minerva

README

Fundamentally achieving ASIC resistant: A brief introduction to Truehash algorithm

As we all know, TaiyueChain adopted hybrid consensus technology, It uses PBFT as fast chain to process transactions and PoW as snailchain to supervise and elect the committee.

PoW algorithm was put forward by Cynthia Dwork and Moni Naor in 1993. The application scenario of it then was to resist the attack of ddos and email spam. In 1999, Jacobsson came up with partial hash inversion algorithm which is also the method that Satoshi Nakamoto utilized in bitcoin network.

Take a look at the coins that issued after the bitcion, it is not hard to see that blockchain developers are experimenting different kinds of mining algorithm to achieve ASIC resistant. And it turns out to be that all the attempts are failed. Their whole idea is to adding the computation against the shortcomings of ASIC to make the algorithms harder. However, algorithm like Ethash and Equihash or others may merely delay the appearance time of ASIC at best but cannot fundamentally resist ASIC. For further research, we must understand why CPU/GPU is way much slower than ASIC. And the main reason is that von Neumann architecture is adopted for Diversity calculation.

Von Neumann architecture is roughly divided into three parts, that is internal storage, control unit and calculating element. The operation program and data are written in DDR and will be transferred to the calculation element when performing calculations. The computational bottleneck caused by transmission bandwidth is called the von neumann bottleneck.

The idea of avoiding the von neumann bottleneck is simple for operations with high uniqueness, such as the calculation of hash values, the idea is to abandon the von neumann architecture and write the algorithm directly into the calculation element to make it become ASIC. That is why ASIC has such an astonishing advantages of computing powers in calculating Bitcoin Sha-256d. Even then Ethash adopted the method of reading the big table, but essentially a computer program can still be hard -coded to calculation elements. Although there's memory grain around the ASIC but that can only increases the cost.

In order to achieve ASIC resistant fundamentally we can borrow the experience from Monero, that is periodic algorithm changing design. The problems with Monero is the unprofessional voting for hard fork by the community. And the biggest problem with them is that no one knows if the project team (they can manipulate the voting results easily) prepared ASIC for new algorithm.

Therefore,we must assure this mechanism is under the control of mining algorithm. And to achieve ASIC resistant fundamentally, the following rules must be satisfied:

1. Command S as a collection of algorithm, all algorithms in S must pass through the von neumann bottleneck.
2. Algorithm will change every T blocks, and the changing methods must meet up with the condition of Verifiable and unpredictable.
3. No human intervention in algorithm switching.

The specific design ideas are as follows

The implementation of Truehash is to set G as a group, for every group member g, make rho_V(G) be the showcase of G in vector space V. In ordinary mining algorithm, when calculate blockheade, nonce and other information with operations like padding and so on, there will form a Vector nonce. By the exhaustion of different nonce to find the results that are less than the difficulty coefficient.

The changing parts of Truehash

Change hash(v(nonce)) to hash(rho(g)*v(nonce)). As long as G (Truehash utilizes permutation group which contains 2048 elements) is complex enough, this algorithm collection will not be entirely hard-coded within the calculation elements. Because the algorithm will switch randomly, the bottleneck of von neumann will be inevitable.

The principle of Truehash's switching algorithm is that the group elements will be changed every 12 thousand PoW block. The production time of 12 thousand poW blocks is 83 days. The new group element information is composed of blocks 1-8192 in the previous cycle, and the composition method is generated by analyzing the hash values of blocks 11001-11256.

Because the hash value of the block is unpredictable in advance, it's impossible for anyone to know anything about the new algorithm until the 11256 block comes along. From the 11,257th block of the last cycle, to the algorithm being invalidated, there is only 88 days in total, so it makes no sense to produce ASIC in such a short period of time.

Documentation

Overview

Package minerva implements the taiyuechain hybrid consensus engine.

Index

• Constants
• Variables
• func LogPrint(info string, addr common.Address, amount *big.Int)
• func MatMuliple(input []uint64, output []uint64, pmat []uint64) int
• func NewDataset(epoch uint64) interface{}
• func SeedHash(block uint64) []byte
• func TruehashLight(dataset []uint64, hash []byte, nonce uint64) ([]byte, []byte)
• type Config
• type ConstSqrt
• type Dataset
  • func (d *Dataset) Hash() common.Hash
• type Minerva
  • func New(config Config, cimList *cim.CimList) *Minerva
  • func NewFakeDelayer(delay time.Duration) *Minerva
  • func NewFakeFailer(fail uint64) *Minerva
  • func NewFaker(cimList *cim.CimList) *Minerva
  • func NewFullFaker() *Minerva
  • func NewShared(cimList *cim.CimList) *Minerva
  • func NewTester(cimList *cim.CimList) *Minerva
  • func (m *Minerva) APIs(chain consensus.ChainReader) []rpc.API
  • func (m *Minerva) Author(header *types.Header) (common.Address, error)
  • func (m *Minerva) Finalize(chain consensus.ChainReader, header *types.Header, state *state.StateDB, ...) (*types.Block, error)
  • func (m *Minerva) GetElection() consensus.CommitteeElection
  • func (m *Minerva) Prepare(chain consensus.ChainReader, header *types.Header) error
  • func (m *Minerva) SetElection(e consensus.CommitteeElection)
  • func (m *Minerva) VerifyHeader(chain consensus.ChainReader, header *types.Header) error
  • func (m *Minerva) VerifyHeaders(chain consensus.ChainReader, headers []*types.Header, seals []bool) (chan<- struct{}, <-chan error)
  • func (m *Minerva) VerifySigns(fastnumber *big.Int, fastHash common.Hash, signs []*types.PbftSign) error
  • func (m *Minerva) VerifySwitchInfo(fastnumber *big.Int, info []*types.CommitteeMember) error
• type Mode
  • func ToMinervaMode(val int) Mode

Constants

const DATALENGTH = 2048 //2048 520

const DGSTSIZE = 32

const HEADSIZE = 32

const OFF_CYCLE_LEN = 8192 //8192  2080

const OFF_SKIP_LEN = 32768 //32768  8230

const OFF_STATR = 12 //in advance updata the algorithm

const PMTSIZE = 4

const SKIP_CYCLE_LEN = 2048 //2048 520

const STARTUPDATENUM = 10240 //10240  2600

const TBLSIZE = 16

const UPDATABLOCKLENGTH = 12000 //12000  3000

Variables

var (

	//SnailBlockRewardsBase Snail block rewards base value is 115.555555555555 * 10^12
	SnailBlockRewardsBase = 115555555555555

	// Big1e6 up to wei  SnailBlockRewardsBase * this is wei
	Big1e6 = big.NewInt(1e6)

	// SnailBlockRewardsChangeInterval Snail block rewards change interval 4500 blocks
	SnailBlockRewardsChangeInterval = 4500

	// SnailBlockRewardsChangePercentage snall block rewards change interval decrease %2
	SnailBlockRewardsChangePercentage = 2

	//BaseBig ...
	BaseBig = big.NewInt(1e18)

	//NetworkFragmentsNuber The number of main network fragments is currently fixed at 1
	NetworkFragmentsNuber = 1

	//MiningConstant Mining constant is 20
	MiningConstant = 20

	//SqrtMin pbft and miner allocation constant
	//Generating formula :TestOutSqrt
	SqrtMin = 25

	//SqrtMax ...
	SqrtMax = 6400
)

var ErrInvalidDumpMagic = errors.New("invalid dump magic")

ErrInvalidDumpMagic errorinfo

Functions

func LogPrint

func LogPrint(info string, addr common.Address, amount *big.Int)

LogPrint log debug

func MatMuliple

func MatMuliple(input []uint64, output []uint64, pmat []uint64) int

func NewDataset

func NewDataset(epoch uint64) interface{}

newDataset creates a new truehash mining dataset

func SeedHash

func SeedHash(block uint64) []byte

SeedHash is the seed to use for generating a verification cache and the mining dataset.

func TruehashLight

func TruehashLight(dataset []uint64, hash []byte, nonce uint64) ([]byte, []byte)

Types

type Config

type Config struct {
	PowMode Mode
}

Config are the configuration parameters of the minerva.

type ConstSqrt

type ConstSqrt struct {
	Num  int     `json:"num"`
	Sqrt float64 `json:"sqrt"`
}

ConstSqrt ...

type Dataset

type Dataset struct {
	// contains filtered or unexported fields
}

dataset wraps an truehash dataset with some metadata to allow easier concurrent use.

func (*Dataset) Hash

func (d *Dataset) Hash() common.Hash

type Minerva

type Minerva struct {
	// contains filtered or unexported fields
}

Minerva consensus

func New

func New(config Config, cimList *cim.CimList) *Minerva

New creates a full sized minerva hybrid consensus scheme.

func NewFakeDelayer

func NewFakeDelayer(delay time.Duration) *Minerva

NewFakeDelayer creates a minerva consensus engine with a fake PoW scheme that accepts all blocks as valid, but delays verifications by some time, though they still have to conform to the Ethereum consensus rules.

func NewFakeFailer

func NewFakeFailer(fail uint64) *Minerva

NewFakeFailer creates a minerva consensus engine with a fake PoW scheme that accepts all blocks as valid apart from the single one specified, though they still have to conform to the Ethereum consensus rules.

func NewFaker

func NewFaker(cimList *cim.CimList) *Minerva

NewFaker creates a minerva consensus engine with a fake PoW scheme that accepts all blocks' seal as valid, though they still have to conform to the Ethereum consensus rules.

func NewFullFaker

func NewFullFaker() *Minerva

NewFullFaker creates an minerva consensus engine with a full fake scheme that accepts all blocks as valid, without checking any consensus rules whatsoever.

func NewShared

func NewShared(cimList *cim.CimList) *Minerva

NewShared creates a full sized minerva shared between all requesters running in the same process.

func NewTester

func NewTester(cimList *cim.CimList) *Minerva

NewTester creates a small sized minerva scheme useful only for testing purposes.

func (*Minerva) APIs

func (m *Minerva) APIs(chain consensus.ChainReader) []rpc.API

APIs implements consensus.Engine, returning the user facing RPC APIs. Currently that is empty.

func (*Minerva) Author

func (m *Minerva) Author(header *types.Header) (common.Address, error)

Author implements consensus.Engine, returning the header's coinbase as the proof-of-work verified author of the block.

func (*Minerva) Finalize

func (m *Minerva) Finalize(chain consensus.ChainReader, header *types.Header, state *state.StateDB,
	txs []*types.Transaction, receipts []*types.Receipt, feeAmount *big.Int) (*types.Block, error)

Finalize implements consensus.Engine, accumulating the block gas rewards, setting the final state and assembling the block.

func (*Minerva) GetElection

func (m *Minerva) GetElection() consensus.CommitteeElection

GetElection return election

func (*Minerva) Prepare

func (m *Minerva) Prepare(chain consensus.ChainReader, header *types.Header) error

Prepare implements consensus.Engine, initializing the difficulty field of a header to conform to the minerva protocol. The changes are done inline.

func (*Minerva) SetElection

func (m *Minerva) SetElection(e consensus.CommitteeElection)

SetElection Append interface CommitteeElection after instantiation

func (*Minerva) VerifyHeader

func (m *Minerva) VerifyHeader(chain consensus.ChainReader, header *types.Header) error

VerifyHeader checks whether a header conforms to the consensus rules of the stock Taiyuechain m engine.

func (*Minerva) VerifyHeaders

func (m *Minerva) VerifyHeaders(chain consensus.ChainReader, headers []*types.Header,
	seals []bool) (chan<- struct{}, <-chan error)

VerifyHeaders is similar to VerifyHeader, but verifies a batch of headers concurrently. The method returns a quit channel to abort the operations and a results channel to retrieve the async verifications.

func (*Minerva) VerifySigns

func (m *Minerva) VerifySigns(fastnumber *big.Int, fastHash common.Hash, signs []*types.PbftSign) error

VerifySigns check the sings included in fast block or fruit

func (*Minerva) VerifySwitchInfo

func (m *Minerva) VerifySwitchInfo(fastnumber *big.Int, info []*types.CommitteeMember) error

type Mode

type Mode uint

Mode defines the type and amount of PoW verification an minerva engine makes.

const (
	ModeNormal Mode = iota
	ModeShared
	ModeTest
	ModeFake
	ModeFullFake
)

constant

func ToMinervaMode

func ToMinervaMode(val int) Mode