Make current Context a global variable in CompasToolkit.jl

NLeSC-COMPAS · Feb 12, 2024 · 7491164 · 7491164
1 parent e1c1630
commit 7491164
Show file tree

Hide file tree

Showing 6 changed files with 123 additions and 105 deletions.
diff --git a/CompasToolkit.jl/examples/fisp.jl b/CompasToolkit.jl/examples/fisp.jl
@@ -5,7 +5,7 @@ using ComputationalResources
 using LinearAlgebra
 using StaticArrays
 
-context = CompasToolkit.make_context(0)
+context = CompasToolkit.init_context(0)
 
 # First we assemble a Shepp Logan phantom with homogeneous T₁ and T₂
 # but non-constant proton density and B₀
@@ -24,7 +24,7 @@ nvoxels = N*N
 
 # Finally we assemble the phantom as an array of `T₁T₂B₀ρˣρʸxy` values
 parameters_ref = map(T₁T₂B₀ρˣρʸxy, T₁, T₂, B₀, real.(ρ), imag.(ρ), X, Y)
-parameters = CompasToolkit.TissueParameters(context, nvoxels, T₁, T₂, B₁, B₀, real.(ρ), imag.(ρ), X, Y)
+parameters = CompasToolkit.TissueParameters(nvoxels, T₁, T₂, B₁, B₀, real.(ρ), imag.(ρ), X, Y)
 
 nTR = N; # nr of TRs used in the simulation
 RF_train = LinRange(1,90,nTR) |> collect .|> complex; # flip angle train
@@ -40,7 +40,7 @@ fisp_ref = FISP(RF_train, sliceprofiles, TR, TE, max_state, TI);
 RF_train = RF_train .|> ComplexF32 # constant flip angle train
 sliceprofiles = collect(sliceprofiles)  .|> ComplexF32 # z locations
 
-fisp = CompasToolkit.FispSequence(context, RF_train, sliceprofiles, Float32(TR), Float32(TE), max_state, Float32(TI))
+fisp = CompasToolkit.FispSequence(RF_train, sliceprofiles, Float32(TR), Float32(TE), max_state, Float32(TI))
 
 
 # isochromat model
@@ -50,7 +50,7 @@ echos_ref = simulate(CUDALibs(), fisp_ref, parameters_ref);
 echos_ref = collect(echos_ref)
 
 echos = zeros(ComplexF32, nvoxels, nTR)
-CompasToolkit.simulate_magnetization(context, echos, parameters, fisp)
+CompasToolkit.simulate_magnetization(parameters, fisp)
 echos = transpose(echos)
 
 println("fraction equal: ", sum(echos .≈ echos_ref) / length(echos))

diff --git a/CompasToolkit.jl/examples/pssfp.jl b/CompasToolkit.jl/examples/pssfp.jl
@@ -5,7 +5,7 @@ using ComputationalResources
 using LinearAlgebra
 using StaticArrays
 
-context = CompasToolkit.make_context(0)
+context = CompasToolkit.init_context(0)
 
 # First we assemble a Shepp Logan phantom with homogeneous T₁ and T₂
 # but non-constant proton density and B₀
@@ -24,7 +24,7 @@ nvoxels = N*N
 
 # Finally we assemble the phantom as an array of `T₁T₂B₀ρˣρʸxy` values
 parameters_ref = map(T₁T₂B₀ρˣρʸxy, T₁, T₂, B₀, real.(ρ), imag.(ρ), X, Y)
-parameters = CompasToolkit.TissueParameters(context, nvoxels, T₁, T₂, B₁, B₀, real.(ρ), imag.(ρ), X, Y)
+parameters = CompasToolkit.TissueParameters(nvoxels, T₁, T₂, B₁, B₀, real.(ρ), imag.(ρ), X, Y)
 
 # Next, we assemble a balanced sequence with constant flip angle of 60 degrees,
 nTR = N
@@ -47,17 +47,16 @@ pssfp_ref = pSSFP(RF_train, TR, γΔtRF, Δt, γΔtGRz, z)
 Δt = (Δt.ex, Δt.inv, Δt.pr) # time intervals during TR
 γΔtGRz = (γΔtGRz.ex, γΔtGRz.inv, γΔtGRz.pr) # slice select gradient strengths during TR
 
-pssfp = CompasToolkit.pSSFPSequence(context, RF_train, Float32(TR), γΔtRF, Δt, γΔtGRz, z)
+pssfp = CompasToolkit.pSSFPSequence(RF_train, Float32(TR), γΔtRF, Δt, γΔtGRz, z)
 
 # isochromat model
 pssfp_ref = gpu(f32(pssfp_ref))
 parameters_ref = gpu(f32(parameters_ref))
 echos_ref = simulate(CUDALibs(), pssfp_ref, parameters_ref);
 echos_ref = collect(echos_ref)
 
-echos = zeros(ComplexF32, nvoxels, nTR)
-CompasToolkit.simulate_magnetization(context, echos, parameters, pssfp)
-echos = transpose(echos)
+echos = CompasToolkit.simulate_magnetization(parameters, pssfp)
+echos = transpose(collect(echos))
 
 println("fraction equal: ", sum(echos .≈ echos_ref) / length(echos))
 

diff --git a/CompasToolkit.jl/examples/signal.jl b/CompasToolkit.jl/examples/signal.jl
@@ -5,7 +5,7 @@ using ComputationalResources
 using LinearAlgebra
 using StaticArrays
 
-context = CompasToolkit.make_context(0)
+context = CompasToolkit.init_context(0)
 
 # First we assemble a Shepp Logan phantom with homogeneous T₁ and T₂
 # but non-constant proton density and B₀
@@ -24,7 +24,7 @@ nvoxels = N*N
 
 # Finally we assemble the phantom as an array of `T₁T₂B₀ρˣρʸxy` values
 parameters_ref = map(T₁T₂B₀ρˣρʸxy, T₁, T₂, B₀, real.(ρ), imag.(ρ), X, Y)
-parameters = CompasToolkit.TissueParameters(context, nvoxels, T₁, T₂, B₁, B₀, real.(ρ), imag.(ρ), X, Y)
+parameters = CompasToolkit.TissueParameters(nvoxels, T₁, T₂, B₁, B₀, real.(ρ), imag.(ρ), X, Y)
 
 # Next, we assemble a balanced sequence with constant flip angle of 60 degrees,
 nTR = N
@@ -60,7 +60,7 @@ k0 = [(-ns/2 * Δkˣ) + im * (py[mod1(r,N)] * Δkʸ) for r in 1:nr]; # starting
 Δk = [Δkˣ + 0.0im for r in 1:nr]; # k-space steps per sample point for each readout
 
 trajectory_ref = CartesianTrajectory(nr,ns,Δt_adc,k0,Δk,py);
-trajectory = CompasToolkit.CartesianTrajectory(context, nr, ns, Float32(Δt_adc), ComplexF32.(k0), ComplexF32.(Δk[1]));
+trajectory = CompasToolkit.CartesianTrajectory(nr, ns, Float32(Δt_adc), ComplexF32.(k0), ComplexF32.(Δk[1]));
 
 # We use two different receive coils
 ncoils = 4
@@ -80,7 +80,6 @@ signal_ref = collect(signal_ref)
 signal_ref = reshape(signal_ref, ns, nr)
 
 signal = CompasToolkit.magnetization_to_signal(
-    context,
     echos,
     parameters,
     trajectory,