Merge pull request #45 from arhik/main

update examples

Author: arhik

examples/elemwise.jl

@@ -1,4 +1,10 @@
+using Test
 using WGPUCompute
 using WGPUCompute: elwise
 x = WgpuArray{Float32, 2}(rand(16, 16))
 y = elwise(+, x, x)
+
+x_cpu = x |> collect
+z_cpu = 2.0f0.*x_cpu
+
+@test z_cpu ≈ (y |> collect)

src/array.jl

@@ -163,7 +163,6 @@ function unsafe_fill!(gpuDevice, dst::WgpuArrayPtr{T}, value::Union{UInt8, Int8}
 	WGPUCore.writeBuffer(gpuDevice.queue, dst.buffer, fill(value, N))
 end
 
-
 mutable struct WgpuArray{T, N} <: AbstractGPUArray{T, N}
 	dims::Dims{N}
 	storageData::Union{Nothing, Vector{T}, Array{T}}

src/ops/elemwise.jl

@@ -8,7 +8,7 @@ function broadcast_kernel(op::Function, x::WgpuArray{T, N}, y::WgpuArray{T, N},
 	out[gId] = op(x[gId], y[gId])
 end
 
-function broadcast_kernel(op::Function, x::WgpuArray{T, N}, y::Number, out::WgpuArray{T, N}) where {T, N}
+function broadcast_kernel(op::Function, x::WgpuArray{T, N}, y::Float32, out::WgpuArray{T, N}) where {T, N}
 	xdim = workgroupDims.x
 	ydim = workgroupDims.y
 	gIdx = workgroupId.x*xdim + localId.x
@@ -44,15 +44,15 @@ function elwise(f::Function, a::Number, b::WgpuArray{T, N}) where {T, N}
 	return out
 end
 
-# Base.:+(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(+, a, b)
-# Base.:-(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(-, a, b)
-# Base.:*(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(*, a, b)
-# Base.:/(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(/, a, b)
-# Base.:+(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(+, a, b)
-# Base.:-(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(-, a, b)
-# Base.:*(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(*, a, b)
-# Base.:/(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(/, a, b)
-# Base.:+(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(+, a, b)
-# Base.:-(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(-, a, b)
-# Base.:*(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(*, a, b)
-# Base.:/(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(/, a, b)
+Base.:+(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(+, a, b)
+Base.:-(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(-, a, b)
+Base.:*(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(*, a, b)
+Base.:/(a::WgpuArray{T, N}, b::WgpuArray{T, N}) where {T, N} = elwise(/, a, b)
+Base.:+(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(+, a, b)
+Base.:-(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(-, a, b)
+Base.:*(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(*, a, b)
+Base.:/(a::WgpuArray{T, N}, b::Number) where {T, N} = elwise(/, a, b)
+Base.:+(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(+, a, b)
+Base.:-(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(-, a, b)
+Base.:*(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(*, a, b)
+Base.:/(a::Number, b::WgpuArray{T, N}) where {T, N} = elwise(/, a, b)