This python script parses remote_load_trace.csv. It produces three outputs:
request_heatmap.pdf: A heatmap showing cache bandwidth utilization of LOAD and STORE requests issued to each cache within a series of time windows (bins). The bandwidth utilization is shown as a percent in the heatmap so that heatmaps can be compared.
meanlat_heatmap.pdf: A heatmap showing mean latency of all LOADS issued to each cache within a series of time windows (bins).
maxnlat_heatmap.pdf: A heatmap showing mean latency of all LOADS issued to each cache within a series of time windows (bins).
Examples are attached.
usage: remote_prof.py [-h] [-s SOURCE] [--first FIRST] [--last LAST]
[--cycbin CYCBIN] [--nbins NBINS] [--filename FILENAME]
Argument parser for the HB Spacetime Heatmap (Remote Load Profiler)
optional arguments:
-h, --help show this help message and exit
-s SOURCE, --source SOURCE
Source tile y,x for remote requests. Can specify
multiple.
--first FIRST First Cycle for Spacetime Graph
--last LAST Last Cycle for Spacetime Graph
--cycbin CYCBIN Cycles per Bin in Heatmap. Cannot specify both
--cycbin and --nbins simultaneously.
--nbins NBINS Number of Bins in Heatmap. Cannot specify both
--cycbin and --nbins simultaneously.
--filename FILENAME Remote Load Trace CSV File
This python script parses remote_load_trace.csv. It produces three outputs:
request_heatmap.pdf
: A heatmap showing cache bandwidth utilization of LOAD and STORE requests issued to each cache within a series of time windows (bins). The bandwidth utilization is shown as a percent in the heatmap so that heatmaps can be compared.meanlat_heatmap.pdf
: A heatmap showing mean latency of all LOADS issued to each cache within a series of time windows (bins).maxnlat_heatmap.pdf
: A heatmap showing mean latency of all LOADS issued to each cache within a series of time windows (bins).Examples are attached.
maxlat_heatmap.pdf meanlat_heatmap.pdf request_heatmap.pdf