Internal:ACDT3D Scitech2018: Difference between revisions

From crtc.cs.odu.edu
Jump to navigation Jump to search
← Older edit
Ctsolakis (talk | contribs)
BrainMapping, CNF, HPC, I2MConversion, NDAM, Segmentation, UltraSound3D, Bureaucrats, external, internal, Administrators
2,152 edits
add 1M complexity data
Ctsolakis (talk | contribs)
BrainMapping, CNF, HPC, I2MConversion, NDAM, Segmentation, UltraSound3D, Bureaucrats, external, internal, Administrators
2,152 edits
 
(43 intermediate revisions by the same user not shown)
Line 1: Line 1:
= Performance Results=
==First Method ==
[[File:cube_strong_scaling.png|600px|case: cube polar-2 metric going from 8K complexity to 500k]]
[[File:Cube polar log log zoom.png|600px]]
'''Right: cube polar-2 metric going from 8K complexity to 500k, Left : Zoom-in for up to 16 cores'''
[[File:delta-wing_strong_scaling.png|600px| case: delta wing going from 5K complexity to 500k]]
[[File:Delta log log zoom.png|600px]]
'''Right: delta wing going from 5K complexity to 500k, Left Zoom in for up to 16 cores'''
[[ Internal:Scitech2018_EPIC_performace | EPIC Data ]]
[[ Internal:Scitech2018_refine_performace | refine Data ]]
== Second Method ==
Starting from a 50K Complexity volume mesh
Results include no boundary adaptation
Results include no boundary adaptation


Size of intial mesh
Size of initial mesh


Vertices : 104,305
Vertices : 104,305
Line 7: Line 27:
Tetrahedra: 581,196
Tetrahedra: 581,196


=== 1M Complexity ( 100M Tetrahedra) ===
=== 1M Complexity (~12M Tetrahedra) ===
 
{| class="wikitable"
  |+ crtc-gpu node
  ! Cores
  ! T<sub>Prep</sub>
  ! T<sub>E2E</sub>
  ! N<sub>e</sub>
  ! N<sub>p</sub>
  ! Speed<sub>e</sub> (elms/sec)
  ! Speed<sub>p</sub> (points/sec)
  ! Specs
|-
| 24
| 2.38
| 88.31
| 11,122,131
| 1,864,227 
| 115,466.48
| 19,278.37
| [1]
|-
| 24 +12HT 
| 2.27
| 82.62
| 11,125,744
| 1,864,711
| 124,214.25
| 20,737.50
| [1]
|-
| 24 + 24HT
| 2.15
| 72.73
| 11,130,150
| 1,865,647
| 140,878.13
| 23,522.20
| [1]
|}
 
{| class="wikitable"
  |+  turing coreV2-himem node
  ! Cores
  ! T<sub>Prep</sub>
  ! T<sub>E2E</sub>
  ! N<sub>e</sub>
  ! N<sub>p</sub>
  ! Speed<sub>e</sub> (elms/sec)
  ! Speed<sub>p</sub> (points/sec)
  ! Specs
|-
| 24
| 2.64
| 110.16
| 11,119,089
| 1,863,716
| 93,421.04
| 15,597.62
| [2]
|-
| 32
| 2.60
| 91.44
| 11,114,943
| 1,862,960
| 115,198.46
| 19,232.88
| [2]
|}
 
{| class="wikitable"
  |+  bridges.psc
  ! Cores
  ! T<sub>Prep</sub>
  ! T<sub>E2E</sub>
  ! N<sub>e</sub>
  ! N<sub>p</sub>
  ! Speed<sub>e</sub> (elms/sec)
  ! Speed<sub>p</sub> (points/sec)
  ! Specs
|-
| 24
| 2.05
| 229.55
| 11,128,838
| 1,865,236
| 45,542.50
| 7,603.33
| [3]
|}
 
 
 
more data here : [[Internal:ACDT3D#1M_Complexity_.28.7E12M_Tetrahedra.29 | here]]
 
=== 10M Complexity (~120M Tetrahedra) ===


{| class="wikitable"
{| class="wikitable"
Line 19: Line 135:
   ! N<sub>p</sub>
   ! N<sub>p</sub>
   ! Speed<sub>e</sub> (elms/sec)
   ! Speed<sub>e</sub> (elms/sec)
   ! Speed<sub>p</sub> (elms/sec)
   ! Speed<sub>p</sub> (points/sec)
   ! Specs
   ! Specs
|-
|-
Line 25: Line 141:
   | 24 + 24HT  
   | 24 + 24HT  
   | 2.41
   | 2.41
   | 1.277.66
   | 1,277.66
   | 109,884,436   
   | 109,884,436   
   |  18,365,004  
   |  18,365,004  
Line 32: Line 148:
   | [1]
   | [1]
|}
|}
more data here : [[Internal:ACDT3D#10M_Complexity_.28.7E120M_Tetrahedra.29 | here]]
=== 100M Complexity (~1.2B Tetrahedra) ===
{| class="wikitable"
  |+
|-
  ! software-name
  ! Cores
  ! T<sub>Prep</sub>
  ! T<sub>E2E</sub>
  ! N<sub>e</sub>
  ! N<sub>p</sub>
  ! Speed<sub>e</sub> (elms/sec)
  ! Speed<sub>p</sub> (points/sec)
  ! Specs
|-
  | CDT3D
  | 24 + 24HT
  | 2.90
  | 21,572.60
  | 1,096,524,832
  | 183,097,400
  | 50,795.75
  |  8,481.52   
  | [1]
|}
more data here : [[Internal:ACDT3D#100M_Complexity_.28.7E1.2B_Tetrahedra.29 | here]]


[1] 24 hardware cores Intel® Xeon® CPU E5-2697v2@2.70 GHz, 757 GB RAM and Hyper-threading Support
[1] 24 hardware cores Intel® Xeon® CPU E5-2697v2@2.70 GHz, 757 GB RAM and Hyper-threading Support
[2] 32 hardware cores Intel® Xeon® E5-4610 v2 @2.30GHz, 757 GB RAM
[3] 64 hardware cores Intel® Xeon® E7-8860 v3 @ 2.20GHz 4TB RAM
== Third Method ==
[[Internal:ACDT3D_Scitech2018_scalable_complexity#Scalable_complexity | Scalable complexity]]
[[Internal:ACDT3D_Scitech2018_scalable_complexity#Scalable_complexity_.28In-memory.29 | Scalable Complexity (in-memory) ]]
{| class="wikitable"
! # Threads
! Initial #Tetrahedra
! Inital #Vertices
! # Tetrahedra
! # Vertices
! Reconstruct Mesh
! Par. Reconnection
! Total Time
! Tetrahedra/sec
! Vertices/sec
! Memory (GB)
|-
| 1
| 581,196
| 104,305
| 5,516,567
| 926,766
| 2.26
| 673.84
| 719.59
| 6,858.59
| 1,146.56
| 1.61
|-
| 3
| 5,516,567
| 926,766
| 15,908,630
| 2,669,073
| 0.18
| 602.93
| 666.35
| 15,595.50
| 2,615.41
| 4.26
|-
| 6
| 15,908,630
| 2,669,073
| 34,169,056
| 5,712,669
| 0.48
| 499.37
| 586.14
| 31,153.69
| 5,196.87
| 11.61
|-
| 12
| 34,169,056
| 5,712,669
| 69,272,450
| 11,571,320
| 0.51
| 513.81
| 648.17
| 54,157.70
| 9,045.87
| 24.13
|-
| 24
| 69,272,450
| 11,571,320
| 138,892,161
| 23,186,740
| 0.97
| 594.65
| 859.06
| 81,041.73
| 13,536.37
| 49.48
|}
= Qualitative Results =
[[Internal:Scitech2018_EPIC_stability | EPIC]]
[[Internal:Scitech2018_CDT3D_stability | CDT3D ]]
[[Internal:Scitech2018_Quality_Comparison | Quality Comparison ]]
=Paper =
public [https://www.overleaf.com/15174959jmkkqdctrvyb https://www.overleaf.com/15174959jmkkqdctrvyb]
private  [https://www.overleaf.com/16946989zckxgznmhwrq https://www.overleaf.com/16946989zckxgznmhwrq]
previous UGAWG papers [https://odu.app.box.com/folder/46906874476 https://odu.app.box.com/folder/46906874476]

Latest revision as of 15:25, 11 October 2018

Performance Results

First Method

case: cube polar-2 metric going from 8K complexity to 500k

Right: cube polar-2 metric going from 8K complexity to 500k, Left : Zoom-in for up to 16 cores

case: delta wing going from 5K complexity to 500k

Right: delta wing going from 5K complexity to 500k, Left Zoom in for up to 16 cores

EPIC Data

refine Data

Second Method

Starting from a 50K Complexity volume mesh

Results include no boundary adaptation

Size of initial mesh

Vertices : 104,305

Tetrahedra: 581,196

1M Complexity (~12M Tetrahedra)

crtc-gpu node
Cores TPrep TE2E Ne Np Speede (elms/sec) Speedp (points/sec) Specs
24 2.38 88.31 11,122,131 1,864,227 115,466.48 19,278.37 [1]
24 +12HT 2.27 82.62 11,125,744 1,864,711 124,214.25 20,737.50 [1]
24 + 24HT 2.15 72.73 11,130,150 1,865,647 140,878.13 23,522.20 [1]
turing coreV2-himem node
Cores TPrep TE2E Ne Np Speede (elms/sec) Speedp (points/sec) Specs
24 2.64 110.16 11,119,089 1,863,716 93,421.04 15,597.62 [2]
32 2.60 91.44 11,114,943 1,862,960 115,198.46 19,232.88 [2]
bridges.psc
Cores TPrep TE2E Ne Np Speede (elms/sec) Speedp (points/sec) Specs
24 2.05 229.55 11,128,838 1,865,236 45,542.50 7,603.33 [3]


more data here : here

10M Complexity (~120M Tetrahedra)

software-name Cores TPrep TE2E Ne Np Speede (elms/sec) Speedp (points/sec) Specs
CDT3D 24 + 24HT 2.41 1,277.66 109,884,436 18,365,004 85,388.49 14,265.39 [1]

more data here : here

100M Complexity (~1.2B Tetrahedra)

software-name Cores TPrep TE2E Ne Np Speede (elms/sec) Speedp (points/sec) Specs
CDT3D 24 + 24HT 2.90 21,572.60 1,096,524,832 183,097,400 50,795.75 8,481.52 [1]

more data here : here

[1] 24 hardware cores Intel® Xeon® CPU E5-2697v2@2.70 GHz, 757 GB RAM and Hyper-threading Support

[2] 32 hardware cores Intel® Xeon® E5-4610 v2 @2.30GHz, 757 GB RAM

[3] 64 hardware cores Intel® Xeon® E7-8860 v3 @ 2.20GHz 4TB RAM

Third Method

Scalable complexity

Scalable Complexity (in-memory)

# Threads Initial #Tetrahedra Inital #Vertices # Tetrahedra # Vertices Reconstruct Mesh Par. Reconnection Total Time Tetrahedra/sec Vertices/sec Memory (GB)
1 581,196 104,305 5,516,567 926,766 2.26 673.84 719.59 6,858.59 1,146.56 1.61
3 5,516,567 926,766 15,908,630 2,669,073 0.18 602.93 666.35 15,595.50 2,615.41 4.26
6 15,908,630 2,669,073 34,169,056 5,712,669 0.48 499.37 586.14 31,153.69 5,196.87 11.61
12 34,169,056 5,712,669 69,272,450 11,571,320 0.51 513.81 648.17 54,157.70 9,045.87 24.13
24 69,272,450 11,571,320 138,892,161 23,186,740 0.97 594.65 859.06 81,041.73 13,536.37 49.48

Qualitative Results

EPIC

CDT3D

Quality Comparison

Paper

public https://www.overleaf.com/15174959jmkkqdctrvyb

private https://www.overleaf.com/16946989zckxgznmhwrq

previous UGAWG papers https://odu.app.box.com/folder/46906874476

Retrieved from "https://crtc.cs.odu.edu/index.php?title=Internal:ACDT3D_Scitech2018&oldid=2523"