Open Pearl Miner v1.7.1. GPU tensor_hash

What's new

• GPU-resident Merkle tensor_hash in job setup (per-grid setup overhead 42.6% → 9.1%, bit-exact) plus a tiled shared-memory transpose.
• Coopmat layout query — the backend requires the native 16×16 sint8 shape or cleanly falls back to the DP4A path (so RDNA2 and other no-matrix-core GPUs still mine).
• Backend C ported off Win32 (timing → clock_gettime, loader → dlopen) so the same source builds as p40vk.dll (Windows) and libp40vk.so (Linux/ARM).

Solo and pool mining carry the same transparent 1% dev fee.

Public release v1.7.0 (AMD modify)

Changes

• Stable support for AMD Vega series cards
• Stable support for AMD 7000 and 8000 series cards
• Nvidia architecture modifications to improve miner performance
• The latest hashrate tests feature stock graphics cards
• Bug and error fixes

Public release Open Pearl Miner (AMD support)

Maintenance release.

Changes
Hashrate display fix. The TH/s readout counted 1e6 hashes per 16×16 tile, but a tile is 16164096 = 2^20 (1,048,576) MACs — the convention the kernel benchmark already uses. The display was under-reporting the true hashrate by ~4.9%; now consistent across the pool and solo paths. Same physical mining — your reported number rises ~5% (e.g. a card showing 28.2 now correctly shows ~29.6).
Internal: cache the GPU compute capability in the per-region hot path instead of re-querying it every region (no perf change).