3 ms·
> All of the BLAKE variants on ARM significantly underperformed Intel (between 20--80% of Intel’s performance) and we believe that this is because the underlyin
by smoldesu 4y ago
> All of the BLAKE variants on ARM significantly underperformed Intel (between 20--80% of Intel’s performance) and we believe that this is because the underlying library being used has no ARM-specific optimizations but it does have optimizations for Intel platforms with AVX2/SSE4.1.
What would ARM-specific optimization look like, in this case? I'm not intimately familiar with the Graviton architecture, but I'm guessing their best option would be NEON-based vectorization which can be quite slow relative to AVX (last I checked).
- ntolia 4y agoYou know, this is a great question. I believe what ARM is pushing now is Scalable Vector Extension (SVE) as the successor to NEON. I haven't seen benchmarks in the wild yet but it does have promise.