perf: use autoresearch to reduce allocations#3225

Open

tac0turtle wants to merge 17 commits intomainfrom

autoresearch/block-perf-2026-04-04

Commits on Apr 4, 2026

Baseline - current block production performance with 100 txs
tac0turtle
committed
sync.Pool for sha256.Hash in leafHashOpt — eliminates 2 sha256.New() allocations per block
tac0turtle
committed
Unsafe reinterpret cast of Txs to [][]byte in ApplyBlock — eliminates make([][]byte, n) allocation
tac0turtle
committed
Direct pb.Data serialization in DACommitment — avoids pruned Data wrapper and txsToByteSlices allocations
tac0turtle
committed
unsafe.Slice in Data.ToProto() — eliminates txsToByteSlices [][]byte allocation
tac0turtle
committed
sync.Pool for protobuf message structs in MarshalBinary — eliminates 10 allocs per block
tac0turtle
committed
pool SignedHeader.MarshalBinary — reuse pb.SignedHeader/pb.Header/pb.Signer/pb.Version
tac0turtle
committed
pool State.MarshalBinary and use it in UpdateState — saves 2 allocs per block
tac0turtle
committed
fix lint
tac0turtle
committed
remove files
tac0turtle
committed

Commits on Apr 7, 2026

Commits on Apr 9, 2026

Merge branch 'main' into autoresearch/block-perf-2026-04-04
tac0turtle
authored

Commits on Apr 10, 2026

go fix
julienrbrt
committed

Commits on Apr 13, 2026

changes++
tac0turtle
committed