Closed stephenworsley closed 3 years ago
Merging #117 (488a1a0) into unstructured_scheme (582502e) will not change coverage. The diff coverage is
n/a
.
@@ Coverage Diff @@
## unstructured_scheme #117 +/- ##
====================================================
Coverage 99.48% 99.48%
====================================================
Files 24 24
Lines 1550 1550
====================================================
Hits 1542 1542
Misses 8 8
Continue to review full report at Codecov.
Legend - Click here to learn more
螖 = absolute <relative> (impact)
,酶 = not affected
,? = missing data
Powered by Codecov. Last update 582502e...488a1a0. Read the comment docs.
Seems like a good pattern to me 馃憤
When these benchmarks are run on my machine (via nox --session="benchmarks(long snapshot)"
) we get the following results
[ 55.56%] 路路路 ======== ===========
height
-------- -----------
100 517卤40ms
200 621卤60ms
400 974卤90ms
600 1.50卤0.1s
800 2.05卤0.1s
1000 2.62卤0.4s
======== ===========
[ 61.11%] 路路路 long.esmf_regridder.PerformScalabilityGridToGrid.time_perform ok
[ 61.11%] 路路路 ======== ============
height
-------- ------------
100 181卤10ms
200 415卤40ms
400 917卤50ms
600 1.74卤0.1s
800 2.53卤0.1s
1000 3.37卤0.03s
======== ============
[ 66.67%] 路路路 long.esmf_regridder.PerformScalabilityGridToMesh.time_lazy_perform ok
[ 66.67%] 路路路 ======== ===========
height
-------- -----------
100 706卤200ms
200 935卤100ms
400 1.37卤0.2s
600 2.38卤0.4s
800 3.15卤0.5s
1000 4.50卤0.6s
======== ===========
[ 72.22%] 路路路 long.esmf_regridder.PerformScalabilityGridToMesh.time_perform ok
[ 72.22%] 路路路 ======== ===========
height
-------- -----------
100 207卤40ms
200 400卤70ms
400 858卤200ms
600 1.70卤0.4s
800 2.97卤0.5s
1000 3.98卤0.7s
======== ===========
[ 77.78%] 路路路 long.esmf_regridder.PerformScalabilityMeshToGrid.time_lazy_perform ok
[ 77.78%] 路路路 ======== ============
height
-------- ------------
100 363卤80ms
200 425卤90ms
400 815卤100ms
600 1.21卤0.05s
800 1.53卤0.1s
1000 2.08卤0.2s
======== ============
[ 83.33%] 路路路 long.esmf_regridder.PerformScalabilityMeshToGrid.time_perform ok
[ 83.33%] 路路路 ======== ============
height
-------- ------------
100 83.7卤6ms
200 192卤9ms
400 452卤20ms
600 704卤20ms
800 956卤10ms
1000 1.26卤0.05s
======== ============
[ 88.89%] 路路路 long.esmf_regridder.PrepareScalabilityGridToGrid.time_prepare 1/6 failed
[ 88.89%] 路路路 ============ ===========
grid width
------------ -----------
50 147卤10ms
100 589卤20ms
200 2.48卤0.1s
400 11.0卤0.4s
600 25.6卤0.4s
800 failed
============ ===========
[ 94.44%] 路路路 long.esmf_regridder.PrepareScalabilityGridToMesh.time_prepare 1/6 failed
[ 94.44%] 路路路 ============ ============
grid width
------------ ------------
50 141卤5ms
100 584卤8ms
200 2.59卤0.05s
400 11.7卤0.1s
600 27.0卤0.01s
800 failed
============ ============
[100.00%] 路路路 long.esmf_regridder.PrepareScalabilityMeshToGrid.time_prepare 2/6 failed
[100.00%] 路路路 ============ ===========
grid width
------------ -----------
50 168卤4ms
100 772卤20ms
200 3.24卤0.1s
400 13.9卤0.3s
600 failed
800 failed
============ ===========
When these benchmarks are run on my machine (via
nox --session="benchmarks(long snapshot)"
) we get the following results
I have successfully replicated.
Add performance tests for preparing a regridder to the CI performance tests and add long running performance tests (with the ability to run them separately via nox) which test the scalability of preparing and performing with real and lazy data for all the regridders.