Mercurial > repos > public > sbplib_julia

comparison src/LazyTensors/lazy_tensor_operations.jl @ 436:cffdac9c612d bugfix/tensor_application_multiplication

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Close branch before merge
author Jonatan Werpers <jonatan@werpers.com>
date Mon, 19 Oct 2020 20:54:23 +0200
parents 2808c41f5efa 2958b4ebd565
children 907b0510699f
comparison
equal deleted inserted replaced
422:f0d6906d8937 436:cffdac9c612d
74 domain_size(tmBinOp::LazyTensorMappingBinaryOperation{Op,T,R,D}) where {Op,T,R,D} = domain_size(tmBinOp.tm1) 74 domain_size(tmBinOp::LazyTensorMappingBinaryOperation{Op,T,R,D}) where {Op,T,R,D} = domain_size(tmBinOp.tm1)
75 75
76 Base.:+(tm1::TensorMapping{T,R,D}, tm2::TensorMapping{T,R,D}) where {T,R,D} = LazyTensorMappingBinaryOperation{:+,T,R,D}(tm1,tm2) 76 Base.:+(tm1::TensorMapping{T,R,D}, tm2::TensorMapping{T,R,D}) where {T,R,D} = LazyTensorMappingBinaryOperation{:+,T,R,D}(tm1,tm2)
77 Base.:-(tm1::TensorMapping{T,R,D}, tm2::TensorMapping{T,R,D}) where {T,R,D} = LazyTensorMappingBinaryOperation{:-,T,R,D}(tm1,tm2) 77 Base.:-(tm1::TensorMapping{T,R,D}, tm2::TensorMapping{T,R,D}) where {T,R,D} = LazyTensorMappingBinaryOperation{:-,T,R,D}(tm1,tm2)
78 78
79 """
80 TensorMappingComposition{T,R,K,D}
79 81
80 # TODO: Write tests and documentation for LazyTensorMappingComposition 82 Lazily compose two TensorMappings, so that they can be handled as a single TensorMapping.
81 # struct LazyTensorMappingComposition{T,R,K,D} <: TensorMapping{T,R,D} 83 """
82 # t1::TensorMapping{T,R,K} 84 struct TensorMappingComposition{T,R,K,D, TM1<:TensorMapping{T,R,K}, TM2<:TensorMapping{T,K,D}} <: TensorMapping{T,R,D}
83 # t2::TensorMapping{T,K,D} 85 t1::TM1
84 # end 86 t2::TM2
85 87
86 # Base.:∘(s::TensorMapping{T,R,K}, t::TensorMapping{T,K,D}) where {T,R,K,D} = LazyTensorMappingComposition(s,t) 88 @inline function TensorMappingComposition(t1::TensorMapping{T,R,K}, t2::TensorMapping{T,K,D}) where {T,R,K,D}
89 @boundscheck if domain_size(t1) != range_size(t2)
90 throw(DimensionMismatch("the first argument has domain size $(domain_size(t1)) while the second has range size $(range_size(t2)) "))
91 end
92 return new{T,R,K,D, typeof(t1), typeof(t2)}(t1,t2)
93 end
94 # Add check for matching sizes as a boundscheck
95 end
96 export TensorMappingComposition
87 97
88 # function range_size(tm::LazyTensorMappingComposition{T,R,K,D}, domain_size::NTuple{D,Integer}) where {T,R,K,D} 98 range_size(tm::TensorMappingComposition) = range_size(tm.t1)
89 # range_size(tm.t1, domain_size(tm.t2, domain_size)) 99 domain_size(tm::TensorMappingComposition) = domain_size(tm.t2)
90 # end
91 100
92 # function domain_size(tm::LazyTensorMappingComposition{T,R,K,D}, range_size::NTuple{R,Integer}) where {T,R,K,D} 101 function apply(c::TensorMappingComposition{T,R,K,D}, v::AbstractArray{T,D}, I::Vararg{S,R} where S) where {T,R,K,D}
93 # domain_size(tm.t1, domain_size(tm.t2, range_size)) 102 apply(c.t1, c.t2*v, I...)
94 # end 103 end
95 104
96 # function apply(c::LazyTensorMappingComposition{T,R,K,D}, v::AbstractArray{T,D}, I::NTuple{R,Int}) where {T,R,K,D} 105 function apply_transpose(c::TensorMappingComposition{T,R,K,D}, v::AbstractArray{T,R}, I::Vararg{S,D} where S) where {T,R,K,D}
97 # apply(c.t1, LazyTensorMappingApplication(c.t2,v), I...) 106 apply_transpose(c.t2, c.t1'*v, I...)
98 # end 107 end
99 108
100 # function apply_transpose(c::LazyTensorMappingComposition{T,R,K,D}, v::AbstractArray{T,D}, I::NTuple{D,Int}) where {T,R,K,D} 109 Base.@propagate_inbounds Base.:∘(s::TensorMapping, t::TensorMapping) = TensorMappingComposition(s,t)
101 # apply_transpose(c.t2, LazyTensorMappingApplication(c.t1',v), I...)
102 # end
103 110
104 # # Have i gone too crazy with the type parameters? Maybe they aren't all needed?
105
106 # export →
107 """ 111 """
108 LazyLinearMap{T,R,D,...}(A, range_indicies, domain_indicies) 112 LazyLinearMap{T,R,D,...}(A, range_indicies, domain_indicies)
109 113
110 TensorMapping defined by the AbstractArray A. `range_indicies` and `domain_indicies` define which indicies of A should 114 TensorMapping defined by the AbstractArray A. `range_indicies` and `domain_indicies` define which indicies of A should
111 be considerd the range and domain of the TensorMapping. Each set of indices must be ordered in ascending order. 115 be considerd the range and domain of the TensorMapping. Each set of indices must be ordered in ascending order.