Page 7 of 7

Choice of matrix library and interface

Posted: December 20th, 2011, 4:03 pm
by karpeev
Ah, of course! I'm more used to it going by FD.It shouldn't be hard to do, but let me take a closer look at it.I imagine that, given my other obligations, some time before the New Year is likely,but possibly sooner.Dmitry.

Choice of matrix library and interface

Posted: January 9th, 2012, 3:54 pm
by Cuchulainn
I propose Boost UBLAS in the short to medium term. I think it is the easiest to use.All those in favour say 'Aye'All those against say 'Nay'(*)(*) but then propose a working alternative. So, arrividerci AAMatrix, BBMatrix, DDMatrix, etc. etc.

Choice of matrix library and interface

Posted: February 4th, 2012, 7:20 pm
by Polter
One more choice: http://viennacl.sourceforge.net/viennac ... mlQuoteThe Vienna Computing Library (ViennaCL) is a scientific computing library written in C++ and based on OpenCL. It allows simple, high-level access to the vast computing resources available on parallel architectures such as GPUs and is primarily focused on common linear algebra operations (BLAS levels 1, 2 and 3) and the solution of large systems of equations by means of iterative methods with optional preconditioner.Some features of interest:QuoteUses OpenCL to support GPUs from NVIDIA and AMDSupport for multi-core CPU (requires AMD APP SDK or Intel OpenCL SDK)Multi-device supportIterative solvers can also be used directly for uBLAS, Eigen and MTL4 objectsConvenient data transfer from and to STL, uBLAS, Eigen and MTL4 objectsInterface similar to Boost uBLASMIT (X11) open source license, which is not bad.

Choice of matrix library and interface

Posted: February 5th, 2012, 10:30 am
by Cuchulainn
QuoteOriginally posted by: PolterOne more choice: http://viennacl.sourceforge.net/viennac ... mlQuoteThe Vienna Computing Library (ViennaCL) is a scientific computing library written in C++ and based on OpenCL. It allows simple, high-level access to the vast computing resources available on parallel architectures such as GPUs and is primarily focused on common linear algebra operations (BLAS levels 1, 2 and 3) and the solution of large systems of equations by means of iterative methods with optional preconditioner.Some features of interest:QuoteUses OpenCL to support GPUs from NVIDIA and AMDSupport for multi-core CPU (requires AMD APP SDK or Intel OpenCL SDK)Multi-device supportIterative solvers can also be used directly for uBLAS, Eigen and MTL4 objectsConvenient data transfer from and to STL, uBLAS, Eigen and MTL4 objectsInterface similar to Boost uBLASMIT (X11) open source license, which is not bad.Interoperability looks good. On Functionality, it looks similar to uBLAS (minus the couple of matrix solvers).

Choice of matrix library and interface

Posted: February 8th, 2012, 7:05 pm
by Cuchulainn
A remark on multiarray versus uBLAS; they have different goals.. MA is for regular nd data with slices and range; you just store data; you cannot multiply 2 multiarrays.. uBLAS is different (patterned matrices, algebra).