h5part AT lists.psi.ch
Subject: H5Part development and discussion
- From: Mark Howison <MHowison AT lbl.gov>
- To: h5part AT lists.psi.ch
- Subject: [H5part] H5hut
- Date: Thu, 1 Jul 2010 10:50:42 -0700
- List-archive: <https://lists.web.psi.ch/pipermail/h5part/>
- List-id: H5Part development and discussion <h5part.lists.psi.ch>
I'm not sure who subscribes to this list, but I would like to announce
that there is currently a project underway to refactor the H5Part
codebase into a new library called H5hut (HDF5 Utility Toolkit), which
will subsume the current H5Part and H5Block modules and add a new
module for finite element meshes called H5Fed that is under
development by Achim and Benedikt.
I have been introducing a number of performance-related changes to
H5Part over the last year or two, and am in the process of merging
those from the H5Part/1.6 branch into the H5hut trunk.
The major design difference for H5hut is that it pools functionality
that is "core" to all the modules into a common h5core module. Achim
refactored the H5Part module to use this new architecture, and I
recently merged in the modifications to H5Part from the LBL side. I am
currently refactoring H5Block to make use of h5core.
Eventually, I also plan to merge a fourth module called H5MultiBlock
that I wrote to more easily handle regular 3D grids with halo regions
that need to be exchanged via MPI. I will also be merging in a
regression test I wrote for H5Part, and plan to write a corresponding
test for H5Block.
From now on, I will start posting our conversations about design
decisions to this list, in case there are users who would like to have
input on our choices.
Lawrence Berkeley National Lab
- [H5part] H5hut, Mark Howison, 07/01/2010
Archive powered by MHonArc 2.6.19.