Merge pull request #102 from LilithHafner/lh/lazy-get-tmp

Make Lazy caches support get_tmp

Author: ChrisRackauckas

README.md

@@ -268,11 +268,11 @@ lbc = GeneralLazyBufferCache(function (p)
 end)
 ```
 
-then `lbc[p]` will be smart and reuse the caches. A full example looks like the following:
+then `lbc[p]` (or, equivalently, `get_tmp(lbc, p)`) will be smart and reuse the caches. A full example looks like the following:
 
 ```julia
 using Random, DifferentialEquations, LinearAlgebra, Optimization, OptimizationNLopt,
-    OptimizationOptimJL, PreallocationTools
+      OptimizationOptimJL, PreallocationTools
 
 lbc = GeneralLazyBufferCache(function (p)
     DifferentialEquations.init(ODEProblem(ode_fnc, y₀, (0.0, T), p), Tsit5(); saveat = t)

docs/src/index.md

@@ -267,7 +267,7 @@ then `lbc[p]` will be smart and reuse the caches. A full example looks like the
 
 ```julia
 using Random, DifferentialEquations, LinearAlgebra, Optimization, OptimizationNLopt,
-    OptimizationOptimJL, PreallocationTools
+      OptimizationOptimJL, PreallocationTools
 
 lbc = GeneralLazyBufferCache(function (p)
     DifferentialEquations.init(ODEProblem(ode_fnc, y₀, (0.0, T), p), Tsit5(); saveat = t)

src/PreallocationTools.jl

@@ -216,14 +216,16 @@ function similar_type(x::AbstractArray{T}, s::NTuple{N, Integer}) where {T, N}
     typeof(similar(x, ntuple(Returns(1), N)))
 end
 
-# override the [] method
-function Base.getindex(b::LazyBufferCache, u::T) where {T <: AbstractArray}
+function get_tmp(b::LazyBufferCache, u::T) where {T <: AbstractArray}
     s = b.sizemap(size(u)) # required buffer size
     get!(b.bufs, (T, s)) do
         similar(u, s) # buffer to allocate if it was not found in b.bufs
     end::similar_type(u, s) # declare type since b.bufs dictionary is untyped
 end
 
+# override the [] method
+Base.getindex(b::LazyBufferCache, u::T) where {T <: AbstractArray} = get_tmp(b, u)
+
 # GeneralLazyBufferCache
 
 """
@@ -246,11 +248,12 @@ struct GeneralLazyBufferCache{F <: Function}
     GeneralLazyBufferCache(f::F = identity) where {F <: Function} = new{F}(Dict(), f) # start with empty dict
 end
 
-function Base.getindex(b::GeneralLazyBufferCache, u::T) where {T}
+function get_tmp(b::GeneralLazyBufferCache, u::T) where {T}
     get!(b.bufs, T) do
         b.f(u)
     end
 end
+Base.getindex(b::GeneralLazyBufferCache, u::T) where {T} = get_tmp(b, u)
 
 export GeneralLazyBufferCache, FixedSizeDiffCache, DiffCache, LazyBufferCache, dualcache
 export get_tmp

test/core_dispatch.jl

@@ -1,6 +1,6 @@
 using LinearAlgebra,
-    Test, PreallocationTools, ForwardDiff, LabelledArrays,
-    RecursiveArrayTools
+      Test, PreallocationTools, ForwardDiff, LabelledArrays,
+      RecursiveArrayTools
 
 function test(u0, dual, chunk_size)
     cache = PreallocationTools.DiffCache(u0, chunk_size)
@@ -53,8 +53,10 @@ results = test(u0, dual, chunk_size)
 
 chunk_size = 5
 u0_B = ones(5, 5)
-dual_B = zeros(ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float64}, Float64,
-        chunk_size}, 2, 2)
+dual_B = zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float64}, Float64,
+        chunk_size},
+    2, 2)
 cache_B = FixedSizeDiffCache(u0_B, chunk_size)
 tmp_du_BA = get_tmp(cache_B, u0_B)
 tmp_dual_du_BA = get_tmp(cache_B, dual_B)
@@ -102,9 +104,11 @@ results = test(u0, dual, chunk_size)
 #ArrayPartition tests
 chunk_size = 2
 u0 = ArrayPartition(ones(2, 2), ones(3, 3))
-dual_a = zeros(ForwardDiff.Dual{ForwardDiff.Tag{nothing, Float64}, Float64,
+dual_a = zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{nothing, Float64}, Float64,
         chunk_size}, 2, 2)
-dual_b = zeros(ForwardDiff.Dual{ForwardDiff.Tag{nothing, Float64}, Float64,
+dual_b = zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{nothing, Float64}, Float64,
         chunk_size}, 3, 3)
 dual = ArrayPartition(dual_a, dual_b)
 results = test(u0, dual, chunk_size)
@@ -128,10 +132,14 @@ results = test(u0, dual, chunk_size)
 @test eltype(results[7]) == eltype(dual)
 
 u0_AP = ArrayPartition(ones(2, 2), ones(3, 3))
-dual_a = zeros(ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float64}, Float64,
-        chunk_size}, 2, 2)
-dual_b = zeros(ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float64}, Float64,
-        chunk_size}, 3, 3)
+dual_a = zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float64}, Float64,
+        chunk_size},
+    2, 2)
+dual_b = zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float64}, Float64,
+        chunk_size},
+    3, 3)
 dual_AP = ArrayPartition(dual_a, dual_b)
 cache_AP = FixedSizeDiffCache(u0_AP, chunk_size)
 tmp_du_APA = get_tmp(cache_AP, u0_AP)

test/core_nesteddual.jl

@@ -1,6 +1,6 @@
 using LinearAlgebra,
-    OrdinaryDiffEq, Test, PreallocationTools, ForwardDiff, Optimization,
-    OptimizationOptimJL
+      OrdinaryDiffEq, Test, PreallocationTools, ForwardDiff, Optimization,
+      OptimizationOptimJL
 
 randmat = rand(5, 3)
 sto = similar(randmat)
@@ -23,7 +23,8 @@ end
 In setting up the DiffCache, we are setting chunk_size to [1, 1], because we differentiate
 only with respect to τ. This initializes the cache with the minimum memory needed. =#
 stod = DiffCache(sto, [1, 1])
-df3 = ForwardDiff.derivative(τ -> ForwardDiff.derivative(ξ -> claytonsample!(stod, ξ, 0.0),
+df3 = ForwardDiff.derivative(
+    τ -> ForwardDiff.derivative(ξ -> claytonsample!(stod, ξ, 0.0),
         τ), 0.3)
 
 #= taking the second derivative of claytonsample! with respect to τ with auto-detected chunk-size.
@@ -32,7 +33,8 @@ than what's needed (1+1), the auto-allocated cache is big enough to handle the n
 if we don't specify the keyword argument levels = 2. This should in general not be relied on to work,
 especially if more levels of nesting occur (see optimization example below). =#
 stod = DiffCache(sto)
-df4 = ForwardDiff.derivative(τ -> ForwardDiff.derivative(ξ -> claytonsample!(stod, ξ, 0.0),
+df4 = ForwardDiff.derivative(
+    τ -> ForwardDiff.derivative(ξ -> claytonsample!(stod, ξ, 0.0),
         τ), 0.3)
 
 @test df3 ≈ df4
@@ -41,7 +43,8 @@ df4 = ForwardDiff.derivative(τ -> ForwardDiff.derivative(ξ -> claytonsample!(s
 For the given size of sto, ForwardDiff's heuristic chooses chunk_size = 8 and with keyword arg levels = 2,
 the created cache size is larger than what's needed (even more so than the last example). =#
 stod = DiffCache(sto, levels = 2)
-df5 = ForwardDiff.derivative(τ -> ForwardDiff.derivative(ξ -> claytonsample!(stod, ξ, 0.0),
+df5 = ForwardDiff.derivative(
+    τ -> ForwardDiff.derivative(ξ -> claytonsample!(stod, ξ, 0.0),
         τ), 0.3)
 
 @test df3 ≈ df5

test/core_odes.jl

@@ -1,6 +1,6 @@
 using LinearAlgebra,
-    OrdinaryDiffEq, Test, PreallocationTools, LabelledArrays,
-    RecursiveArrayTools
+      OrdinaryDiffEq, Test, PreallocationTools, LabelledArrays,
+      RecursiveArrayTools
 
 #Base array
 function foo(du, u, (A, tmp), t)

test/general_lbc.jl

@@ -1,6 +1,6 @@
 using Random,
-    OrdinaryDiffEq, LinearAlgebra, Optimization, OptimizationOptimJL,
-    PreallocationTools
+      OrdinaryDiffEq, LinearAlgebra, Optimization, OptimizationOptimJL,
+      PreallocationTools
 
 lbc = GeneralLazyBufferCache(function (p)
     init(ODEProblem(ode_fnc, y₀,
@@ -40,6 +40,7 @@ x = rand(1000)
 y = view(x, 1:900)
 @inferred cache[y]
 @test 0 == @allocated cache[y]
+@test cache[y] === get_tmp(cache, y)
 
 cache_17 = LazyBufferCache(Returns(17))
 x = 1:10
@@ -52,3 +53,4 @@ cache = GeneralLazyBufferCache(T -> Vector{T}(undef, 1000))
 # @inferred cache[Float64]
 cache[Float64] # generate the buffer
 @test 0 == @allocated cache[Float64]
+@test get_tmp(cache, Float64) === cache[Float64]

test/gpu_all.jl

@@ -1,14 +1,16 @@
 using LinearAlgebra,
-    OrdinaryDiffEq, Test, PreallocationTools, CUDA, ForwardDiff
+      OrdinaryDiffEq, Test, PreallocationTools, CUDA, ForwardDiff
 
 # upstream
 OrdinaryDiffEq.DiffEqBase.anyeltypedual(x::FixedSizeDiffCache, counter = 0) = Any
 
 #Dispatch tests
 chunk_size = 5
 u0_CU = cu(ones(5, 5))
-dual_CU = cu(zeros(ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float32}, Float32,
-        chunk_size}, 2, 2))
+dual_CU = cu(zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float32}, Float32,
+        chunk_size},
+    2, 2))
 dual_N = ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float32}, Float32, 5}(0)
 cache_CU = DiffCache(u0_CU, chunk_size)
 tmp_du_CUA = get_tmp(cache_CU, u0_CU)
@@ -32,8 +34,10 @@ tmp_dual_du_CUN = get_tmp(cache_CU, dual_N)
 
 chunk_size = 5
 u0_B = cu(ones(5, 5))
-dual_B = cu(zeros(ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float32}, Float32,
-        chunk_size}, 2, 2))
+dual_B = cu(zeros(
+    ForwardDiff.Dual{ForwardDiff.Tag{typeof(something), Float32}, Float32,
+        chunk_size},
+    2, 2))
 cache_B = FixedSizeDiffCache(u0_B, chunk_size)
 tmp_du_BA = get_tmp(cache_B, u0_B)
 tmp_dual_du_BA = get_tmp(cache_B, dual_B)