nx_qwen_native_gate.nx
buildroot/runtime/nx_qwen_native_gate.nx
about
nx_qwen_native_gate.nx -- NATIVE WINDOWS " Paris" forward.
Loads the real Qwen2.5-0.5B from a Windows path via sys_read_file (native
CreateFile/ReadFile bridge), runs the ST greedy forward (NX_Q8_ST_MAC_MAX
forced huge -> no pool/futex/clone), prints GEN[...] via sys_write. Compile
with the FULL compiler (nx_compile_x86.elf), wrap with nx_pe_natbw, run native.
license_tier: ORIGINAL expect_exit: n_emit
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
| 30 | const NX_GATE_NTOK: i64 = 24 // tokens to emit (use 2 vs 34 to load-subtract per-tok decode) |
| 31 | const NX_GATE_NATIVE: i64 = 1 // 0=all serial ; 1=matmul->pteam-12, block->inline (PROVEN 1.8x, 2026-07-13; BEST) ; 2=block->capped-pool (REJECTED: two-team contention, 13-25s vs 12s, 2026-07-14) |
functions
| 33 | func _put_num(v: i64) -> i64 |
| 46 | func main() -> i64 |