Sorry for the delay, last week we were at OGF23 so I've been playing "catch-up" all week. For those interested in my take on the state of the grid, check out my report of the event

Activities

• Attending OGF23
  


• Linking mpitrace with WRF
  


• Applying PAPI kernel patch to GCB kernel for mpitrace
  

Accomplishments

Amon and aprof were modified to compile on MareNostrum, although we still have not tested their functionality.

Problems

• A function name conflict was found between WRF and mpitrace. I tried to fix this by changing the name in WRF, but was not successful due to the intricacies of the WRF codebase (code generation, for one). German Llort (the tracing utilities developer) solved this problem by modifying mpitrace. This led to a different problem, so now we are trying to use the LD_PRELOAD method, which does not require the mpitrace libraries to be linked in at compile time.
  
  

  
  



• There is a problem compiling the kernel of GCB, which is preventing us from being able to patch it. I am working with David Villegas (the GCB administrator) to solve this problem.
  
  

  
  

Plans

• Successfully patch GCB kernel with PAPI
  


• Get mpitrace working with WRF on GCB
  


• Test amon and aprof on MareNostrum. First with basic executables and then with WRF