code wiki / (root) / nx_f32x8_mt_matmul.nx

nx_f32x8_mt_matmul.nx

buildroot/runtime/nx_f32x8_mt_matmul.nx

5915 B111 linesdepth 2pulls 2 transitivereach 0 importersview sourcekind tooltopic f32x8
docsdependenciesstructsconstsfunctions

about

nx_f32x8_mt_matmul.nx -- the FULL sovereign compute stack toward physics: AVX2 8-wide x MULTICORE. __f32x8_dot (256-bit vmovups+vmulps via .byte VEX) inside forked workers (sys_fork+mmap_shared+wait4). Measures scalar-ST -> AVX2-ST -> AVX2-MT and the gap to the derived silicon f32 peak (~1.84 TFLOP/s). This is the deepest "get to exceed" climb with the levers built this session; what remains = FMA (2x, vfmadd231ps, VEX encoder already proven) + better multicore scaling. Bit-exact vs scalar. Sovereign: nx_cc AVX2 + raw fork, NO BLAS/CUDA/OpenMP. No hw writes (Rule 26). expect_exit: 0 tier: ORIGINAL

dependencies 1 imports · 0 importers

nx_syscalls.nx nx_f32x8_mt_matmul.nx

imports: nx_syscalls.nx

imported by: nobody (leaf or entry point)

call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown

main mp_puts sys_write sys_mmap_shared sys_mmap pack4 sys_now_us sys_mmap ↻ sys_clock_gettime_mono band_scalar band_avx2 sys_fork sys_exit sys_wait4 mp_num sys_mmap ↻ sys_write ↻

structs

none

consts

8const K_MAGIC_1842560: i64 = 1842560

functions

10func mp_puts(s: *u8) -> i64 { var n: i64=0; while s[n]!=(0 as u8){n=n+1} sys_write(1,s,n); return 0 }
called by 1: main calls 1: sys_write
11func mp_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 }
called by 1: main calls 2: sys_mmapsys_write
12func pack4(buf: *u8, idx: i64, bits: i64) -> i64 { buf[idx*4+0]=(bits) as u8; buf[idx*4+1]=(bits>>8) as u8; buf[idx*4+2]=(bits>>16) as u8; buf[idx*4+3]=(bits>>24) as u8; return 0 }
called by 1: main
14func band_scalar(a: *i64, b: *i64, c: *i64, M: i64, N: i64, K: i64, w: i64, nw: i64) -> i64
called by 1: main
25func band_avx2(pa: *u8, pbt: *u8, c: *i64, M: i64, N: i64, K: i64, w: i64, nw: i64) -> i64
called by 1: main
39func main() -> i64