https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/dss-reorder/m-p/1687983#M37118 <P>I am realizing that this step is not only reordering but also symbolic factorization, which of course is costly. This is not clear (to me) in the documentation but is pretty clear when you look at the call sequence. No need to reply.</P> Wed, 07 May 2025 01:15:53 GMT Js23 2025-05-07T01:15:53Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/dss-reorder/m-p/1687979#M37117 <P>I have a question about dss_reorder.</P><P>When you are using dss, it looks like dss_reorder needs to be called, even if you want to use natural ordering.&nbsp; (That is, feeding it a permutation vector that is just 1 to nrows.) When I omit it, I get "MKL-DSS-DSS-Error, Invalid call sequence" from dss_factor_real.</P><P>At the same time, it's surprisingly expensive to run -- in my case the timing is about the same whether I run it with a full permutation vector or MKL_DSS_DEFAULTS.</P><P>Is there a way around this? And would it be possible to help me understand what work dss_reorder is doing in this case? I'd like this step to do as little work as possible.</P><P>&nbsp;</P><P>Thanks.</P> Wed, 07 May 2025 01:01:09 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/dss-reorder/m-p/1687979#M37117 Js23 2025-05-07T01:01:09Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/dss-reorder/m-p/1687983#M37118 <P>I am realizing that this step is not only reordering but also symbolic factorization, which of course is costly. This is not clear (to me) in the documentation but is pretty clear when you look at the call sequence. No need to reply.</P> Wed, 07 May 2025 01:15:53 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/dss-reorder/m-p/1687983#M37118 Js23 2025-05-07T01:15:53Z