Use bucket sort to schedule comm batches in distributed-memory
This avoids quadratic runtime in the previous "batchy toposort".
Co-authored-by: Andreas Kloeckner <inform@tiker.net>
parent
dea23730
Loading
Loading
Pipeline
#483669
passed
with stage
in
1 hour, 8 minutes, and 9 seconds
Loading
Please register or sign in to comment