Open reetzelhaft opened 5 years ago
Merging #17 into master will not change coverage. The diff coverage is
0%
.
@@ Coverage Diff @@
## master #17 +/- ##
=======================================
Coverage 26.95% 26.95%
=======================================
Files 63 63
Lines 15328 15328
Branches 7254 7254
=======================================
Hits 4132 4132
Misses 11159 11159
Partials 37 37
Impacted Files | Coverage Δ | |
---|---|---|
channelflow/flowfield.cpp | 40.58% <0%> (ø) |
:arrow_up: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update d334948...4b664ce. Read the comment docs.
The number of actual grid points of saved FlowFields (in physical space) can differ between HDF5 and NetCDF formats. This only happens if dealiasing using the 2/3-rule was active while computing the FlowField. Using NetCDF the x-grid without the padded modes is of size 2(Nx/3). Using HDF5 the x-grid without the padded modes was of size (2Nx)/3. That was too conservative.
Thanks Pavan Kashyap for the bug report.
I adjusted the size for HDF5 files to be consistent with NetCDF. The 2/3-rule says that all modes
k
should be set to zero ifabs(k)>=N/3
(see Canuto et al. 2006, p.139). This way it is and was always implemented in Channelflow. Such a condition allows for the NetCDF style of saving using less points.This fix does not impair backward compatibility of HDF5 since it affects the writing only.