remove timed enable multi-threading (#14)

* remove TimerOutputs

Project.toml

@@ -1,7 +1,7 @@
 name = "ONNXRunTime"
 uuid = "e034b28e-924e-41b2-b98f-d2bbeb830c6a"
 authors = ["Jan Weidner <[email protected]> and contributors"]
-version = "0.2.3"
+version = "0.3.0"
 
 [deps]
 ArgCheck = "dce04be8-c92d-5529-be00-80e4d2c0e197"
@@ -12,15 +12,13 @@ LazyArtifacts = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
 Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
 Requires = "ae029012-a4dd-5104-9daa-d747884805df"
-TimerOutputs = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
 
 [compat]
 ArgCheck = "2"
 CEnum = "0.4"
 DataStructures = "0.18"
 DocStringExtensions = "0.8"
 Requires = "1"
-TimerOutputs = "0.5"
 julia = "1.6"
 
 [extras]

src/ONNXRunTime.jl

@@ -1,8 +1,5 @@
 module ONNXRunTime
 using Requires:@require
-import TimerOutputs
-
-const TIMER = TimerOutputs.TimerOutput()
 
 function _perm(arr::AbstractArray{T,N}) where {T,N}
     ntuple(i->N+1-i, N)

src/capi.jl

@@ -5,8 +5,7 @@ This module closely follows the offical onnxruntime [C-API](https://github.com/m
 See [here](https://github.com/microsoft/onnxruntime-inference-examples/blob/d031f879c9a8d33c8b7dc52c5bc65fe8b9e3960d/c_cxx/fns_candy_style_transfer/fns_candy_style_transfer.c) for a C code example.
 """
 module CAPI
-using ONNXRunTime: TIMER, reversedims_lazy
-using TimerOutputs: @timeit
+using ONNXRunTime: reversedims_lazy
 
 using DocStringExtensions
 using Libdl
@@ -901,14 +900,10 @@ end
     $TYPEDSIGNATURES
 
 """
-@timeit TIMER function GetTensorMutableData(api::OrtApi, tensor::OrtValue)::AbstractArray
+function GetTensorMutableData(api::OrtApi, tensor::OrtValue)::AbstractArray
     GC.@preserve tensor begin
-        @timeit TIMER "unsafe_GetTensorMutableData" begin
-            data_owned_by_tensor::PermutedDimsArray = unsafe_GetTensorMutableData(api, tensor)
-        end
-        @timeit TIMER "copy" begin
-            reversedims_lazy(copy(parent(data_owned_by_tensor)))
-        end
+        data_owned_by_tensor::PermutedDimsArray = unsafe_GetTensorMutableData(api, tensor)
+        reversedims_lazy(copy(parent(data_owned_by_tensor)))
     end
 end
 

src/highlevel.jl

@@ -2,7 +2,6 @@ using ArgCheck
 using LazyArtifacts
 using DataStructures: OrderedDict
 using DocStringExtensions
-using TimerOutputs: @timeit, TimerOutput
 ################################################################################
 ##### testdatapath
 ################################################################################
@@ -29,7 +28,6 @@ struct InferenceSession
     allocator::OrtAllocator
     _input_names::Vector{String}
     _output_names::Vector{String}
-    timer::TimerOutput
 end
 function Base.show(io::IO, o::InferenceSession)
     print(io,
@@ -60,8 +58,8 @@ end
     $TYPEDSIGNATURES
 """
 function load_inference(path::AbstractString; execution_provider::Symbol=:cpu,
-            envname::AbstractString="defaultenv", timer=TIMER,
-                       )::InferenceSession
+        envname::AbstractString="defaultenv",
+    )::InferenceSession
     api = GetApi(;execution_provider)
     env = CreateEnv(api, name=envname)
     if execution_provider === :cpu
@@ -87,17 +85,14 @@ function load_inference(path::AbstractString; execution_provider::Symbol=:cpu,
     @check allunique(_input_names)
     @check allunique(_output_names)
     return InferenceSession(api, execution_provider, session, meminfo, allocator,
-        _input_names,
-        _output_names,
-        timer,
-    )
+                            _input_names,
+                            _output_names,
+                           )
 end
 
-@timeit o.timer function make_input_tensor(o::InferenceSession, inputs, key)
+function make_input_tensor(o::InferenceSession, inputs, key)
     arr = inputs[keytype(inputs)(key)]
-    @timeit o.timer "clayout" begin
-        cstorage = vec(reversedims(arr)::Array)
-    end
+    cstorage = vec(reversedims(arr)::Array)
     CreateTensorWithDataAsOrtValue(o.api, o.meminfo, cstorage, size(arr))
 end
 
@@ -136,46 +131,39 @@ be a `NamedTuple` or an `AbstractDict`. Optionally `output_names` can be passed.
 In this case only the outputs whose name is contained in `output_names` are computed.
 """
 function (o::InferenceSession)(
-        inputs,
-        output_names=nothing
-    )
-    timer = o.timer
-    @timeit timer "pre CAPI.Run" begin
-        if output_names === nothing
-            output_names = @__MODULE__().output_names(o)
-        end
-        @argcheck o.execution_provider in EXECUTION_PROVIDERS
-        @argcheck eltype(output_names) <: Union{AbstractString, Symbol}
-        @argcheck keytype(inputs) <: Union{AbstractString, Symbol}
-        expected_input_names = ONNXRunTime.input_names(o)
-        for key in keys(inputs)
-            if !(String(key) in expected_input_names)
-                msg = """
-                Invalid input name.
-                Expected: $(expected_input_names)
-                Got: $(key)
-                """
-                throw(ArgumentError(msg))
-            end
-        end
-        expected_output_names = ONNXRunTime.output_names(o)
-        for name in output_names
-            if !(String(name) in expected_output_names)
-                msg = """
-                Invalid output name.
-                Expected: $(expected_output_names)
-                Got: $(name)
-                """
-                throw(ArgumentError(msg))
-            end
-        end
-        inp_names, input_tensors = prepare_inputs(o, inputs)
-        run_options    = nothing
+                               inputs,
+                               output_names=nothing
+                              )
+    if output_names === nothing
+        output_names = @__MODULE__().output_names(o)
     end
-    @timeit timer "CAPI.Run" begin
-        output_tensors = Run(o.api, o.session, run_options, inp_names, input_tensors, output_names)
+    @argcheck o.execution_provider in EXECUTION_PROVIDERS
+    @argcheck eltype(output_names) <: Union{AbstractString, Symbol}
+    @argcheck keytype(inputs) <: Union{AbstractString, Symbol}
+    expected_input_names = ONNXRunTime.input_names(o)
+    for key in keys(inputs)
+        if !(String(key) in expected_input_names)
+            msg = """
+            Invalid input name.
+            Expected: $(expected_input_names)
+            Got: $(key)
+            """
+            throw(ArgumentError(msg))
+        end
     end
-    @timeit timer "post CAPI.Run" begin
-        make_output(o, inputs, output_names, output_tensors)
+    expected_output_names = ONNXRunTime.output_names(o)
+    for name in output_names
+        if !(String(name) in expected_output_names)
+            msg = """
+            Invalid output name.
+            Expected: $(expected_output_names)
+            Got: $(name)
+            """
+            throw(ArgumentError(msg))
+        end
     end
+    inp_names, input_tensors = prepare_inputs(o, inputs)
+    run_options    = nothing
+    output_tensors = Run(o.api, o.session, run_options, inp_names, input_tensors, output_names)
+    make_output(o, inputs, output_names, output_tensors)
 end