clean up Context

src/context.jl

@@ -1,18 +1,19 @@
 @enum CONTEXT_TYPE CPU=1 GPU=2 CPU_PINNED=3
 
-type Context
+immutable Context
   device_type :: CONTEXT_TYPE
-  device_id   :: Cint
-
-  old_ctx     :: Nullable{Context}
+  device_id   :: Int
 end
-Context(dev_type :: Union{CONTEXT_TYPE, Integer}, dev_id :: Integer = 0) =
-    Context(convert(CONTEXT_TYPE, dev_type), convert(Cint, dev_id), Nullable{Context}())
+Context(dev_type :: Union{CONTEXT_TYPE, Int}, dev_id :: Int = 0) =
+    Context(convert(CONTEXT_TYPE, dev_type), dev_id)
 
 function Base.show(io :: IO, ctx :: Context)
   print(io, "$(ctx.device_type)$(ctx.device_id)")
 end
 
-
-# global default context
-DEFAULT_CONTEXT = Context(CPU)
+function cpu(dev_id::Int=0)
+  return Context(CPU, dev_id)
+end
+function gpu(dev_id::Int=0)
+  return Context(GPU, dev_id)
+end

src/ndarray.jl

@@ -43,7 +43,7 @@ function Base.show(io :: IO, arr :: NDArray)
 end
 
 function NDArray{T<:Real}(data :: Array{T})
-  copy(data, mx.DEFAULT_CONTEXT)
+  copy(data, cpu())
 end
 
 function Base.unsafe_convert(::Type{MX_handle}, obj::NDArray)
@@ -63,7 +63,10 @@ function context(arr :: NDArray)
   return Context(ref_typeid[], ref_devid[])
 end
 
-function empty{N}(shape :: NTuple{N, Int}, ctx :: Context = DEFAULT_CONTEXT)
+function empty{N}(shape :: NTuple{N, Int})
+  empty(shape, cpu())
+end
+function empty{N}(shape :: NTuple{N, Int}, ctx :: Context)
   NDArray(_ndarray_alloc(shape, ctx, false))
 end
 function empty(shape :: Int...)
@@ -99,7 +102,10 @@ function eltype(arr :: NDArray)
 end
 
 "Create zero-ed NDArray of specific shape"
-function zeros{N}(shape :: NTuple{N, Int}, ctx :: Context = DEFAULT_CONTEXT)
+function zeros{N}(shape :: NTuple{N, Int})
+  zeros(shape, cpu())
+end
+function zeros{N}(shape :: NTuple{N, Int}, ctx :: Context)
   arr = empty(shape, ctx)
   arr[:] = 0
   return arr
@@ -109,7 +115,10 @@ function zeros(shape :: Int...)
 end
 
 "Create NDArray and initialize with 1"
-function ones{N}(shape :: NTuple{N, Int}, ctx :: Context = DEFAULT_CONTEXT)
+function ones{N}(shape :: NTuple{N, Int})
+  ones(shape, cpu())
+end
+function ones{N}(shape :: NTuple{N, Int}, ctx :: Context)
   arr = empty(shape, ctx)
   arr[:] = 1
   return arr

src/random.jl

@@ -1,15 +1,21 @@
 function rand!(low::Real, high::Real, out::NDArray)
   _random_uniform(low, high, out)
 end
-function rand(low::Real, high::Real, shape::Tuple, ctx::Context=DEFAULT_CONTEXT)
+function rand{N}(low::Real, high::Real, shape::NTuple{N, Int})
+  rand(low, high, shape, cpu())
+end
+function rand{N}(low::Real, high::Real, shape::NTuple{N, Int}, ctx::Context)
   out = empty(shape, ctx)
   rand!(low, high, out)
 end
 
 function randn!(mean::Real, stdvar::Real, out::NDArray)
   _random_gaussian(mean, stdvar, out)
 end
-function randn(mean::Real, stdvar::Real, shape::Tuple, ctx::Context=DEFAULT_CONTEXT)
+function randn{N}(mean::Real, stdvar::Real, shape::NTuple{N,Int})
+  randn(mean, stdvar, shape, cpu())
+end
+function randn{N}(mean::Real, stdvar::Real, shape::NTuple{N,Int}, ctx::Context)
   out = empty(shape, ctx)
   randn!(mean, stdvar, out)
 end

test/unittest/ndarray.jl

@@ -9,7 +9,7 @@ using ..Main: rand_dims, reldiff
 ################################################################################
 function rand_tensors{N}(dims::NTuple{N, Int})
   tensor = rand(mx.MX_float, dims)
-  array  = copy(tensor, mx.DEFAULT_CONTEXT)
+  array  = copy(tensor, mx.cpu())
   return (tensor, array)
 end
 
@@ -20,12 +20,12 @@ function test_copy()
   info("NDArray::copy::dims = $dims")
 
   # copy to NDArray and back
-  array   = copy(tensor, mx.DEFAULT_CONTEXT)
+  array   = copy(tensor, mx.cpu())
   tensor2 = copy(array)
   @test reldiff(tensor, tensor2) < 1e-6
 
   # copy between NDArray
-  array2  = copy(array, mx.DEFAULT_CONTEXT)
+  array2  = copy(array, mx.cpu())
   tensor2 = copy(array2)
   @test reldiff(tensor, tensor2) < 1e-6
 end