Skip to content

ftrace: Homogenize timestamps conversions

Created by: valschneider

By default, the str to float conversion done when reading from csv is different from the one used when reading from the trace.txt file.

Here's an example:

  • trace.txt string timestamps: [76.402065, 80.402065, 80.001337]

  • parsed dataframe timestamps: [76.402065000000007, 80.402065000000007, 82.001337000000007]

  • csv string timestamps: [76.402065, 80.402065, 80.001337]

  • cached dataframe timestamps: [76.402064999999993, 80.402064999999993, 82.001337000000007]

To fix this, the timestamps read from the cache are rounded to microsecond precision, which results in cache-read timestamps being identical to trace-read timestamps.

Tests have also been added/tweaked to ensure this stays true.

Merge request reports