Closed petschge closed 4 years ago
Merging #119 into hdf5 will increase coverage by
0.33%
. The diff coverage is86.97%
.
@@ Coverage Diff @@
## hdf5 #119 +/- ##
==========================================
+ Coverage 63.62% 63.95% +0.33%
==========================================
Files 115 116 +1
Lines 10474 10491 +17
Branches 1491 1496 +5
==========================================
+ Hits 6664 6710 +46
+ Misses 3287 3252 -35
- Partials 523 529 +6
Impacted Files | Coverage Δ | |
---|---|---|
src/vpic/dump.cc | 67.08% <ø> (ø) |
|
src/sf_interface/clear_array.cc | 50.00% <50.00%> (ø) |
|
src/sf_interface/reduce_array.cc | 50.00% <50.00%> (ø) |
|
src/sf_interface/pipeline/reduce_array_pipeline.cc | 73.17% <73.17%> (ø) |
|
src/sf_interface/hydro_array.cc | 92.64% <84.61%> (-0.91%) |
:arrow_down: |
src/sf_interface/pipeline/clear_array_pipeline.cc | 87.87% <87.87%> (ø) |
|
...cies_advance/standard/pipeline/hydro_p_pipeline.cc | 95.78% <95.78%> (ø) |
|
src/species_advance/standard/hydro_p.cc | 100.00% <100.00%> (+2.53%) |
:arrow_up: |
src/util/pipelines/pipelines_thread.cc | 61.68% <0.00%> (-1.87%) |
:arrow_down: |
... and 2 more |
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 d4a1507...827983a. Read the comment docs.
The first commit just cherry-picks the pipelines hydro_p from devel.
The second commit adds a flag to accumulate_hydro_p to weigh contributions of a species by the mass instead of the charge. This makes computation of the bulk flow speed much easier. Additionally the flag disables computation of higher moments of the distribution function as that is not required for the bulk flow speed.
The first commit brought my testcase from 280 seconds down to 88 seconds. The second commit does not change the speed when the flag is true (still 88 seconds, so not much slow-down despite a if-in-a-loop situation) and brings the case when the flag is false down to 62 seconds. Furthermore we can cut our memory overhead in half when all we want is bulk speed and the code in the deck reduces from ~40 lines to ~5.