[llvm-bugs] [Bug 40880] New: [llvm-exegesis] Analysis: smarter clustering?

via llvm-bugs llvm-bugs at lists.llvm.org
Wed Feb 27 03:00:54 PST 2019


            Bug ID: 40880
           Summary: [llvm-exegesis] Analysis: smarter clustering?
           Product: tools
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: llvm-exegesis
          Assignee: unassignedbugs at nondot.org
          Reporter: lebedev.ri at gmail.com
                CC: clement.courbet at gmail.com, gchatelet at google.com,
                    llvm-bugs at lists.llvm.org

Even with all the fixes, i'm still seeing some confusing behavior.
Even with `-analysis-clustering-epsilon=0.001`, the clusters end up
with very different instructions (e.g. rthroughput of 1 vs rthroughput of 2)
in the same cluster. I'm not sure if it is a bug or not, but i'm very
sure this is not the behavior one wants.

Has anyone else seen this? I'm not sure how to show an reproducer,
without uploading the entire measurement set..

Any thoughts on this? 
Perhaps the clustering algo is simply misimplemented?

Or maybe some other algorithm should be used?
E.g. go through ever SchedWrite class, group all the measurements
for all the instructions that are within this SchedWrite class into
a cluster, then crop off the outliers into their own clusters?

You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20190227/c78ce81a/attachment.html>

More information about the llvm-bugs mailing list