Staff Priorities - Scott: Difference between revisions

Revision as of 18:08, 28 January 2014

Refactored and tested vmesh gen code
Compared runtimes of GPU code with different size meshes - slower, more efficient on multiple-of-200 mesh
Queued CVM-H no GTL with 4 sites
Tried installing UCVM w/BBP 1D, tests failed
Tested performance of more CPUs for AWP - 50x50x50 is faster and more efficient
Updated algorithm for synth runtimes
Changed bookkeeping so that UCVM 1D model is tracked as "SCEC 1D", not HK
Added BBP 1D to workflow and scripts

@@ Line 1: / Line 1: @@
 == Yesterday ==
-*Integrated code to check AWP SGTs for NaNs
+*Refactored and tested vmesh gen code
-*Completed 1D curves
+*Compared runtimes of GPU code with different size meshes - slower, more efficient on multiple-of-200 mesh
-*Worked on Pegasus paper
+*Queued CVM-H no GTL with 4 sites
-*Worked on updating algorithm which calculates synth runtimes
+*Tried installing UCVM w/BBP 1D, tests failed
-*Single velocity model generation code takes too much memory, needs to be refactored
+*Tested performance of more CPUs for AWP - 50x50x50 is faster and more efficient
+*Updated algorithm for synth runtimes
+*Changed bookkeeping so that UCVM 1D model is tracked as "SCEC 1D", not HK
+*Added BBP 1D to workflow and scripts
 == Today ==
-*Refactor vmesh gen code
-*Compare runtimes of GPU code with different size meshes
 *Try 7x PP test on BW
-*Run CVM-H no GTL with 4 sites
+*Check on curves with CVM-H, no GTL
 *Queue hazard curves with GPU, CPU code and CVM-S4.26
 *Continue preparing readiness review
-*Test performance of 8K CPUs for AWP
+*Generate BBP 1D curves
-*Finish updating synth runtimes
-*Change bookkeeping so that UCVM 1D model is tracked as "SCEC 1D", not HK
 == Blocked ==
@@ Line 21: / Line 20: @@
 == Follow-ups ==
-*No
+*Summer School call
+*High F call
+*CyberShake calls
 == Areas of Responsibilities ==