Skip to content

Investigate writing custom c++/CUDA function #7

Description

@rdaly525

Note, 'N' refers to the number of select bits of a mux (or the number of input bits of a LUT).
'K' refers to the number of LUTs in a particular layer.

layers.py:135
The function LutLayer is a single layer of LUTs. This specifies the connectivity as well as instantiating the Luts.

layers.py:113
The function LutN creates the LUT weights and the Mux.

layers.py:66
The function MuxSTriangle does the actual Mux computation. 'I' is mux input, 'S' is mux select. For a LUT, the 'I' is the weights, and the 'S' is the LUT input.

For custom code, we need to speedup either the full LutLayer (which instantiates a LutN which instantiates a MuxSTriangle) or just the MuxSTriangle.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions