|
|
Glas :Re: [glas] dense and sparse vectors |
From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2005-08-07 15:32:31
C J Kenneth Tan -- OptimaNumerics wrote:
> The benchmarks of dense daxpy and dgemm, running against system
> vendors' implementations, will be interesting to see.
>
> More interesting would be higher level code, such as dposv, dppsv,
> dgesv, dgeev, dgesvd, dsyev, dgels, or their equivalents. While daxpy
> and dgemm code are relatively straight forward, tuning these higher
> level code is often non-trivial, especially if the code is to be
> platform independent.
I was mainly planning to benchmark 'sparse axpy' of different libraries
(I'm currently using mtl-v2.1.22, oski-1.0 and ublas). I indeed
mentioned 'dense' also but mainly because I also plan to benchmark the
sparse axpy against the dense versions.
As for having a 'dense axpy' implementation in glas itself: Glas should
probably not try to outperform vendor-blas axpy's: the vendor-optimised
axpy's are already there, hard to compete with although not generic.
That's also we stated in the requirements of glas to be able to use 3rd
party backends (http://glas.sourceforge.net/doc/requirements.html).
toon