fixed oneapi and amdgpu issues

Author: sshin23

ext/ExaModelsAMDGPU.jl

@@ -4,16 +4,4 @@ import ExaModels, AMDGPU
 
 ExaModels.convert_array(v, backend::AMDGPU.ROCBackend) = AMDGPU.ROCArray(v)
 
-ExaModels.sort!(array::A; lt = isless) where {A<:AMDGPU.ROCVector} =
-    copyto!(array, sort!(Array(array); lt = lt))
-
-# Below are type piracy
-function Base.findall(f::F, bitarray::A) where {F<:Function,A<:AMDGPU.ROCVector}
-    a = Array(bitarray)
-    b = findall(f, a)
-    c = similar(bitarray, eltype(b), length(b))
-
-    return copyto!(c, b)
-end
-Base.findall(bitarray::A) where {A<:AMDGPU.ROCVector} = Base.findall(identity, bitarray)
 end

ext/ExaModelsCUDA.jl

@@ -3,7 +3,6 @@ module ExaModelsCUDA
 import ExaModels: ExaModels, NLPModels
 import CUDA: CUDA, CUDABackend, CuArray
 
-ExaModels.ExaCore(backend::CUDABackend) = ExaModels.ExaCore(Float64, backend)
 ExaModels.convert_array(v, backend::CUDABackend) = CuArray(v)
 
 end

ext/ExaModelsOneAPI.jl

@@ -7,15 +7,15 @@ function ExaModels.append!(
     a::A,
     b::Base.Generator{UnitRange{I}},
     lb,
-) where {I,A<:oneAPI.oneVector}
+) where {I,A<:oneAPI.oneArray}
     la = length(a)
     aa = similar(a, la + lb)
     copyto!(view(aa, 1:la), a)
     map!(b.f, view(aa, (la+1):(la+lb)), b.iter)
     return aa
 end
 
-function ExaModels.append!(backend, a::A, b::Base.Generator, lb) where {A<:oneAPI.oneVector}
+function ExaModels.append!(backend, a::A, b::Base.Generator, lb) where {A<:oneAPI.oneArray}
     la = length(a)
     aa = similar(a, la + lb)
     copyto!(view(aa, 1:la), a)
@@ -29,7 +29,7 @@ function ExaModels.append!(
     a::A,
     b::V,
     lb,
-) where {A<:oneAPI.oneVector,V<:AbstractVector}
+) where {A<:oneAPI.oneArray,V<:AbstractArray}
     la = length(a)
     aa = similar(a, la + lb)
     copyto!(view(aa, 1:la), a)
@@ -38,7 +38,7 @@ function ExaModels.append!(
 end
 
 
-function ExaModels.append!(backend, a::A, b::Number, lb) where {A<:oneAPI.oneVector}
+function ExaModels.append!(backend, a::A, b::Number, lb) where {A<:oneAPI.oneArray}
     la = length(a)
     aa = similar(a, la + lb)
     copyto!(view(aa, 1:la), a)
@@ -48,16 +48,6 @@ end
 
 ExaModels.convert_array(v, backend::oneAPI.oneAPIBackend) = oneAPI.oneArray(v)
 
-ExaModels.sort!(array::A; lt = isless) where {A<:oneAPI.oneVector} =
+ExaModels.sort!(array::A; lt = isless) where {A<:oneAPI.oneArray} =
     copyto!(array, sort!(Array(array); lt = lt))
-
-# below is type piracy
-function Base.findall(f::F, bitarray::A) where {F<:Function,A<:oneAPI.oneVector}
-    a = Array(bitarray)
-    b = findall(f, a)
-    c = similar(bitarray, eltype(b), length(b))
-    return copyto!(c, b)
-end
-Base.findall(bitarray::A) where {A<:oneAPI.oneVector} = Base.findall(identity, bitarray)
-
 end # module

src/nlp.jl

@@ -690,20 +690,20 @@ end
     f.o0 + f.f.first(itr[i], nothing)
 @inbounds @inline offset0(f::F, itr, i) where {T<:Tuple,P<:Pair{T},F<:SIMDFunction{P}} = f.o0 + idxx(coord(itr, i, f.f.first), Base.size(itr))
 
-idx(itr, I) = @inbounds itr[I]
-idx(itr::Base.Iterators.ProductIterator{V}, I) where V =  _idx(I-1, itr.iterators, Base.size(itr))
-function _idx(n, (vec1, vec...), (si1, si...))
+@inline idx(itr, I) = @inbounds itr[I]
+@inline idx(itr::Base.Iterators.ProductIterator{V}, I) where V =  _idx(I-1, itr.iterators, Base.size(itr))
+@inline function _idx(n, (vec1, vec...), (si1, si...))
     d, r = divrem(n, si1)
     return (vec1[r + 1], _idx(d, vec, si)...)
 end
-_idx(n, (vec,), ::Tuple{Int}) = @inbounds vec[n + 1]
+@inline _idx(n, (vec,), ::Tuple{Int}) = @inbounds vec[n + 1]
 
-idxx(coord, si) = _idxx(coord, si, 1) + 1
-_idxx((c,coord...), (s,si...), a) = a * (c - 1) + _idxx(coord, si, a*s)
-_idxx(::Tuple{}, ::Tuple{}, a) = 0
+@inline idxx(coord, si) = _idxx(coord, si, 1) + 1
+@inline _idxx(coord, si, a) = a * (coord[1] - 1) + _idxx(coord[2:end], si[2:end], a*si[1])
+@inline _idxx(::Tuple{}, ::Tuple{}, a) = 0
 
-coord(itr, i, (f,fs...)) = (f(idx(itr,i), nothing), coord(itr, i, fs)...)
-coord(itr, i, ::Tuple{}) = ()
+@inline coord(itr, i, (f,fs...)) = (f(idx(itr,i), nothing), coord(itr, i, fs)...)
+@inline coord(itr, i, ::Tuple{}) = ()
 
 for (thing, val) in [(:solution, 1), (:multipliers_L, 0), (:multipliers_U, 2)]
     @eval begin