Latest revision as of 06:13, 29 March 2018

Decoupled Approach for CDT3D

Take advantage of CDT3D performance and mesh quality capability and create a ‘coarse’ mesh
Decompose the mesh into N subdomains (N > P)
Distribute and refine each subdomain in parallel with no communication.

Due to the absence of appropriate software simple heuristics have been employed

Initial Mesh: 880k Tetrahedra; Final Mesh 150M Tetrahedra ; Number of Subdomains 640
empty	User Handler threads	Total Time	Subdomain Migrations
Pthreads (32)	-	3288	-
PREMA 64 processes, 128 cores without ILB	1	579	0
PREMA 64 processes, 128 cores with ILB.	1	333	123
PREMA 42 processes, 126 cores with ILB.	2	263	215
PREMA 25 processes, 125 cores with ILB.	3	197	136

@@ Line 10: / Line 10: @@
 |[[Image:Decoupled_partitioned.png|400px]]
 |[[Image:Decoupled_partitioned2.png|400px]]
+|}
+== Preliminary Results (May 2017) ==
+{| class="wikitable" style="margin-left: auto; margin-right: auto; border: none;"
+|+ Initial Mesh: 880k Tetrahedra; Final Mesh 150M Tetrahedra ; Number of Subdomains 640
+|-
+! empty || User Handler threads || Total Time || Subdomain Migrations
+|-
+|Pthreads (32)                                   || - ||    3288    || -
+|-
+|PREMA<br> 64 processes, 128 cores without ILB || 1 || 579 || 0
+|-
+|PREMA<br> 64 processes, 128 cores with ILB. || 1 || 333 || 123
+|-
+|PREMA<br> 42 processes, 126 cores with ILB. || 2 || 263 || 215
+|-
+|PREMA<br> 25 processes, 125 cores with ILB. || 3 || 197 || 136
 |}