Difference between revisions of "Product Backlog - Mei"
From SCECpedia
Jump to navigationJump to search| Line 2: | Line 2: | ||
== Project/Product Backlog == | == Project/Product Backlog == | ||
| + | #Doc how La Habra mesh is made with ucvm2mesh_mpi_layer | ||
#More crossing data for Kevin, Kevin said CCA06 plot looks more fuzzier than Scott's UCVMpython generated one | #More crossing data for Kevin, Kevin said CCA06 plot looks more fuzzier than Scott's UCVMpython generated one | ||
#Compare UCVM python CCA06 with UCVM C CCA06 | #Compare UCVM python CCA06 with UCVM C CCA06 | ||
Revision as of 20:20, 4 June 2018
Mei-Hui Su's Project/Product Backlog
Project/Product Backlog
- Doc how La Habra mesh is made with ucvm2mesh_mpi_layer
- More crossing data for Kevin, Kevin said CCA06 plot looks more fuzzier than Scott's UCVMpython generated one
- Compare UCVM python CCA06 with UCVM C CCA06
- Discuss mesh plotting Scripts with Scott Cal.
- Generate large (100M+ point) AWP mesh using ucvmc and ucvmp and compare meshes
DONE
- Generated various plots highlight CS17.3-H, https://scec.usc.edu/scecpedia/UCVMC_CS17.3-H_plots#Plots
- Registered new Central California Basin Model from Havard. CS17.3-H with ELY GTL
- built a useable UCVMC installation on Titan, /lustre/atlas/proj-shared/geo112/ucvm_18_5/
- made UCVMC statically linkable and the plugin models worked as in dynamically linked UCVMC
- reworked logic in ucvm2mesh-mpi to have a more 'feasible' compute node requirement for very large mesh like La Habra, ucvm2mesh_mpi_layer, remember to strip result directory when running at blue waters. lfs setstripe -c 4 /path/to/working/directory
- extract UCVMC cs173 info for Scott's gridpoints (off alittle but okayed by Scott)
- reworked basin_query tests in slurm on hpc
- created a pull request to official UCVM-C, fix up each models that is in github -- merged into official UCVMC's dev branch
- Ran tests tiling cencal+cca+cvms5+bbp1d, the 1d part gets surface gtl
- Tiled cencal+cca(with gtl)+cvms5(with gtl)+1d tested and verified
- GTL for CCA06- verified by generating identical plots to CyberShake_Study_18.5
- Fixed plugin/dynamic loading of modules, Scott's interleaving bug, fixes for cvmsi is also needed for the plugin module. There was the missed case for left-over points in chunking/processing code, fixed that. Also the code was not setup to take more than one dynamic library
- Setup to run my latest UCVMC on blue water - query works but plotting scripts do not work with BW's installed default pthread library(too old, conflicting with matplotlib's pyplot from Anaconda2)
- Fixed segfaults flagged from a working ucvmc installation on local mac (fixed cs173_final(), cca_final()), (mac gdb installation hints: https://gist.github.com/gravitylow/fb595186ce6068537a6e9da6d8b5b96d)
- Ran basin_query_mpi on hpc with SLURM
- Doc projection fix on scecpedia, https://scec.usc.edu/scecpedia/Final_Projection_Review_Result
- Add the fixed config file into CCA's github, also edits from below(pending moving to scec space): https://github.com/meihuisu/CCA.git
- Copy the CCA06 data on scec-01/Intensity to hypocenter, change the dependency to hypocenter in the make-data script
- adding elygtl to CCA06
- new cs173 github location(pending moving to scec space):https://github.com/meihuisu/cs173.git
- Register Scott NorCal model into UCVM - CS173, compare CS173 plots from ScottC with my version
- Doc CCA06 Crossing review, https://scec.usc.edu/scecpedia/Final_CCA06_Crossings_Result
- Update basin_query_mpi to return list of crossing per grid point, also return one crossing output file for each algorithm.
- To check projection fix, trace processing from en-jui data files to configuration files
- Cross section plots showing 0-10km depth through CCA06, showing velocity inversions, starting at one end of region at 5km spacing to other end of region. **No need, posted inversion plot in scecpedia**
