code wiki / _hdl_build / nx_serving_census.nx
nx_serving_census.nx
buildroot/runtime/_hdl_build/nx_serving_census.nx
about
nx_serving_census.nx -- HONESTY GATE on the SERVING / QUEUING / RESOURCE-SHARING subsystem (operator recall:
"we were reusing the z-image LLM -- images worked, chat was hit-or-miss because it wasn't fine-tuned, BUT the
QUEUING and RESOURCE SHARING appeared to be getting state of the art"). This census tests THAT claim: how does
our sovereign serving stack (continuous batching + KV-paging + VRAM budgeting + fair sched + dead-letter + the
:11434 seat, all .nx) compare to the inference-serving SOTA (vLLM continuous-batching+PagedAttention · TGI ·
NVIDIA Triton/TensorRT-LLM · SGLang prefix-cache · Ray Serve · NVIDIA MPS/MIG GPU-partition)? verdict =
xcd_verdict(us,SOTA) COMPUTED never asserted; sovereignty tagged [FLOOR]; AHEAD liar-killed; neg-control must
fire. Grounded in the real organs via have(). Separates PATTERN-parity (algorithms) from PERF (the honest gap).
license_tier: ORIGINAL expect_exit: 0
dependencies 1 imports · 0 importers
imports: nx_cms_exceed.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
| none |
functions
| 12 | func sw(s: *u8) -> i64 { var n: i64=0; while s[n]!=(0 as u8){n=n+1} sys_write(1,s,n); return 0 } |
| 13 | func sn(v: i64) -> i64 { let bb: *u8=sys_mmap(28); var m: i64=v; if m<0{m=0} let t: *u8=sys_mmap(28); var k: i64=0; if m==0{t[0]=48 as u8;k=1} while m>0{t[k]=(48+(m%10)) as u8;m=m/10;k=k+1} var i: i64=0; while i<k{bb[i]=t[k-1-i];i=i+1} sys_write(1,bb,k); return 0 } |
| 14 | func have(path: *u8) -> i64 { let fd: i64=sys_openat_rd(path); if fd<0 {return 0} sys_close(fd); return 1 } called by 1: row |
| 17 | func row(dom: *u8, cap: *u8, our_c: i64, bar_c: i64, n: i64, ev: *u8, ground: *u8, floor: i64, tot: *i64) -> i64 |
| 33 | func main() -> i64 |