Each cold PLE row demand-faults a 4 KiB page for ~90 bytes of data,
capping cold diverse-text prefill at 218-360 tok/s vs ~785 warm on
GB10. All n-gram row indices of a ubatch are known host-side before
the graph runs, so under the new LLAMA_LAZY_MODE_DIRECT
(--lazy-mode on-direct) they are staged into an input tensor with
sorted, deduplicated, parallel pread()s and dequantized exactly like
ggml_get_rows; downstream kernels unchanged, table stays on disk.
Cold diverse prefill on qwen3.8-flash-next: 542-741 tok/s (2.0-3.1x,
within ~6% of warm); warm, decode and greedy outputs bit-identical.
Josh Leverette
302bd55b5f