nx_nofloat_qwen_fastgen_i8_gate.nx
buildroot/runtime/nx_nofloat_qwen_fastgen_i8_gate.nx
about
nx_nofloat_qwen_fastgen_i8_gate.nx -- Stage 2a of the decode-speed arc: W8A8 SIMD decode (__i16x16_madd,
probe-proven 13x matmul, ~0.8% lossy) with the LOSSLESS i32 head kept for token selection. This ISOLATES the
faithfulness impact of quantizing the DECODE projections: prefill (lossless i64) picks token 0 exactly; every
token after comes from the i8 SIMD decode. The gate MEASURES whether the real completion survives quantization
(does it still say ' Paris. It is the largest'?) and how fast it runs. LOSSY by design -> the lossless i32
path stays the hero; this is the optional fast mode. No hw writes (Rule 26). expect_exit: 0 license_tier: ORIGINAL
dependencies 10 imports · 0 importers
imports: nx_syscalls.nxnx_tier.nxnx_le.nxnx_tensor.nxnx_gguf.nxnx_gguf_load.nxnx_gguf_meta.nxnx_nofloat_llm.nxnx_nofloat_tok.nxnx_gate_verdict.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
| 18 | func f8_puts(s: *u8) -> i64 { var n: i64=0; while s[n]!=(0 as u8){n=n+1} sys_write(1,s,n); return 0 } |
| 19 | func f8_num(v: i64) -> i64 { let b: *u8=sys_mmap(28); var m: i64=v; if m<0{m=0-m;sys_write(1,"-" as *u8,1)} 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{b[i]=t[k-1-i];i=i+1} sys_write(1,b,k); return 0 } |
| 20 | func f8_slen(s: *u8) -> i64 { var n: i64=0; while s[n]!=(0 as u8){n=n+1} return n } called by 1: main |
| 22 | func main() -> i64 |