Source/Packages

Realization.Nvidia.SM86.StreamingAttentionSM86

packages/realizations/cooperative/nvidia-sm86/src/Realization/Nvidia/SM86/StreamingAttentionSM86.alpha

1,129 lines205 declarations66.1 KiBSHA-256 aefcbc3b00ca

def · lines 304–308

saRescaleOutput

Full file
O's row half h rescaled by alpha (the previous tile's products have finished: the first multiply waits for them)
304def saRescaleOutput = (lambda unrestricted h : Nat . (lambda unrestricted tail : (family SM86Program) .
305  (saFor 16 (lambda unrestricted i : Nat .
306    (let unrestricted e = (saElem (saO (naturalDivideUnchecked i 2)) h (naturalModuloUnchecked i 2)) in
307      (saFmul e e (saTmp (naturalAdd 4 h)) (naturalSelect (naturalIsZero i) saWait3 saWaitNone))))
308    tail)))

The compiler supplied declaration spans and resolved links from this source snapshot. This page does not assert that this file belongs to a checked closure.