| Thread Previous • Date Previous • Date Next • Thread Next |
DOLFIN wrote:
One or more new changesets pushed to the primary DOLFIN repository.
A short summary of the last three changesets is included below.
changeset: 2758:533a363c904a8da2d301ec1fad8f021033dba188
tag: tip
user: "Anders Logg <logg@xxxxxxxxx>"
date: Fri Mar 16 11:13:05 2007 +0100
files: src/sandbox/assembly/main.cpp
description:
Test new assembly with old matrix (which now inherits from GenericTensor).
Seems to work, but is slower (at least uBlas without precomputing sparsity)
than AssemblyMatrix-based assembly:
Old assembly, DOLFIN Matrix: 14.5
New assembly, DOLFIN Matrix: 0.94
New assembly, AssemblyMatrix: 0.54
The apply() function to finalise assembly was not being called by the Assembler, so the "New assembly, DOLFIN Matrix" case will be a bit slower than reported above.
I just did some tests with precomputation of the sparsity pattern, and the difference between a DOFLIN Matrix (uBLAS) and an AssemblyMatrix was smaller (about 30%). Timings for both include the generation of the sparsity pattern. uBlasMatrix is not being initialised in the cleverest way, so there is still room for improvement.
Garth
changeset: 2757:ccf82b8253c92f9c723f6976af6a98be6731b597 user: "Anders Logg <logg@xxxxxxxxx>" date: Fri Mar 16 11:06:08 2007 +0100 files: src/kernel/la/dolfin/AssemblyMatrix.h description: Fix bug in AssemblyMatrix.h (missing ]) changeset: 2756:05032167c3be0a8db4ee6782aea2aa8fe6a6021d user: "Anders Logg <logg@xxxxxxxxx>" date: Thu Mar 15 22:28:39 2007 +0100 files: src/kernel/la/dolfin/Scalar.h description: Add class Scalar implementing GenericTensor interface for scalars ---------------------------------------------------------------------- For more details, visit http://www.fenics.org/hg/dolfin _______________________________________________ DOLFIN-dev mailing list DOLFIN-dev@xxxxxxxxxx http://www.fenics.org/mailman/listinfo/dolfin-dev
| Thread Previous • Date Previous • Date Next • Thread Next |
