Closed edoddridge closed 5 years ago
Merging #212 into master will increase coverage by
0.68%
. The diff coverage is95.21%
.
@@ Coverage Diff @@
## master #212 +/- ##
==========================================
+ Coverage 91.14% 91.82% +0.68%
==========================================
Files 19 20 +1
Lines 1603 1822 +219
Branches 82 96 +14
==========================================
+ Hits 1461 1673 +212
- Misses 116 124 +8
+ Partials 26 25 -1
Impacted Files | Coverage Δ | |
---|---|---|
src/bernoulli.f90 | 100% <100%> (ø) |
:arrow_up: |
src/advection_schemes.f90 | 100% <100%> (ø) |
:arrow_up: |
src/exchange.f90 | 100% <100%> (ø) |
|
src/state_deriv.f90 | 100% <100%> (+8.33%) |
:arrow_up: |
src/aronnax.f90 | 94.18% <100%> (+4.18%) |
:arrow_up: |
src/momentum.f90 | 92.24% <100%> (+0.06%) |
:arrow_up: |
src/enforce_thickness.f90 | 70% <100%> (ø) |
:arrow_up: |
aronnax/driver.py | 93.85% <100%> (+0.16%) |
:arrow_up: |
src/end_run.f90 | 58.82% <100%> (ø) |
:arrow_up: |
src/thickness.f90 | 97.67% <100%> (+0.02%) |
:arrow_up: |
... and 9 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 539fd8d...9691557. Read the comment docs.
Aronnax can now use MPI to run on multiple processors. Closes #47
I have only benchmarked on my laptop, but going from one core to two cores increased the speed by 65%. While there's room for improvement, this is still excellent.
There are currently far too many exchange calls. These were required to keep the results identical to previous single-core results which wrapped fields far too often. Judicious use of the halos should remove the need for this, and will be addressed in the future. These calls will likely limit the speed up that can be obtained with large numbers (O(10s-100s)) of processors on big clusters.
All previous tests pass. In addition to this, there are many new multicore tests:
Once merged this is worthy of a new release! 0.3.0 here we come.