Issue with GPU Kernel Execution in Unified Memory Mode

[DDantalion]

I'm running the sample implementation of PageRank with the following command:
./pagerank -timing -trace-mem -unified-gpus 1,2,3,4 -use-unified-memory -report-all
However, I noticed that only GPU1 shows kernel execution time, while GPUs 2, 3, and 4 remain idle. According to the code, workgroups should have been distributed across all four GPUs.

[DX990307]

The size of the benchmark is not enough to drive all GPUs. You can increase the size of the benchmark like

mgpusim/samples/pagerank/main.go

Line 12 in 383c963

var numNode = flag.Int("node", 16, "The number of nodes")

change the node counts from 16 to 512.