Internal:Metric Adaptive CDT3D: Difference between revisions

From crtc.cs.odu.edu
Jump to navigation Jump to search
← Older editNewer edit →
Ctsolakis (talk | contribs)
BrainMapping, CNF, HPC, I2MConversion, NDAM, Segmentation, UltraSound3D, Bureaucrats, external, internal, Administrators
2,152 edits
Ctsolakis (talk | contribs)
BrainMapping, CNF, HPC, I2MConversion, NDAM, Segmentation, UltraSound3D, Bureaucrats, external, internal, Administrators
2,152 edits
Line 68: Line 68:
* Using a scaled unit cube ([0,1]<sup>3</sup>) and Polar-1 metric  from [https://github.com/UGAWG/adapt-benchmarks/tree/master/cube-cylinder#polar-1 UGAWG]
* Using a scaled unit cube ([0,1]<sup>3</sup>) and Polar-1 metric  from [https://github.com/UGAWG/adapt-benchmarks/tree/master/cube-cylinder#polar-1 UGAWG]
* No boundary adaptation,due to the increasing prepossessing time and the difficulty to recover the boundary.
* No boundary adaptation,due to the increasing prepossessing time and the difficulty to recover the boundary.
* Grouping (that enhances significantly performance) is disabled due to Delaunay criterion issues.
* <code> mrecm2 </code> needs to be disabled due to problems with inpshere falling in infinite loop
* <code> mrec4 </code> needs to be on or highly anisotropic elements will not be refined ( this is due to the double flag bug of CDT3D)
* The experiments performed on a DELL workstation with Linux Red Hat Enterprise, 24 hardware cores Intel® Xeon® CPU E5-2697v2@2.70 GHz, and 757 GB RAM
* The experiments performed on a DELL workstation with Linux Red Hat Enterprise, 24 hardware cores Intel® Xeon® CPU E5-2697v2@2.70 GHz, and 757 GB RAM
* Using 24+24 (HyperThreading) threads
* Using 24+24 (HyperThreading) threads

Revision as of 16:07, 10 May 2018

Preliminary data

All time data are given in seconds.
# Tetrahedra # Vertices Isotropic Boundary

Refinement (MMGS)

Boundary

Recovery

Mesh Refinement Total Time

(without MMGS)

Memory (GB)
# Iterations Par. Deactivation Par. Vertex Creation Vertex Insertion Partition Par. Reconnection
case 1 (x8) 14,094,106 2,390,948 3.04 4.93 78 7.42 12.38 5.62 82.91 238.17 352.66 6.1
case 2 (x10) 26,859,966 4,546,172 3.78 11.64 80 14.69 25.17 9.85 154.10 437.97 655.05 10.4
case 3 (x20) 209,865,483 35,344,507 25.64 169.53 276 311.71 381.31 74.69 3,781.33 9,321.10 14,045.37 74.31

Notes

  • Using a scaled unit cube ([0,1]3) and Polar-1 metric from UGAWG
  • No boundary adaptation,due to the increasing prepossessing time and the difficulty to recover the boundary.
  • mrecm2 needs to be disabled due to problems with inpshere falling in infinite loop
  • mrec4 needs to be on or highly anisotropic elements will not be refined ( this is due to the double flag bug of CDT3D)
  • The experiments performed on a DELL workstation with Linux Red Hat Enterprise, 24 hardware cores Intel® Xeon® CPU E5-2697v2@2.70 GHz, and 757 GB RAM
  • Using 24+24 (HyperThreading) threads
  • parameters --met=1 --meval=2 --mrecm2=1 --mrec4=1 --nadapt=0 --mpp=2 --qstatr=1 --outmesh=${NAME}.meshb --ndfs=0 --nthreads=24 --nbuckets=640 --nthreadsaf=24 --nbucketsaf=640 --blockmul=500000 --nqual=0 --nminpts=30 --ordt=1 --pardeact=1 --mdse=0
  • raw data is under crtc-gpu.cs.odu.edu:/research/ctsolakis/QCIRCLE

TO DO (shorterm)

  • Try enabling sorting ( sortp)
  • export quality to tecplot format

TO DO (long term)

Retrieved from "https://crtc.cs.odu.edu/index.php?title=Internal:Metric_Adaptive_CDT3D&oldid=1195"