More AMD updates and fixes

Author: Katharine Hyatt

Project.toml

@@ -25,9 +25,9 @@ FiniteDifferences = "26cc04aa-876d-5657-8c51-4c34ba976000"
 cuTENSOR = "011b41b2-24ef-40a8-b3eb-fa098493e9e1"
 
 [sources]
-GPUArrays = {rev = "ksh/more_diag", url = "https://github.com/JuliaGPU/GPUArrays.jl"}
+GPUArrays = {rev = "master", url = "https://github.com/JuliaGPU/GPUArrays.jl"}
 MatrixAlgebraKit = {rev = "ksh/tk", url = "https://github.com/QuantumKitHub/MatrixAlgebraKit.jl"}
-AMDGPU = {rev = "ksh/diag_norm", url = "https://github.com/JuliaGPU/AMDGPU.jl"}
+AMDGPU = {rev = "master", url = "https://github.com/JuliaGPU/AMDGPU.jl"}
 cuTENSOR = {subdir = "lib/cutensor", url = "https://github.com/JuliaGPU/CUDA.jl"}
 
 [extensions]

ext/TensorKitAMDGPUExt/TensorKitAMDGPUExt.jl

@@ -7,6 +7,7 @@ import AMDGPU: rand as rocrand, rand! as rocrand!, randn as rocrandn, randn! as
 using TensorKit
 import TensorKit.VectorInterface: scalartype as vi_scalartype
 using TensorKit.Factorizations
+using TensorKit.Strided
 using TensorKit.Factorizations: AbstractAlgorithm
 using TensorKit: SectorDict, tensormaptype, scalar, similarstoragetype, AdjointTensorMap
 

ext/TensorKitAMDGPUExt/roctensormap.jl

@@ -224,3 +224,25 @@ function LinearAlgebra.isposdef(t::ROCTensorMap)
     end
     return true
 end
+
+# Conversion to ROCArray:
+#----------------------
+# probably not optimized for speed, only for checking purposes
+function Base.convert(::Type{ROCArray}, t::AbstractTensorMap)
+    I = sectortype(t)
+    if I === Trivial
+        convert(ROCArray, t[])
+    else
+        cod = codomain(t)
+        dom = domain(t)
+        T = sectorscalartype(I) <: Complex ? complex(scalartype(t)) :
+            sectorscalartype(I) <: Integer ? scalartype(t) : float(scalartype(t))
+        A = AMDGPU.zeros(T, dims(cod)..., dims(dom)...)
+        for (f₁, f₂) in fusiontrees(t)
+            F = convert(ROCArray, (f₁, f₂))
+            Aslice = StridedView(A)[axes(cod, f₁.uncoupled)..., axes(dom, f₂.uncoupled)...]
+            add!(Aslice, StridedView(TensorKit._kron(convert(ROCArray, t[f₁, f₂]), F)))
+        end
+        return A
+    end
+end