nx_qwen_wsl_timing_gate.nx
buildroot/runtime/nx_qwen_wsl_timing_gate.nx
about
nx_qwen_wsl_timing_gate.nx -- WSL2 forward on the DEFAULT threaded pool (the
production path: matmul pool_force + SwiGLU pool, hw_worker_count workers),
SELF-TIMED around nx_f32_llm_run_v3 via sys_now_us (Linux clock). For the
same-box WSL2-vs-native fair A/B: run at NX_WSL_NTOK=2 and =34, then
decode_tok/s = 32e6 / (RUN_US_34 - RUN_US_2) (subtracts the common prefill).
Model = /tmp/nx_real_model.gguf (symlink to the shared 491MB gguf).
license_tier: ORIGINAL expect_exit: 0
dependencies 22 imports · 0 importers
diagram shows first 10 each side; +12 more imports, +0 more importers in the complete lists below.
imports: nx_syscalls.nxnx_tier.nxnx_bpe.nxnx_gguf.nxnx_gguf_load.nxnx_gguf_meta.nxnx_f32.nxnx_f32_kv_cache.nxnx_f32_lazy_weight.nxnx_f32_llama_block.nxnx_f32_llama_block_v4.nxnx_f32_llama_stack_v4.nxnx_f32_llama_layer_lazy_load.nxnx_f32_llm.nxnx_f32_llm_v4.nxnx_f32_llm_read_dims.nxnx_f32_bpe_load.nxnx_f32_llm_special_tokens.nxnx_f32_sampler.nxnx_f32_llm_run_v2.nxnx_f32_llm_run_v3.nxnx_thread_pool.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| 31 | const NX_WSL_NTOK: i64 = 34 // emit count; run at 2 and 34 -> decode = 32/(us34-us2) |
functions
| 33 | func _put_num(v: i64) -> i64 |
| 46 | func main() -> i64 |