Aleo Testnet 3
Bird's Eye View
blockHash0
blockHash1
blockHashk \(= B_k\)
- Who will add the next block?
- Decide by creating a puzzle
- Whoever solves it first wins the race to add the new block
- The puzzle must be a function of the block hash
- Should it be deterministic w.r.t. block hash?
Miner \(M\)
Solution \(1\)
Solution \(2\)
Solution \(c\)
\(\vdots\)
Bird's Eye View
blockHash0
blockHash1
blockHashk \(= B_k\)
Miner \(M\)
Solution \(1\)
Solution \(2\)
Solution \(c\)
\(\vdots\)
\textsf{mine}(B_k, M, c)
p = \textsf{createPuzzle}(B_k)
s = \textsf{solvePuzzle}(p, B_k, M, c)
\text{if } \ \textsf{meetsTarget}(s):
\text{return } \ s
\text{else} :
\textsf{mine}(B_k, M, c+1)
Bird's Eye View
\textsf{createPuzzle}(B_k)
\textsf{solvePuzzle}(p, B_k, M, c)
R = [\dots]
\text{// register table}
I_N = \textsf{sampleInst}(B_k, R)
\text{// sample $N$ instructions}
S = \textsf{createStack}(I_N)
p = \textsf{program}(S, R)
\text{return } \ p
\text{// move instructions to an empty stack}
\textsf{rng} = \textsf{sha256}(\textsf{sha256}(B_k, M, c))
\text{// create a random number generator}
x = \textsf{randomInputs}(p, \textsf{rng})
\text{// generates random inputs, } x \in \mathbb{B}/\mathbb{F}/i8/i16/i32/i64/i128
t = \textsf{r1cs}(y = p(x))
\text{// generates r1cs trace}
L = \textsf{toBits}(t)
\text{// creates leaves by converting the trace to bits}
\text{return } \ \textsf{merkleRoot}_{8, 8}(L)
\text{// solution is merkle root of depth-8, arity-8 tree}
Bird's Eye View
\textsf{createPuzzle}(B_k)
\textsf{solvePuzzle}(p, B_k, M, c)
R = [\dots]
I_N = \textsf{sampleInst}(B_k, R)
S = \textsf{createStack}(I_N)
p = \textsf{program}(S, R)
\text{return } \ p
\textsf{rng} = \textsf{sha256}(\textsf{sha256}(B_k, M, c))
x = \textsf{randomInputs}(p, \textsf{rng})
t = \textsf{r1cs}(y = p(x))
L = \textsf{toBits}(t)
\text{return } \ \textsf{merkleRoot}_{8, 8}(L)
- What are the instruction types?
- What is the distribution?
- Can we predict leaves based on the distribution?
- Sizes of random inputs are known.
- We don't need to "build" circuit, just need witnesses.
- Padding in merkle-tree with arity 8.
- Leaves are 0 or 1.
L
\frac{L}{4}
\frac{L}{8}
Convolution
3@128x128
364@128x128
364@128x128
364@128x128
48@128x128
48@1x1
N Convolution layers
Convolution
Avg pooling
Fully connected
