add back uniform_vertical_levels #2016
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
cc @kmdeck |
|
Im not sure if it matters for this, but just in case: we had to make some changes to CC MatrixFields to even create the Jacobian on the spectral element space :#1884 |
|
Can we add a unit test covering this case? Maybe a separate PR with a broken test first, and then we can re-base this PR. |
|
Sorry about the regression! |
9 tasks
charleskawczynski
force-pushed
the
js/matrixfields
branch
from
e85dad4 to
520090b
Compare
|
Hi @juliasloan25, I tried running the reproducer with this branch, and it seems like this PR doesn't fix the issue. I did see that if I define function _single_field_solve!(
::ClimaComms.CUDADevice,
cache,
x,
A,
b,
)
endthe kernel does run. So, it seems related to dispatch. |
|
Looking at the types, I see that this passes: @testset "types" begin
@test column_A(xs[1], 1, 1, 1) isa Fields.ColumnField
@test column_A(As[1], 1, 1, 1) isa Fields.ColumnField
@test column_A(bs[1], 1, 1, 1) isa Fields.ColumnField
@test column_A(xs[2], 1, 1, 1) isa Fields.ColumnField
@test column_A(As[2], 1, 1, 1) isa Fields.ColumnField
@test column_A(bs[2], 1, 1, 1) isa Fields.ColumnField
@test column_A(xs[3], 1, 1, 1) isa Fields.PointDataField
@test column_A(As[3], 1, 1, 1) isa Fields.PointDataField
@test column_A(bs[3], 1, 1, 1) isa Fields.PointDataField
@test column_A(xs[4], 1, 1, 1) isa Fields.ColumnField
@test column_A(As[4], 1, 1, 1) isa UniformScaling
@test column_A(bs[4], 1, 1, 1) isa Fields.ColumnField
@test column_A(xs[5], 1, 1, 1) isa Fields.ColumnField
@test column_A(As[5], 1, 1, 1) isa UniformScaling
@test column_A(bs[5], 1, 1, 1) isa Fields.ColumnField
@test column_A(xs[6], 1, 1, 1) isa Fields.PointDataField
@test column_A(As[6], 1, 1, 1) isa UniformScaling
@test column_A(bs[6], 1, 1, 1) isa Fields.PointDataField
end |
|
So, it seems like we need to define function _single_field_solve!(
::ClimaComms.CUDADevice,
cache::Fields.PointDataField,
x::Fields.PointDataField,
A::Fields.PointDataField,
b::Fields.PointDataField,
)
end |
This was referenced Oct 2, 2024
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
#1992 removed the
uniform_vertical_levelsfunction. This led to a GPU error downstream in ClimaLand (see error message below). This PR reverts that change and makes a patch release - this has been tested from ClimaLand (see run here).The error came up in ClimaLand in a PR switching the canopy temperature variable from being stepped explicitly to implicitly. We also step soil water content implicitly, but there has been no error with that case. In the soil-only case, our implicit variable is defined on a
ExtrudedFiniteDifferenceSpace. In the new combined soil/canopy case, we have one implicit variable on aExtrudedFiniteDifferenceSpace, and the other on aSpectralElementSpace2D. Maybe the bug comes from having two variables on two different spaces, or from theSpectralElementSpace2Dspecifically?Here is the error (e.g. from this failing run):