Skip to content

BTX 0.32.3

Choose a tag to compare

View all tags
@long long released this 09 Jun 01:28
· 22 commits to main since this release
v0.32.3
898b170
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

BTX version 0.32.3 is now available from:

https://github.com/btxchain/btx/releases/tag/v0.32.3

This v0.32.3 point release updates the CUDA and Metal MatMul mining backends for
post-nMatMulNonceSeedHeight nonce-bound seed derivation. It also includes pool
share-target handling fixes, clearer CUDA fallback diagnostics, shielded-state
repair observability, ZMQ release-binary packaging fixes, and wallet/RPC
follow-ups from the 0.32.x block-125,000 activation series.

Please report bugs using the issue tracker at GitHub:

https://github.com/btxchain/btx/issues

To receive release and update notifications, please subscribe to:

https://btx.dev/

How to Upgrade

If you are running an older version, shut it down. Wait until it has completely
shut down, then install the new binaries or replace the existing btxd,
btx-cli, and GUI binaries with the 0.32.3 release artifacts.

BTX 0.32.3 does not introduce a new protocol activation height. Nodes, miners,
pools, exchanges, services, and explorers should upgrade from earlier 0.32.x
builds for the mining backend and operational fixes.

Compatibility

BTX is supported on Linux, macOS 13+, and Windows 10+.

Notable changes

  • CUDA mining now has a post-nMatMulNonceSeedHeight nonce-seed path. The
    solver batches nonce-seeded pre-hash scans on CUDA, returns only candidates
    that pass the consensus sigma gate, and runs variable-base digest batches so
    per-candidate A/B matrix generation happens on the GPU instead of one nonce at
    a time on the CPU. The CUDA optimization notes recorded a representative
    height-125,000 benchmark improving from about 14.1k nonces/sec to about 2.45M
    nonces/sec while restoring high GPU utilization.

  • Metal mining now uses the same post-activation structure: Metal pre-hash scan,
    variable-base digest batching, and miner-loop routing through the GPU batch
    path. The Metal nonce-seed default batch size scales from the detected Apple
    GPU core count, and btx-matmul-backend-info / btx-matmul-solve-bench now
    report the Metal device name, GPU core count, and detection source.

  • CUDA nonce-seed batch sizing was tuned for production mining hosts. Operators
    can use BTX_MATMUL_NONCE_SEED_BATCH_SIZE for an exact override, and CUDA
    auto-sizing can be capped with BTX_MATMUL_CUDA_NONCE_SEED_MEMORY_PERCENT.
    The nonce-seed CUDA path currently uses the first selected visible CUDA device
    when BTX_MATMUL_CUDA_DEVICES lists multiple devices.

  • Pool share_target_override handling was restored for the MatMul solvers.
    The override relaxes only the digest early-exit target used to return pool
    shares; the consensus pre-hash gate and pre-hash scan window continue to use
    the block target derived from nBits.

  • CUDA fallback behavior is now visible in logs and diagnostics. When CUDA is
    requested but unavailable, the mining backend line includes the concrete
    reason. Runtime CUDA-to-CPU fallback warnings are emitted immediately and then
    throttled for sustained failures, and the new
    doc/btx-cuda-mining-troubleshooting.md guide covers common driver,
    architecture-list, WSL2, and kernel-load failures.

  • CUDA compute-capability gating now supports an explicit older-GPU opt-in.
    BTX_CUDA_ALLOW_OLDER_GPUS=1 lowers the default sm_80+ floor to sm_60+
    for operators who have independently validated their hardware and built
    kernels for those architectures.

  • Shielded-state operations have improved repair visibility. Startup rebuilds
    now log progress more clearly, -resetshieldedstate can force a local
    shielded-state repair, and the new getshieldedstateinfo RPC exposes current
    shielded-state status for operators and automation.

  • Static release binaries now include ZMQ support by building and packaging the
    static libzmq.a dependency. Nodes also warn when ZMQ publish options are set
    on a build without ZMQ support.

  • RPC and wallet follow-ups fix getblock verbosity-2 fee reporting for
    shielded transactions and tighten C-002 PSBT boundary handling.

  • Release automation examples now point at the 0.32.3 artifacts.

Credits

Thanks to everyone who contributed code, testing, operational validation, and
release engineering to this release.

Assets 14
Loading