code wiki / (root) / nx_f32x8_matmul_gate.nx

nx_f32x8_matmul_gate.nx

buildroot/runtime/nx_f32x8_matmul_gate.nx

6505 B128 linesdepth 3pulls 3 transitivereach 0 importersview sourcekind gate/prooftopic f32x8
docsdependenciesstructsconstsfunctions

about

nx_f32x8_matmul_gate.nx -- PROVES the AVX2 8-wide lever past SSE 4-wide. Three matmuls, same data: scalar (__f32_mul, 1 MAC/instr) vs SSE packed (__f32x4_dot, 4) vs AVX2 packed (__f32x8_dot, 8 -- vmovups+vmulps 256-bit via .byte VEX, proven nxasm_vex_kat 7/7). Verifies all three agree BIT-EXACT (small-int f32 -> order-independent) + a known-answer dot + MEASURES both packed speedups. The AVX2 rung is the gating compute lever for racing sdcpp. Honest: vextractf128+hsum per 8 is overhead; a vector-accumulating FMA kernel approaches the full 8x -- this first AVX2 kernel measures what it measures. No hw writes (Rule 26). expect_exit: 0 license_tier: ORIGINAL

dependencies 2 imports · 0 importers

nx_syscalls.nx nx_gate_verdict.nx nx_f32x8_matmul_gate.nx

imports: nx_syscalls.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

main gx_puts sys_write sys_mmap pack4 mm_scalar mm_x4 mm_x8 sys_now_us sys_mmap ↻ sys_clock_gettime_mono gx_num sys_mmap ↻ sys_write ↻ gv_ctr sys_mmap ↻ gv_verdict gv_puts sys_write ↻ gv_num sys_mmap ↻ sys_write ↻ sys_munmap gv_journal sys_openat_append sys_mmap ↻ gv_catn sys_mmap ↻ sys_munmap ↻ sys_now_realtime_sec sys_mmap ↻ sys_clock_gettime_real gv_cat sys_write ↻ sys_close sys_munmap ↻ sys_exit

structs

none

consts

none

functions

11func gx_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
12func gx_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
13func 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
15func mm_scalar(sa: *i64, sb: *i64, c: *i64, M: i64, N: i64, K: i64) -> i64
called by 1: main
24func mm_x4(pa: *u8, pb: *u8, c: *i64, M: i64, N: i64, K: i64) -> i64
called by 1: main
35func mm_x8(pa: *u8, pb: *u8, c: *i64, M: i64, N: i64, K: i64) -> i64
called by 1: main
47func main() -> i64
calls 11: gx_putssys_mmappack4mm_scalarmm_x4mm_x8+5