Modeling Hot Outflows and Cold Clouds with the Cholla Code
Modeling Hot Outflows and Cold Clouds with the Cholla Code
Brant Robertson (really Evan Schneider)
What is Cholla?
•
•
Cholla is a new three-dimensional fixed grid, Cartesian mesh hydrodynamics code that harnesses the power of GPUs to accelerate astrophysical simulations. Written by my PhD student at Evan Schneider (UA) for her thesis. Cholla is very fast, MPI parallelized, and shows demonstrated weak scaling to > 16000 GPUs.
Total runtime (excluding initialization) Boundary conditions CTU (all GPU functions) 100 Time (s)
•
Weak Scaling test, ~3223 / GPU
10
1 1
Schneider, E. & Robertson, B. 2015, ApJS, 217, 24
10 100 1000 MPI processes (GPUs)
10000
Simulations of Wind-Cloud Interactions
z
vwind
Cloud
y
Fixed grid resolution everywhere Long boxes with 8x cloud radii transverse dimensions Radiative cooling with photoionizing background
Wind Properties (Chevalier & Clegg 1985)
whole box
nc = 1 / cc r = 5pc 64 cells/cloud radius 512x512x2048
3.5x zoom
Schneider & Robertson (in prep)
Phase diagram
r
ba iso initial sphere
relative mass
hot wind
heating balances cooling
Schneider & Robertson (in prep)
Velocity-density diagram
relative mass
hot wind
initial sphere Schneider & Robertson (in prep)
n = 0.1 / cc
n = 0.5 / cc
Schneider & Robertson (in prep)
n = 1 / cc
Convergence?
16 cells / cloud radius
32 cells / 64 cells / cloud radius cloud radius
128 cells / cloud radius
Schneider & Robertson (in prep)
Cloud Density Distributions Affect Their Crushing Time
Klein et al. 1994 estimate works well for uniform spheres.
Turbulent clouds disrupt faster than Klein et al. 1994 estimate.
Schneider & Robertson (2015)
Cloud Density Distributions Affect Their Crushing Time
Klein et al. 1994 estimate works well for uniform spheres.
Turbulent clouds disrupt faster than Klein et al. 1994 estimate.
Schneider & Robertson (2015)
Cloud Density Distribution and Entrainment
sphere
cloud
Schneider & Robertson (in prep)
Cloud Density Distribution and Entrainment
sphere
cloud
Schneider & Robertson (in prep)
Phase Structure of Entrained Flow: Sphere vs. Cloud sphere
cloud
Schneider & Robertson (in prep)
Momentum Structure of Entrained Flow: Sphere vs. Cloud sphere
cloud
Schneider & Robertson (in prep)
Column Densities: Sphere vs. Cloud
sphere
cloud
Schneider & Robertson (in prep)
Column Densities: Sphere vs. Cloud
sphere
cloud
Schneider & Robertson (in prep)
Momentum Column Densities: Sphere vs. Cloud
wind
sphere
cloud
wind Schneider & Robertson (in prep)
Brant Robertson (really Evan Schneider)
What is Cholla?
•
•
Cholla is a new three-dimensional fixed grid, Cartesian mesh hydrodynamics code that harnesses the power of GPUs to accelerate astrophysical simulations. Written by my PhD student at Evan Schneider (UA) for her thesis. Cholla is very fast, MPI parallelized, and shows demonstrated weak scaling to > 16000 GPUs.
Total runtime (excluding initialization) Boundary conditions CTU (all GPU functions) 100 Time (s)
•
Weak Scaling test, ~3223 / GPU
10
1 1
Schneider, E. & Robertson, B. 2015, ApJS, 217, 24
10 100 1000 MPI processes (GPUs)
10000
Simulations of Wind-Cloud Interactions
z
vwind
Cloud
y
Fixed grid resolution everywhere Long boxes with 8x cloud radii transverse dimensions Radiative cooling with photoionizing background
Wind Properties (Chevalier & Clegg 1985)
whole box
nc = 1 / cc r = 5pc 64 cells/cloud radius 512x512x2048
3.5x zoom
Schneider & Robertson (in prep)
Phase diagram
r
ba iso initial sphere
relative mass
hot wind
heating balances cooling
Schneider & Robertson (in prep)
Velocity-density diagram
relative mass
hot wind
initial sphere Schneider & Robertson (in prep)
n = 0.1 / cc
n = 0.5 / cc
Schneider & Robertson (in prep)
n = 1 / cc
Convergence?
16 cells / cloud radius
32 cells / 64 cells / cloud radius cloud radius
128 cells / cloud radius
Schneider & Robertson (in prep)
Cloud Density Distributions Affect Their Crushing Time
Klein et al. 1994 estimate works well for uniform spheres.
Turbulent clouds disrupt faster than Klein et al. 1994 estimate.
Schneider & Robertson (2015)
Cloud Density Distributions Affect Their Crushing Time
Klein et al. 1994 estimate works well for uniform spheres.
Turbulent clouds disrupt faster than Klein et al. 1994 estimate.
Schneider & Robertson (2015)
Cloud Density Distribution and Entrainment
sphere
cloud
Schneider & Robertson (in prep)
Cloud Density Distribution and Entrainment
sphere
cloud
Schneider & Robertson (in prep)
Phase Structure of Entrained Flow: Sphere vs. Cloud sphere
cloud
Schneider & Robertson (in prep)
Momentum Structure of Entrained Flow: Sphere vs. Cloud sphere
cloud
Schneider & Robertson (in prep)
Column Densities: Sphere vs. Cloud
sphere
cloud
Schneider & Robertson (in prep)
Column Densities: Sphere vs. Cloud
sphere
cloud
Schneider & Robertson (in prep)
Momentum Column Densities: Sphere vs. Cloud
wind
sphere
cloud
wind Schneider & Robertson (in prep)
