Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-14781

Upgrade Eigen to 3.2.10

    XMLWordPrintable

    Details

    • Type: Improvement
    • Status: Done
    • Resolution: Done
    • Fix Version/s: None
    • Component/s: eigen

      Description

      As part of DM-14305 upgrading Eigen to 3.3 we need to transition to a version new enough that pybind11 supports it. The earliest version of Eigen that our pybind11 supports is 3.2.7. This intermediate step is required in order to release a version of ndarray that no longer uses EigenView for pybind11, and thus allows us to switch to the standard pybind11 wrappers.

      I will go to 3.2.10, the latest 3.2 release, in order to get as many bug fixes and other improvements as we can.

        Attachments

          Issue Links

            Activity

            Hide
            rowen Russell Owen added a comment -

            Note that a few values changed by a tiny amount in the lsst_dm_stack_demo (details below). I assume this is due to bug fixes in Eigen, but in any case the changes are small enough not to worry.

            https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/28055/pipeline

            On linux 2 values changed:

            Failed (absolute difference 1.00001e-15, relative difference 2.604e-12 over tolerance 0) in column coord_dec.
            Failed (absolute difference 5.00001e-17, relative difference 4.04846e-11 over tolerance 0) in column coord_dec.
            

            On macOS 4 values changed:

            Failed (absolute difference 1.00001e-16, relative difference 4.49102e-12 over tolerance 0) in column coord_dec.
            Failed (absolute difference 1.00001e-15, relative difference 2.604e-12 over tolerance 0) in column coord_dec.
            Failed (absolute difference 1.99999e-17, relative difference 1.61937e-11 over tolerance 0) in column coord_dec.
            Failed (absolute difference 9.99981e-15, relative difference 1.59698e-12 over tolerance 0) in column coord_dec.
            

            Show
            rowen Russell Owen added a comment - Note that a few values changed by a tiny amount in the lsst_dm_stack_demo (details below). I assume this is due to bug fixes in Eigen, but in any case the changes are small enough not to worry. https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/28055/pipeline On linux 2 values changed: Failed (absolute difference 1.00001e-15, relative difference 2.604e-12 over tolerance 0) in column coord_dec. Failed (absolute difference 5.00001e-17, relative difference 4.04846e-11 over tolerance 0) in column coord_dec. On macOS 4 values changed: Failed (absolute difference 1.00001e-16, relative difference 4.49102e-12 over tolerance 0) in column coord_dec. Failed (absolute difference 1.00001e-15, relative difference 2.604e-12 over tolerance 0) in column coord_dec. Failed (absolute difference 1.99999e-17, relative difference 1.61937e-11 over tolerance 0) in column coord_dec. Failed (absolute difference 9.99981e-15, relative difference 1.59698e-12 over tolerance 0) in column coord_dec.
            Hide
            Parejkoj John Parejko added a comment -

            Given that we don't have a ground truth for the "correct" values in the stack demo, I'd take a change that small to mean the same thing you did: Eigen fixed some things.

            We should really have a new stack demo built on lsstSim images where we know "truth" and can compare directly to it. I wonder if that'll ever happen?

            Show
            Parejkoj John Parejko added a comment - Given that we don't have a ground truth for the "correct" values in the stack demo, I'd take a change that small to mean the same thing you did: Eigen fixed some things. We should really have a new stack demo built on lsstSim images where we know "truth" and can compare directly to it. I wonder if that'll ever happen?

              People

              Assignee:
              rowen Russell Owen
              Reporter:
              rowen Russell Owen
              Reviewers:
              John Parejko
              Watchers:
              Jim Bosch, John Parejko, Russell Owen
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved:

                  Jenkins

                  No builds found.