# public jedbrown /talks

### Subversion checkout URL

You can clone with HTTPS or Subversion.

• 2 commits
• 4 files changed
• 1 contributor
Apr 03, 2012
Update font sizes to better fit on the slides 0f0cf48
Add SCA talk 0119137
163  20120403-SCA.tex
22  slides/Dohp/PMultigrid.tex
 ... ... @@ -0,0 +1,22 @@ 1 +\begin{frame}{One level of $p$-multigrid} 2 + \begin{itemize} 3 + \item Want to skip assembly on finest level (for better throughput) 4 + \item High order operators lack $h$-ellipticity 5 + \begin{itemize} 6 + \item Necessary and sufficient condition for existence of pointwise smoother 7 + \end{itemize} 8 + \item Use embedded low-order operator as smoother 9 + \begin{itemize} 10 + \item Rescaled to recover a consistent inner product 11 + \item Does not destroy symmetry for point-block Jacobi 12 + \end{itemize} 13 + \item Polynomial smoothers 14 + \begin{itemize} 15 + \item Target upper part of PBJacobi-preconditioned spectrum 16 + \item Efficient GPU implementation 17 + \end{itemize} 18 + \item Reordered incomplete factorization to couple columns'' 19 + \item Operator-dependent interpolation is more delicate 20 + \item Strict semi-coarsening requires semi-structured grid 21 + \end{itemize} 22 +\end{frame}
6  slides/GroundingLine/Steepness.tex
 ... ... @@ -1,4 +1,4 @@ 1 -\begin{frame}{Bathymetry and stickyness distribution} 1 +\begin{frame}[shrink=5]{Bathymetry and stickyness distribution} 2 2  \begin{itemize} 3 3  \item Bathymetry: 4 4  \begin{itemize} @@ -18,12 +18,12 @@ 18 18  \item[$\lambda \ll 1$] Shallow Shelf Approximation (SSA), 2D elliptic solve in map-plane 19 19  \item Hydrostatic and various hybrids, 2D or 3D elliptic solves 20 20  \end{itemize} 21 - \item<2> \alert{\large Bed slope is discontinuous and of order 1.} 21 + \item \alert{\large Bed slope is discontinuous and of order 1.} 22 22  \begin{itemize} 23 23  \item Taylor expansions no longer valid 24 24  \item Numerics require high resolution, subgrid parametrization, short time steps 25 25  \item Still get low quality results in the regions of most interest. 26 26  \end{itemize} 27 - %\item<2> \alert{\LARGE Basal sliding parameters are discontinuous.} 27 + \item \alert{\large Basal sliding parameters are discontinuous.} 28 28  \end{itemize} 29 29  \end{frame}
10  slides/PETSc/RediscretizedMultigrid.tex
 @@ -3,28 +3,38 @@ 3 3  {\scriptsize 4 4  \item \texttt{DM} manages problem data beyond purely algebraic objects 5 5  \begin{itemize} 6 +{\scriptsize 6 7  \item structured, redundant, and (less mature) unstructured implementations in PETSc 7 8  \item third-party implementations 9 +} 8 10  \end{itemize} 9 11  \item \texttt{DMCoarsen(dmfine,coarse\_comm,\&coarsedm)} to create geometric'' coarse level 10 12  \begin{itemize} 13 +{\scriptsize 11 14  \item Also \texttt{DMRefine()} for grid sequencing and convenience 12 15  \item \texttt{DMCoarsenHookAdd()} for external clients to move resolution-dependent data for rediscretization and FAS 16 +} 13 17  \end{itemize} 14 18  \item \texttt{DMCreateInterpolation(dmcoarse,dmfine,\&Interp,\&Rscale)} 15 19  \begin{itemize} 20 +{\scriptsize 16 21  \item Usually uses geometric information, can be operator-dependent 17 22  \item Can be improved subsequently, e.g. using energy-minimization from AMG 23 +} 18 24  \end{itemize} 19 25  \item \texttt{DMCreateDecomposition(dm,\&nsplits,\&splitnames,\&splits,\&dms)} 20 26  \begin{itemize} 27 +{\scriptsize 21 28  \item New/tentative API to expose split information to preconditioner and nonlinear solvers 22 29  \item Can have multiple named decompositions 30 +} 23 31  \end{itemize} 24 32  \item Resolution-dependent solver-specific callbacks use attribute caching on \texttt{DM}. 25 33  \begin{itemize} 34 +{\scriptsize 26 35  \item Managed by solvers, not visible to users unless they need exotic things (e.g. custom homogenization, reduced models) 27 36  %\item This implementation aspect is subject to change, but should not affect user interface. 37 +} 28 38  \end{itemize} 29 39  } 30 40  \end{itemize}

### No commit comments for this range

Something went wrong with that request. Please try again.