bit changed, and output result 1bit changed.
*`DS` : is distances of previous result and current result , ABS(prev_fnvresult,current_fnvresult).
** Basically, `DF` is higher is best on hash algorithm.
`F___RC=fefffe6d, DF:111e47a9(14)` : `fnv_reduce = fnv(fnv(fnv(a,b),c),d) ` result is fefffe6d , and Different Bits counts are `14` bits.
## Rationale
In case of ethash algorithm, it can't prevent ASIC forever.
Tethash miner has 2~3% of hashrate degrade on GPU, due to more core computation time.
## Copyright
This work is licensed under a [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License](https://creativecommons.org/licenses/by-nc-sa/4.0/).
