Add support for diagm

[kshyatt]

Pull over the logic from CUDA.jl so that we can use it on AMD and other backends.

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic master) to apply these changes.

Click here to view the suggested changes.

diff --git a/src/host/linalg.jl b/src/host/linalg.jl
index 0cc40bf..f81460a 100644
--- a/src/host/linalg.jl
+++ b/src/host/linalg.jl
@@ -64,12 +64,12 @@ end
 
 ## diagm
 
-LinearAlgebra.diagm(kv::Pair{<:Integer,<:AbstractGPUVector}...) = _gpu_diagm(nothing, kv...)
-LinearAlgebra.diagm(m::Integer, n::Integer, kv::Pair{<:Integer,<:AbstractGPUVector}...) = _gpu_diagm((Int(m),Int(n)), kv...)
+LinearAlgebra.diagm(kv::Pair{<:Integer, <:AbstractGPUVector}...) = _gpu_diagm(nothing, kv...)
+LinearAlgebra.diagm(m::Integer, n::Integer, kv::Pair{<:Integer, <:AbstractGPUVector}...) = _gpu_diagm((Int(m), Int(n)), kv...)
 LinearAlgebra.diagm(v::AbstractGPUVector) = LinearAlgebra.diagm(0 => v)
 LinearAlgebra.diagm(m::Integer, n::Integer, v::AbstractGPUVector) = LinearAlgebra.diagm(m, n, 0 => v)
 
-function _gpu_diagm(size, kv::Pair{<:Integer,<:AbstractGPUVector}...)
+function _gpu_diagm(size, kv::Pair{<:Integer, <:AbstractGPUVector}...)
     A = LinearAlgebra.diagm_container(size, kv...)
     for p in kv
         inds = LinearAlgebra.diagind(A, p.first)
@@ -78,7 +78,7 @@ function _gpu_diagm(size, kv::Pair{<:Integer,<:AbstractGPUVector}...)
     return A
 end
 
-function LinearAlgebra.diagm_container(size, kv::Pair{<:Integer,<:AbstractGPUVector}...)
+function LinearAlgebra.diagm_container(size, kv::Pair{<:Integer, <:AbstractGPUVector}...)
     T = promote_type(map(x -> eltype(x.second), kv)...)
     U = promote_type(T, typeof(zero(T)))
     A = similar(kv[1].second, U, LinearAlgebra.diagm_size(size, kv...)...)
@@ -86,13 +86,13 @@ function LinearAlgebra.diagm_container(size, kv::Pair{<:Integer,<:AbstractGPUVec
     return A
 end
 
-function LinearAlgebra.diagm_size(size::Nothing, kv::Pair{<:Integer,<:AbstractGPUVector}...)
-    mnmax = mapreduce(x -> length(x.second) + abs(Int(x.first)), max, kv; init=0)
+function LinearAlgebra.diagm_size(size::Nothing, kv::Pair{<:Integer, <:AbstractGPUVector}...)
+    mnmax = mapreduce(x -> length(x.second) + abs(Int(x.first)), max, kv; init = 0)
     return mnmax, mnmax
 end
-function LinearAlgebra.diagm_size(size::Tuple{Int,Int}, kv::Pair{<:Integer,<:AbstractGPUVector}...)
-    mmax = mapreduce(x -> length(x.second) - min(0,Int(x.first)), max, kv; init=0)
-    nmax = mapreduce(x -> length(x.second) + max(0,Int(x.first)), max, kv; init=0)
+function LinearAlgebra.diagm_size(size::Tuple{Int, Int}, kv::Pair{<:Integer, <:AbstractGPUVector}...)
+    mmax = mapreduce(x -> length(x.second) - min(0, Int(x.first)), max, kv; init = 0)
+    nmax = mapreduce(x -> length(x.second) + max(0, Int(x.first)), max, kv; init = 0)
     m, n = size
     (m ≥ mmax && n ≥ nmax) || throw(DimensionMismatch(lazy"invalid size=$size"))
     return m, n