Source/Packages

Realization.Nvidia.SM86.AttentionHeadLayoutSM86

packages/realizations/cooperative/nvidia-sm86/src/Realization/Nvidia/SM86/AttentionHeadLayoutSM86.alpha

1,710 lines162 declarations85.3 KiBSHA-256 ad8cef82929b

def · lines 850–909

attentionHeadLayoutSM86TokenToHeadProgram

Full file
Reverse the preceding layout for the attention-output adjoint: fp32 token-major [position][512] to fp16 head-major [head][position][64].
850def attentionHeadLayoutSM86TokenToHeadProgram : (family SM86Program) =
851  (attentionHeadLayoutSM86Prologue
852    (byte 5)
853    attentionHeadLayoutSM86U31
854    (attentionHeadLayoutSM86Next
855      (attentionHeadLayoutSM86IMAD (byte 6) (byte 3) attentionHeadLayoutSM86U2 (byte 7))
856      (attentionHeadLayoutSM86Next
857        (attentionHeadLayoutSM86IMAD (byte 8) (byte 1) attentionHeadLayoutSM86U512 (byte 7))
858        (attentionHeadLayoutSM86Next
859          (attentionHeadLayoutSM86IMAD (byte 8) (byte 2) attentionHeadLayoutSM86U64 (byte 8))
860          (attentionHeadLayoutSM86Next
861            (attentionHeadLayoutSM86IMAD (byte 8) (byte 6) attentionHeadLayoutSM86U1 (byte 8))
862            (attentionHeadLayoutSM86Next
863              (attentionHeadLayoutSM86Wide
864                (byte 10)
865                (byte 8)
866                (byte 4)
867                attentionHeadLayoutSM86InputPointer)
868              (attentionHeadLayoutSM86Next
869                (attentionHeadLayoutSM86Load (byte 12) (byte 10) attentionHeadLayoutSM86U0)
870                (attentionHeadLayoutSM86Next
871                  (attentionHeadLayoutSM86Load (byte 13) (byte 10) attentionHeadLayoutSM86U4)
872                  (attentionHeadLayoutSM86Next
873                    (attentionHeadLayoutSM86IMAD
874                      (byte 16)
875                      (byte 2)
876                      attentionHeadLayoutSM86U65536
877                      (byte 7))
878                    (attentionHeadLayoutSM86Next
879                      (attentionHeadLayoutSM86IMAD
880                        (byte 16)
881                        (byte 1)
882                        attentionHeadLayoutSM86U64
883                        (byte 16))
884                      (attentionHeadLayoutSM86Next
885                        (attentionHeadLayoutSM86IMAD
886                          (byte 16)
887                          (byte 6)
888                          attentionHeadLayoutSM86U1
889                          (byte 16))
890                        (attentionHeadLayoutSM86Next
891                          (attentionHeadLayoutSM86Wide
892                            (byte 18)
893                            (byte 16)
894                            (byte 9)
895                            attentionHeadLayoutSM86FirstOutputPointer)
896                          (attentionHeadLayoutSM86Next
897                            (attentionHeadLayoutSM86Pack
898                              (byte 14)
899                              (byte 13)
900                              (byte 12)
901                              attentionHeadLayoutSM86WaitLoads)
902                            (attentionHeadLayoutSM86Next
903                              (attentionHeadLayoutSM86Store
904                                (byte 18)
905                                (byte 14)
906                                attentionHeadLayoutSM86U0)
907                              (attentionHeadLayoutSM86Next
908                                (constructor SM86InstructionBody SM86Exit sm86SafeControl)
909                                attentionHeadLayoutSM86End)))))))))))))))

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.