Solana developers have advanced a major efficiency rewrite of the network’s Associated Token Account program, one of the most frequently invoked pieces of token infrastructure on the chain. SIMD-0567, which proposes replacing the existing ATA program with a Pinocchio-based implementation called p-ATA, was merged into the Solana Improvement Documents repository on November 1 after completing the required technical reviews and formal verification. The specification preserves the existing program address while replacing its underlying implementation.
The milestone was followed on November 2 by another implementation step: Anza merged the corresponding p-ATA feature gate into Agave. That code provides the mechanism for swapping the legacy Loader v2 ATA program for the new implementation through an in-place migration. Neither merge means p-ATA is already active on mainnet, because the feature still has to progress through network activation before the replacement takes effect.
p-ATA Targets One of Solana’s Largest Compute Consumers
Associated Token Accounts provide the canonical token account derived from a wallet, token mint and token program, making ATA instructions common across transfers, swaps and other token activity. SIMD-0567 analyzed a stratified sample of 30,000 confirmed mainnet blocks and found ATA calls in 11.85% of all transactions and 31.3% of non-vote transactions. The legacy ATA program accounted for an estimated 13.28% of total compute-unit consumption in that sample, making it a high-impact optimization target.
p-ATA rebuilds the program using Pinocchio, a lightweight Rust library designed around no_std, zero-copy parsing and reduced dependency overhead. Benchmark results included in SIMD-0567 show an 81% weighted-average reduction in compute consumption across ATA instruction patterns, with individual improvements ranging from roughly 63% to more than 85%. Based on the measured ATA share of network compute, the authors estimate that deployment could reduce cluster-wide CU consumption by approximately 10%-11%.
Those figures represent modeled savings rather than production results. The proposal also reports about a 35% reduction in local wall-clock execution when comparing p-ATA with the legacy implementation, but that result comes from benchmark testing. Lower CU consumption frees block capacity, yet it does not automatically translate into an equivalent percentage increase in mainnet transaction throughput.
The approach has already been tested on another core component. SIMD-0266 replaced the legacy SPL Token implementation with p-token, also built around Pinocchio. Its benchmarks reduced TransferChecked from 6,200 CUs to 105 CUs, while subsequent network analysis cited an overall reduction of roughly 93% in Token Program compute usage after activation. p-ATA applies the same optimization strategy to the account infrastructure that applications repeatedly invoke alongside token operations.
Feature Gate Still Requires Network Activation
Compatibility is central to the migration. p-ATA keeps the current program ID and maintains the existing ABI, address derivation, instruction discriminators and state-transition behavior for Create, CreateIdempotent and RecoverNested. It also introduces an optional CreateWithArgs instruction that lets advanced callers provide optimization hints when doing so reduces compute use. The design is intended to deliver lower execution overhead without forcing existing applications to migrate to a new program address.
The implementation has also passed a formal-verification stage that had previously delayed the SIMD merge. A reproducible build of the p-ATA binary was deployed to a buffer, and developers verified that its executable hash matches the hash specified in the proposal. The Agave feature-gate code for performing the migration is now merged, but the project’s tracking issue still lists Devnet, Testnet and Mainnet activation as remaining deployment stages.
p-ATA arrives during a broader performance overhaul across Solana. The network recently reduced its mainnet target slot time to 250 milliseconds, while Transaction V1 expanded Solana’s transaction envelope to 4,096 bytes and Alpenglow replaced the network’s previous consensus system. These upgrades target different constraints: p-ATA reduces execution cost inside frequently used program calls, while slot timing, transaction format and consensus changes address separate parts of the processing pipeline.
The next material milestone is therefore not another benchmark but activation. Until the p-ATA feature reaches mainnet, the projected 10%-11% cluster-wide compute reduction remains an engineering estimate rather than a production result.
Jack Reynolds is SatoshiPick’s infrastructure mind. Based in Denmark, he follows Bitcoin, Ethereum, Layer 1 networks, stablecoins and DeFi with a practical question always in the background: does this actually make crypto work better?
His articles focus on the systems beneath the headlines: settlement rails, protocol upgrades, stablecoin usage, DeFi coordination and the regulatory limits around them. Jack avoids turning technical coverage into a maze. He explains what is live, what is still experimental and why the difference matters.