AlphaGitHub
Source/Packages

Realization.Nvidia.SM86.EmbeddingGatherF32SM86

packages/realizations/cooperative/nvidia-sm86/src/Realization/Nvidia/SM86/EmbeddingGatherF32SM86.alpha

672 lines106 declarations27.6 KiBSHA-256 1a27ac9fb9de
Download exact fileView at revisionf5034773ad4bb52133

family · lines 87–91

EmbeddingGatherF32Width

Full file
Each thread moves four floats. A width choice determines both the row stride in the program and the number of threads in its launch.
87family EmbeddingGatherF32Width : Type 0
88constructor EmbeddingGatherF32Width1024
89constructor EmbeddingGatherF32Width576
90constructor EmbeddingGatherF32Width512
91constructor EmbeddingGatherF32Width256

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.