more specific logfile limitations

1 files changed, 2 insertions, 1 deletions

diff --git a/README.md b/README.md
index 38294e5..a57b9c0 100644
--- a/README.md
+++ b/README.md
@@ -70,7 +70,8 @@ We also have a short [video example](https://ess.cs.uos.de/static/videos/splc22-
 ## Log-Based Performance Model Generation
 
 Here, dfatool works with lines of the form "`[::]` *Key* *Attribute* | *parameters* | *NFP values*", where *parameters* is a space-separated series of *param=value* entries (i.e., benchmark configuration) and *NFP values* is a space-separate series of *NFP=value* entries (i.e., benchmark output).
-All measurements of a given *Key* must use the same set of parameter names and NFP names.
+All measurements of a given *Key* *Attribute* combination must use the same set of NFP names.
+Parameter names may be different -- parameters that are present in other lines of the same *Key* *Attribute* will be treated as undefined in those lines where they are missing.
 
 Use `bin/analyze-log.py file1.txt file2.txt ...` for analysis.