2 comments

  • Neywiny15 minutes ago
    Just checking here- this systemverilog is a hypothetical telling of what you think is going on? Or do you have the actual source of the RTL?
  • eiln2 days ago
    RTL performance erratum in the Apple M3 Neural Engine throttles DRAM weight streaming throughput down to 17–19 GB/s from the nominal 45–60 GB/s. Avoiding the problematic path in the kernel DMA engine's speculative prefetch ring increased Llama 3.2 1B token throughput from 10.0 to 24.3 tokens/s.