Skip to content

Commit

Permalink
Merge pull request #335 from doraemonho/main
Browse files Browse the repository at this point in the history
Add LSRK54 time stepping method
  • Loading branch information
navidcy committed Oct 23, 2022
2 parents 32fa72e + ff63faa commit 4cdccae
Show file tree
Hide file tree
Showing 4 changed files with 109 additions and 7 deletions.
2 changes: 1 addition & 1 deletion Project.toml
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ authors = ["Gregory L. Wagner <[email protected]>", "Navid C. Constantinou <
description = "Tools for building fast, hackable, pseudospectral partial differential equation solvers on periodic domains."
documentation = "https://fourierflows.github.io/FourierFlowsDocumentation/stable/"
repository = "https://github.com/FourierFlows/FourierFlows.jl"
version = "0.10.0"
version = "0.10.1"

[deps]
CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
Expand Down
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -102,7 +102,7 @@ and [Navid C. Constantinou][] (@navidcy).

The code is citable via [zenodo](https://zenodo.org). Please cite as:

> Gregory L. Wagner, Navid C. Constantinou, and contributors. (2022). FourierFlows/FourierFlows.jl: FourierFlows v0.10.0 (Version v0.10.0). Zenodo. [http://doi.org/10.5281/zenodo.1161724](http://doi.org/10.5281/zenodo.1161724)
> Gregory L. Wagner, Navid C. Constantinou, and contributors. (2022). FourierFlows/FourierFlows.jl: FourierFlows v0.10.1 (Version v0.10.1). Zenodo. [http://doi.org/10.5281/zenodo.1161724](http://doi.org/10.5281/zenodo.1161724)

## Contributing
Expand Down
111 changes: 106 additions & 5 deletions src/timesteppers.jl
Original file line number Diff line number Diff line change
Expand Up @@ -40,7 +40,8 @@ const fullyexplicitsteppers= [
:AB3,
:FilteredForwardEuler,
:FilteredRK4,
:FilteredAB3
:FilteredAB3,
:LSRK54
]

isexplicit(stepper) = any(Symbol(stepper) .== fullyexplicitsteppers)
Expand Down Expand Up @@ -73,6 +74,7 @@ end
# * RK4
# * Filtered RK4
# * ETDRK4
# * LSRK54
# * Filtered ETDRK4
# * AB3
# * Filtered AB3
Expand Down Expand Up @@ -166,6 +168,11 @@ k₂ = RHS(uⁿ + k₁ * dt/2, tⁿ + dt/2)
k₃ = RHS(uⁿ + k₂ * dt/2, tⁿ + dt/2)
k₄ = RHS(uⁿ + k₃ * dt, tⁿ + dt)
```
!!! info "Usage"
If you are limited by memory then consider switching to [`LSRK54TimeStepper`](@ref).
The [`LSRK54TimeStepper`](@ref) timestepper has half the memory footprint compared
to the `RK4TimeStepper` with a 25~30% performance trade off.
"""
struct RK4TimeStepper{T} <: AbstractTimeStepper{T}
sol₁ :: T
Expand Down Expand Up @@ -277,10 +284,104 @@ function stepforward!(sol, clock, ts::FilteredRK4TimeStepper, equation, vars, pa
return nothing
end

# --
# LSRK(5)4
# --

"""
struct LSRK54TimeStepper{T} <: AbstractTimeStepper{T}
A 4th-order 5-stages 2-storage Runge-Kutta timestepper for time-stepping
`∂u/∂t = RHS(u, t)` via:
```
S² = 0
for i = 1:5
S² = Aᵢ * S² + dt * RHS(uⁿ, t₀ + Cᵢ * dt)
uⁿ += Bᵢ * S²
end
uⁿ⁺¹ = uⁿ
```
where `Aᵢ`, `Bᵢ`, and `Cᵢ` are the ``A``, ``B``, and ``C`` coefficients from
the LSRK tableau table at the ``i``-th stage. For details, please refer to
> Carpenter, M. H. and Kennedy, C. A. (1994). Fourth-order 2N-storage Runge–Kutta schemes, Technical Report NASA TM-109112, NASA Langley Research Center, VA.
!!! info "Usage"
The `LSRK54TimeStepper` is *slower* than the [`RK4TimeStepper`](@ref) but
with *less* memory footprint; half compared to [`RK4TimeStepper`](@ref).
If you are bound by performance then use [`RK4TimeStepper`](@ref); if your
simulation is bound by memory then consider using `LSRK54TimeStepper`.
"""
struct LSRK54TimeStepper{T,V} <: AbstractTimeStepper{T}
:: T
RHS :: T
A :: V
B :: V
C :: V
end

"""
LSRK54TimeStepper(equation::Equation, dev::Device=CPU())
# ------
Construct a 4th-order 5-stages low storage Runge-Kutta timestepper for `equation` on device `dev`.
"""
function LSRK54TimeStepper(equation::Equation, dev::Device=CPU())
@devzeros typeof(dev) equation.T equation.dims S² RHS

T = equation.T
A = T[0,
-567301805773//1357537059087,
-2404267990393//2016746695238,
-3550918686646//2091501179385,
-1275806237668//842570457699]

B = T[1432997174477//9575080441755,
5161836677717//13612068292357,
1720146321549//2090206949498,
3134564353537//4481467310338,
2277821191437//14882151754819]
C = T[0,
1432997174477//9575080441755,
2526269341429//6820363962896,
2006345519317//3224310063776,
2802321613138//2924317926251]

return LSRK54TimeStepper(S², RHS, Tuple(A), Tuple(B), Tuple(C))
end

function LSRK54!(sol, clock, ts, equation, vars, params, grid, t, dt)
T = equation.T
A, B, C = ts.A, ts.B, ts.C

# initialize the S² term
@. ts.= 0

for i = 1:5
equation.calcN!(ts.RHS, sol, t + C[i] * dt , clock, vars, params, grid)
addlinearterm!(ts.RHS, equation.L, sol)

@. ts.= A[i] * ts.+ dt * ts.RHS
@. sol += B[i] * ts.
end

return nothing
end

function stepforward!(sol, clock, ts::LSRK54TimeStepper, equation, vars, params, grid)
LSRK54!(sol, clock, ts, equation, vars, params, grid, clock.t, clock.dt)
clock.t += clock.dt
clock.step += 1

return nothing
end

# --
# ETDRK4
# ------
# --

"""
struct ETDRK4TimeStepper{T,TL} <: AbstractTimeStepper{T}
Expand Down Expand Up @@ -430,9 +531,9 @@ function stepforward!(sol, clock, ts::FilteredETDRK4TimeStepper, equation, vars,
end


# ---
# --
# AB3
# ---
# --

const ab3h1 = 23/12
const ab3h2 = 16/12
Expand Down
1 change: 1 addition & 0 deletions test/runtests.jl
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@ const rtol_timesteppers = 1e-12
const steppers = [
"ForwardEuler",
"RK4",
"LSRK54",
"ETDRK4",
"AB3",
"FilteredForwardEuler",
Expand Down

2 comments on commit 4cdccae

@navidcy
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@JuliaRegistrator
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Registration pull request created: JuliaRegistries/General/70845

After the above pull request is merged, it is recommended that a tag is created on this repository for the registered package version.

This will be done automatically if the Julia TagBot GitHub Action is installed, or can be done manually through the github interface, or via:

git tag -a v0.10.1 -m "<description of version>" 4cdccae5cd3368cceb44644a459af1563d6642e5
git push origin v0.10.1

Please sign in to comment.