fix calculation for errorbars with multibar support

The old calculation was wrong and only worked with 3
benchmark runs.

This spells out the math more cleanly and has been tested
for with 2,3,4 and 5 benchmark runs to ensure it scales properly.

Signed-off-by: Andy Doan <andy.doan@linaro.org>
1 file changed
tree: 2eea38fca1da8d613085306c7453c520d3c2b890
  1. android_benchmark_views_app/
  2. .gitignore
  3. MANIFEST.in
  4. setup.cfg
  5. setup.py