Skip to content

Keryx-node-v1.2.6-OPoI

Latest

Choose a tag to compare

@slashkrx slashkrx released this 08 Jun 16:29

Keryx Node v1.2.6 — SALT v4 & Virtual Processor Fix

Mandatory update for ALL node operators and miners. v1.2.6 activates a new PoW salt (SALT v4)

  • Critical fix: resolves the virtual-processor stall that was throttling the chain to ~0.3 BPS and starving miners.
  • Clean relaunch on stock Kaspa difficulty (SALT v4) — the experimental difficulty re-anchoring from v1.2.4/v1.2.5 is gone. Difficulty is back to vanilla Kaspa auto-adjustment.
  • No data loss: full history, balances and archival state are preserved.

What was wrong

Since the relaunch, the node's single-threaded virtual processor was spending ~3.8 seconds per block inside sink_search_algorithm. It was repeatedly trying to reorg onto an abandoned, higher-blue-work branch left in the datadir (the old salt-v3 chain, stored header-only), walking the entire dead branch on every block — ~425,000 iterations per resolve. The GPU sat idle waiting for templates, the chain crawled at ~0.3 BPS, and difficulty never recovered.

This was not a difficulty problem — every previous diff re-anchor only masked it cosmetically.

The fix

sink_search_algorithm now skips candidates that can never be the sink — header-only blocks (no UTXO state) and blocks already disqualified from the chain — without walking their parents. The valid sink is always reachable through the body tips already in the heap, so the result is identical (same sink, same candidate set) — it just no longer wastes the walk.

Impact:

  • Sink_search iterations: 425,047 > 22 iterations/block
  • Sink_search time: 3.78s > ~2ms/block
  • Block rate: ~0.3 BPS > ~10 BPS
  • GPU Hashrate: 6 MH/s > 761 Mh/s

This is a correct, fork-safe change (it never alters which sink is selected) and it generalizes: any node will now shrug off abandoned high-blue-work branches instead of choking on them.