Fix #963: PETSc 3.20.0 broke the world #982

guyer · 2023-12-01T16:50:05Z

Fixes #963
(Actually, PETSC 3.19.0 broke the world.)

This PR:

Assembles before multTranspose to prevent newly added exception
Renames bandwidth to nonZerosPerRow and removes sizeHint

The two were confusingly redundant:
- PySparse takes sizeHint, the number of non-zeros in the matrix.
- PyTrilinos takes NumEntriesPerRow.
- petsc4py didn't used to be clear what it took, but is now
  documented as number of non-zeros per row (of the local portion
  of the matrix, but we'll ignore that part).
- scipy doesn't preallocate.
- Linear algebra
  defines bandwidth
  as "number $k$ such that $a_{i,j}=0$ if $|i-j| > k$", which is
  roughly half the number of non-zeros per row (and only applies
  to a band-diagonal matrix).
  Better to be explicit about what we really mean.
Now all take same parameter and PySparse adjusts as needed.

sizeHint was introduced in @a15d696 (in 2006!) to
"allow smaller preallocations", but it was never used that way.
Now, nonZerosPerRow can take an array_like to specify row-by-row
preallocations, which are directly supported by PyTrilinos and petsc4py,
and can be simulated for PySparse.

Added exactNonZeros, which may have performance benefits for
PyTrilinos and petsc4py. Currently unused.
Uses Term's knowledge of own stencil to preallocate more effectively.
Still doesn't do a good job with vector equations, but that's a deeper change
(the resolution of which might help Should be possible to couple scalar and vector equations #920).

The two were confusingly redundant: - PySparse takes `sizeHint`, the number of non-zeros in the matrix. - PyTrilinos takes `NumEntriesPerRow`. - petsc4py didn't used to be clear what it took, but is now documented as number of non-zeros per row (of the local portion of the matrix, but we'll ignore that part). - scipy doesn't preallocate. - Linear algebra [defines bandwidth](https://en.wikipedia.org/wiki/Band_matrix#Bandwidth) as "number $k$ such that $a_{i,j}=0$ if $|i-j| > k$", which is roughly half the number of non-zeros per row (and only applies to a band-diagonal matrix). Better to be explicit about what we really mean. Now all take same parameter and PySparse adjusts as needed. `sizeHint` was introduced in @a15d696 (in 2006!) to "allow smaller preallocations", but it was never used that way. Now, `nonZerosPerRow` can take an array_like to specify row-by-row preallocations, which are directly supported by PyTrilinos and petsc4py, and can be simulated for PySparse. Added `exactNonZeros`, which may have performance benefits for PyTrilinos and petsc4py. Currently unused.

Use Term's knowledge of its stencil to preallocate its matrix more effectively.

guyer · 2023-12-04T15:57:48Z

The one failure is due to an unrelated intermittent race condition

guyer · 2023-12-04T16:32:37Z

Now, the one failure is due to cosmic rays ☢️ 😠

wd15 · 2023-12-04T17:10:06Z

.azure/pipelines.yml

@@ -258,7 +258,7 @@ stages:
        condition: startsWith(variables.image, 'windows')

      - bash: |
-          mamba create --quiet --name wheelEnvironment --channel conda-forge python=3.10 mamba
+          conda create --quiet --name wheelEnvironment --channel conda-forge python=3.10


Why back to conda?

Because conda now uses libmamba (or at least can be configured to do so, which I do)

wd15 · 2023-12-04T17:14:22Z

fipy/matrices/pysparseMatrix.py

        if matrix is None:
            tmpMatrix = spmatrix.ll_mat(1, 1, 1)
+            try:


Can you do this without a try / except? It just feels wrong here. Does the asarray call raise the TypeError?

The len(nonZerosPerRow) is what throws the TypeError. This is the most effective way to figure out if nonZerosPerRow is iterable in Python 2.7. In Py3k, you can (and I do) do an isinstance(nonZerosPerRow, collections.abc.Iterable).

Oh yeah, FiPy still needs to work with 2.7.

Hopefully not for much longer. My plan is to finish the solver harmonization I'm working on, make a release with everything, and then immediately drop PySparse and Python 2.7

tkphd

Did not test myself, but the changes look good to me.

Fixes #963 (Actually, PETSC 3.19.0 broke the world.) This PR: - Assembles before `multTranspose` to prevent newly added exception - Renames `bandwidth` to `nonZerosPerRow` and removes `sizeHint` The two were confusingly redundant: - PySparse takes `sizeHint`, the number of non-zeros in the matrix. - PyTrilinos takes `NumEntriesPerRow`. - petsc4py didn't used to be clear what it took, but is now documented as number of non-zeros per row (of the local portion of the matrix, but we'll ignore that part). - scipy doesn't preallocate. - Linear algebra [defines bandwidth](https://en.wikipedia.org/wiki/Band_matrix#Bandwidth) as "number $k$ such that $a_{i,j}=0$ if $|i-j| > k$", which is roughly half the number of non-zeros per row (and only applies to a band-diagonal matrix). Better to be explicit about what we really mean. Now all take same parameter and PySparse adjusts as needed. `sizeHint` was introduced in @a15d696 (in 2006!) to "allow smaller preallocations", but it was never used that way. Now, `nonZerosPerRow` can take an array_like to specify row-by-row preallocations, which are directly supported by PyTrilinos and petsc4py, and can be simulated for PySparse. Added `exactNonZeros`, which may have performance benefits for PyTrilinos and petsc4py. Currently unused. - Uses `Term`'s knowledge of own stencil to preallocate more effectively. Still doesn't do a good job with vector equations, but that's a deeper change (the resolution of which might help #920). - Fixes(?) conda/mamba installs - Fixes(?) race condition

guyer added 8 commits November 27, 2023 19:09

Assemble before multTranspose

c29d032

Test preallocation

d00de10

Preallocate terms

fdfc8b2

Use Term's knowledge of its stencil to preallocate its matrix more effectively.

Fix spelling

333df13

Account for ghost cells and element shape in preallocation

bdaa4b8

Fix(?) conda/mamba installs

efcc78b

Configure libmamba

0b2b295

guyer requested review from wd15 and tkphd December 4, 2023 15:58

Fix(?) race condition

c68bea8

wd15 approved these changes Dec 4, 2023

View reviewed changes

tkphd approved these changes Dec 4, 2023

View reviewed changes

guyer merged commit e1cb886 into master Dec 4, 2023
24 checks passed

guyer deleted the issue963-PETSc_3_20_0_broke_the_world branch December 4, 2023 22:54

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix #963: PETSc 3.20.0 broke the world #982

Fix #963: PETSc 3.20.0 broke the world #982

guyer commented Dec 1, 2023 •

edited

Loading

guyer commented Dec 4, 2023

guyer commented Dec 4, 2023

wd15 Dec 4, 2023

guyer Dec 4, 2023

wd15 Dec 4, 2023

guyer Dec 4, 2023

wd15 Dec 4, 2023

guyer Dec 4, 2023

tkphd left a comment

Fix #963: PETSc 3.20.0 broke the world #982

Fix #963: PETSc 3.20.0 broke the world #982

Conversation

guyer commented Dec 1, 2023 • edited Loading

guyer commented Dec 4, 2023

guyer commented Dec 4, 2023

wd15 Dec 4, 2023

Choose a reason for hiding this comment

guyer Dec 4, 2023

Choose a reason for hiding this comment

wd15 Dec 4, 2023

Choose a reason for hiding this comment

guyer Dec 4, 2023

Choose a reason for hiding this comment

wd15 Dec 4, 2023

Choose a reason for hiding this comment

guyer Dec 4, 2023

Choose a reason for hiding this comment

tkphd left a comment

Choose a reason for hiding this comment

guyer commented Dec 1, 2023 •

edited

Loading