code wiki / _hdl_build / nx_f32_gqa_gate.nx
nx_f32_gqa_gate.nx
buildroot/runtime/_hdl_build/nx_f32_gqa_gate.nx
dependencies 4 imports · 0 importers
imports: nx_gate_gn.nxnx_gate_base.nxnx_f32_hw.nxnx_syscalls.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 grow(name: *u8, ok: i64) -> i64 { if ok==1 { gw(" PASS " as *u8) } else { gw(" FAIL " as *u8) } gw(name); gw(" calls 1: gw |
| 20 | func gm(x: i64) -> i64 { return gn(f32_int(f32_mul(x, f32_of(1000)))) } |
| 21 | func f32_le(x: i64, y: i64) -> i64 { let d: i64=f32_sub(x,y) & 0xFFFFFFFF; if ((d>>31)&1)==1 { return 1 } if (d & 0x7FFFFFFF)==0 { return 1 } return 0 } called by 1: main |
| 22 | func f32_abs(x: i64) -> i64 { return x & 0x7FFFFFFF } |
| 23 | func f32_max2(a: i64, b: i64) -> i64 { if f32_le(a,b)==1 { return b } return a } |
| 24 | func f32_sqrt(x: i64) -> i64 { if (x & 0x7FFFFFFF)==0 { return f32_of(0) } var y: i64=x; var i: i64=0; while i<16 { y=f32_div(f32_add(y, f32_div(x,y)), f32_of(2)); i=i+1 } return y } |
| 25 | func f32_exp(x: i64) -> i64 { // range-reduced (the fixed version) |
| 32 | func softmax2(z0: i64, z1: i64, out: *i64) -> i64 { let mx: i64=f32_max2(z0,z1); let e0: i64=f32_exp(f32_sub(z0,mx)); let e1: i64=f32_exp(f32_sub(z1,mx)); let s: i64=f32_add(e0,e1); out[0]=f32_div(e0,s); out[1]=f32_div(e1,s); return 0 } |
| 36 | func out1_0(Q: *i64, K: *i64, V: *i64) -> i64 |
| 44 | func main() -> i64 |