def tdot
bendtorch-gpu@1.3.0.0 · main.bend:658
import bendtorch-gpu@1.3.0.0/main.bend as BendtorchGpu
import Base
import bendtorch@1.3.0.0/main.bend as T
658
659
660def tdot(k: Nat, +ia: U32, +ib: U32, +i: U32, +sb: U32, acc: F32, ra: Array<F32> & F32, rb: Array<F32> & F32) -> T.D2:
661 match k ra rb:
662 case 0n Tuple{a, x} Tuple{b, y}: T.D2{a, b, acc}
663 case 1n+p Tuple{a, +x} Tuple{b, +y}: tdot(p, (ia + (i + 1 : U32) : U32), (ib + sb : U32), (i + 1 : U32), sb, (acc + (x * y : F32) : F32), Array.get(F32, a, (ia + (i + 1 : U32) : U32)), Array.get(F32, b, (ib + sb : U32)))