Merge remote-tracking branch 'upstream/pr/417/head' into source

* upstream/pr/417/head:
  add note on thesis committee
  update report and bib

Signed-off-by: Robert Speck <[email protected]>

Author: pancetta

_bibliography/external/reducing_communication_sparse.bib

@@ -20,8 +20,7 @@
 %   Aaaa           first author's surname in CamelCase
 %   Bbbb           second author's surname in CamelCase
 %
-% - for single-author publications:   AaaaYYYY
-% - for two-authors publications:     AaaaBbbbYYYY
+% - for single-author publications:   AaaaYYYY % - for two-authors publications:     AaaaBbbbYYYY
 % - for three or more authors:        AaaaEtAlYYYY
 %
 % In case an author has published more than one publication in the same year, append lower case
@@ -68,6 +67,14 @@ @article{BienzEtAl2017
   author = {Amanda Bienz and William D. Gropp and Luke N. Olson},
   journal = {SIAM Journal on Scientific Computing},
   notes = {submitted},
-  title = {TAPSpMV: Topology-Aware Parallel Sparse Matrix Vector Multiplication},
+  title = {NAPSpMV: Topology-Aware Parallel Sparse Matrix Vector Multiplication},
   year = {2017},
 }
+
+@misc{BiOl2017,
+      author = {Bienz, Amanda and Olson, Luke N.},
+      title = {{RAPtor}: parallel algebraic multigrid v0.1},
+      year = {2017},
+      url = {https://github.com/lukeolson/raptor},
+      note = {Release 0.1}
+      }

_projects/reducing_communication_sparse.md

@@ -1,7 +1,7 @@
 ---
 layout: page_project
 title: Reducing Communication in Sparse Iterative and Direct Solvers
-date: 2016-03-23
+date: 2018-03-01
 updated:
 navbar: Research
 subnavbar: Projects
@@ -11,7 +11,9 @@ status: running
 topics:
   - numerics
 keywords:
-  -
+  - sparse matrix
+  - multigrid
+  - communication
 head: olson_l
 members:
   - bienz_a
@@ -37,13 +39,12 @@ and Luke Olson, and INRIA professor Laura Grigori. The work has highlighted
 several aspects in moving toward topology-based algorithm decisions. Here, we investigated the
 communication overhead of an algebraic multigrid method at scale, where coarse grids push the
 strong scaling limit and exhibit irregular memory access patterns. As a result, high communication
-lead to reduced efficiency. We developed an alpha-beta type communication model that attempts
+lead to reduced efficiency. We continue to develop a communication model that attempts
 to account for two aspects in the communication: message distance and message contention in the
 network. The result is a model that may help make point-wise decisions in the algorithm. For
 example, sparse entries could be strengthened or weakened depending on their communication
 burden.
 
-
 ## Results for 2015/2016
 
 Recent efforts have focused on overlapping communication with computation in the
@@ -58,6 +59,14 @@ The work was extended to aggregate communication at on the nodes and was present
 
 The next steps of the collaboration will look at adding pushing the topology aware process to sparse algorithms at INRIA.
 
+## Results for 2017/2018
+
+The work resulted in presentations at the 7th JLESC Workshop in July, 2017, along with presentation of multigrid and sparse-matrix multiplication results at SC17 and Copper Mountain Multigrid (2017) and Copper Mountain Multigrid (2018).  In addition, the main code Raptor, was released as open source.
+
+Grigori (INRIA) serves on Bienz's thesis committee; Bienz is expected to finish in 2018.
+
+The next steps include completing contention modeling.
+
 ## Visits and meetings
 
 Completed: JointLab Meeting in Barcelona, June 2015.
@@ -66,13 +75,14 @@ Completed: JointLab Meeting in Bonn, December 2015.
 
 Completed: Amanda Bienz visit to INRIA, Spring 2016 for 4 months.
 
-Planned: JointLab Meeting in Champaign, July 2017.
+Completed: JointLab Meeting in Champaign, July 2017.
+
+Planned: JointLab Meeting in Barcelona, April 2018.
 
 ## Impact and publications
 
 {% bibliography --cited --file jlesc.bib %}
 
-
 ## Future plans
 
 Going forward, we have outlined a plan to work with a broader collection of