Merge pull request #39 from JuliaMath/artifact

Adding MKL_jll artifact

Author: Crown421

.github/workflows/main.yml

@@ -1,4 +1,4 @@
-name: julia 1.0/1.3
+name: julia 1.3
 on:
   push:
     branches:
@@ -13,7 +13,6 @@ jobs:
       fail-fast: false
       matrix:
         version:
-          - '1.0'
           - '1.3'
         os:
           - ubuntu-latest

Project.toml

@@ -1,16 +1,13 @@
 name = "IntelVectorMath"
 uuid = "c8ce9da6-5d36-5c03-b118-5a70151be7bc"
-version = "0.3.1"
+version = "0.4.0"
 
 [deps]
-BinaryProvider = "b99e7846-7c00-51b0-8f62-c81ae34c0232"
-CpuId = "adafc99b-e345-5852-983c-f28acb93d879"
-Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+MKL_jll = "856f044c-d86e-5d09-b602-aeab76dc8ba7"
 
 [compat]
-BinaryProvider = "0.5.8"
-CpuId = "0.2"
-julia = "0.7, 1.0"
+julia = "1.3"
+MKL_jll = "2020"
 
 [extras]
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

README.md

@@ -4,7 +4,7 @@
 [pkgeval-url]: https://juliaci.github.io/NanosoldierReports/pkgeval_badges/report.html
 
 [![PkgEval][pkgeval-img]][pkgeval-url]
-![](https://github.com/JuliaMath/VML.jl/workflows/julia%201.0/1.3/badge.svg)
+![](https://github.com/JuliaMath/VML.jl/workflows/julia%201.3/badge.svg)
 ![](https://github.com/JuliaMath/VML.jl/workflows/julia%20nightly/badge.svg)
 
 This package provides bindings to the Intel MKL [Vector Mathematics Functions](https://software.intel.com/en-us/node/521751).
@@ -13,14 +13,18 @@ Until Julia 0.6 the package was registered as `VML.jl`.
 
 Similar packages are [Yeppp.jl](https://github.com/JuliaMath/Yeppp.jl), which wraps the open source Yeppp library, and [AppleAccelerate.jl](https://github.com/JuliaMath/AppleAccelerate.jl), which provides access to macOS's Accelerate framework.
 
+### Warning for macOS
+There is currently [the following](https://github.com/JuliaPackaging/BinaryBuilder.jl/issues/700) issue between the `CompilerSupportLibraries_jll` artifact, which is used for example by `SpecialFunctions.jl`, and `MKL_jll`. Unless `MKL_jll` is loaded first, there might be wrong results coming from a small number of function for particular input array lengths. If you are unsure which, if any, your used packages might load this artifact, loading `IntelVectorMath` as the very first package should be fine. 
+
 ## Basic install
 
 To install IntelVectorMath.jl run
 ```julia
 julia> ] add IntelVectorMath
 ```
-Since version 0.3 IntelVectorMath downloads its own version of MKL and keeps only the required files in its own directory. As such installing MKL.jl or MKL via intel are no longer required, and may mean some duplicate files if they are present. However, this package will adopt the new artifact system in the next minor version update and fix this issue. 
+Since version 0.4 `IntelVectorMath` uses the `MKL_jll` artifact, which is shared with other packages uses MKL, removing several other dependencies. This has the side effect that from version 0.4 onwards this package requires at least Julia 1.3. 
 
+For older versions of Julia `IntelVectorMath v0.3` downloads its own version of MKL and keeps only the required files in its own directory. As such installing MKL.jl or MKL via intel are no longer required, and may mean some duplicate files if they are present. However, this package will adopt the new artifact system in the next minor version update and fix this issue. 
 In the event that MKL was not installed properly you will get an error when first `using` it. Please try running 
 ```julia
 julia> ] build IntelVectorMath

deps/.gitignore

@@ -6,11 +6,15 @@ export IVM
 const IVM = IntelVectorMath
 
 # import Base: .^, ./
-# using Libdl
-include("../deps/deps.jl")
-
 include("setup.jl")
 
+function __init__()
+    compilersupportlibaries_jll_uuid = Base.UUID("e66e0078-7015-5450-92f7-15fbd957f2ae")
+    if Sys.isapple() && haskey(Base.loaded_modules, Base.PkgId(compilersupportlibaries_jll_uuid, "CompilerSupportLibraries_jll"))
+        @warn "It appears CompilerSupportLibraries_jll was loaded prior to this package, which currently on mac may lead to wrong results in some cases. For further details see github.com/JuliaMath/IntelVectorMath.jl"
+    end
+end
+
 for t in (Float32, Float64, ComplexF32, ComplexF64)
     # Unary, real or complex
     def_unary_op(t, t, :acos, :acos!, :Acos)

deps/build.jl

@@ -1,14 +1,4 @@
-
-function __init__()
-    check_deps()
-
-    Libdl.dlopen(libmkl_core, Libdl.RTLD_GLOBAL)
-    Libdl.dlopen(libmkl_rt, Libdl.RTLD_GLOBAL) # maybe only needed on mac
-    Libdl.dlopen(libmkl_vml_avx, Libdl.RTLD_GLOBAL)
-
-end
-
-__init__()
+import MKL_jll
 
 struct VMLAccuracy
     mode::UInt
@@ -20,15 +10,15 @@ const VML_EP = VMLAccuracy(0x00000003)
 
 Base.show(io::IO, m::VMLAccuracy) = print(io, m == VML_LA ? "VML_LA" :
                                               m == VML_HA ? "VML_HA" : "VML_EP")
-vml_get_mode() = ccall((:_vmlGetMode, libmkl_vml_avx), Cuint, ())
-vml_set_mode(mode::Integer) = (ccall((:_vmlSetMode, libmkl_vml_avx), Cuint, (UInt,), mode); nothing)
+                                              
+vml_get_mode() = ccall((:vmlGetMode, MKL_jll.libmkl_rt), Cuint, ())
+vml_set_mode(mode::Integer) = (ccall((:vmlSetMode, MKL_jll.libmkl_rt), Cuint, (UInt,), mode); nothing)
 
 vml_set_accuracy(m::VMLAccuracy) = vml_set_mode((vml_get_mode() & ~0x03) | m.mode)
 vml_get_accuracy() = VMLAccuracy(vml_get_mode() & 0x3)
 
-vml_set_mode((vml_get_mode() & ~0x0000FF00))
 function vml_check_error()
-    vml_error = ccall((:_vmlClearErrStatus, libmkl_vml_avx), Cint, ())
+    vml_error = ccall((:vmlClearErrStatus, MKL_jll.libmkl_rt), Cint, ())
     if vml_error != 0
         if vml_error == 1
             throw(DomainError(-1, "This function does not support arguments outside its domain"))
@@ -45,13 +35,13 @@ end
 
 function vml_prefix(t::DataType)
     if t == Float32
-        return "_vmls"
+        return "vs"
     elseif t == Float64
-        return "_vmld"
+        return "vd"
     elseif t == Complex{Float32}
-        return "_vmlc"
+        return "vc"
     elseif t == Complex{Float64}
-        return "_vmlz"
+        return "vz"
     end
     error("unknown type $t")
 end
@@ -63,24 +53,24 @@ function def_unary_op(tin, tout, jlname, jlname!, mklname;
     (@isdefined jlname) || push!(exports, jlname)
     (@isdefined jlname!) || push!(exports, jlname!)
     @eval begin
-        function ($jlname!)(out::Array{$tout,N}, A::Array{$tin,N}) where {N}
+        function ($jlname!)(out::Array{$tout}, A::Array{$tin})
             size(out) == size(A) || throw(DimensionMismatch())
-            ccall(($mklfn, libmkl_vml_avx), Nothing, (Int, Ptr{$tin}, Ptr{$tout}), length(A), A, out)
+            ccall(($mklfn, MKL_jll.libmkl_rt), Nothing, (Int, Ptr{$tin}, Ptr{$tout}), length(A), A, out)
             vml_check_error()
             return out
         end
         $(if tin == tout
             quote
                 function $(jlname!)(A::Array{$tin})
-                    ccall(($mklfn, libmkl_vml_avx), Nothing, (Int, Ptr{$tin}, Ptr{$tout}), length(A), A, A)
+                    ccall(($mklfn, MKL_jll.libmkl_rt), Nothing, (Int, Ptr{$tin}, Ptr{$tout}), length(A), A, A)
                     vml_check_error()
                     return A
                 end
             end
         end)
         function ($jlname)(A::Array{$tin})
             out = similar(A, $tout)
-            ccall(($mklfn, libmkl_vml_avx), Nothing, (Int, Ptr{$tin}, Ptr{$tout}), length(A), A, out)
+            ccall(($mklfn, MKL_jll.libmkl_rt), Nothing, (Int, Ptr{$tin}, Ptr{$tout}), length(A), A, out)
             vml_check_error()
             return out
         end
@@ -95,16 +85,16 @@ function def_binary_op(tin, tout, jlname, jlname!, mklname, broadcast)
     (@isdefined jlname!) || push!(exports, jlname!)
     @eval begin
         $(isempty(exports) ? nothing : Expr(:export, exports...))
-        function ($jlname!)(out::Array{$tout,N}, A::Array{$tin,N}, B::Array{$tin,N}) where {N}
-            size(out) == size(A) == size(B) || $(broadcast ? :(return broadcast!($jlname, out, A, B)) : :(throw(DimensionMismatch())))
-            ccall(($mklfn, libmkl_vml_avx), Nothing, (Int, Ptr{$tin}, Ptr{$tin}, Ptr{$tout}), length(A), A, B, out)
+        function ($jlname!)(out::Array{$tout}, A::Array{$tin}, B::Array{$tin}) 
+            size(out) == size(A) == size(B) || throw(DimensionMismatch("Input arrays and output array need to have the same size"))
+            ccall(($mklfn, MKL_jll.libmkl_rt), Nothing, (Int, Ptr{$tin}, Ptr{$tin}, Ptr{$tout}), length(A), A, B, out)
             vml_check_error()
             return out
         end
-        function ($jlname)(A::Array{$tout,N}, B::Array{$tin,N}) where {N}
-            size(A) == size(B) || $(broadcast ? :(return broadcast($jlname, A, B)) : :(throw(DimensionMismatch())))
+        function ($jlname)(A::Array{$tout}, B::Array{$tin}) 
+            size(A) == size(B) || throw(DimensionMismatch("Input arrays need to have the same size"))
             out = similar(A)
-            ccall(($mklfn, libmkl_vml_avx), Nothing, (Int, Ptr{$tin}, Ptr{$tin}, Ptr{$tout}), length(A), A, B, out)
+            ccall(($mklfn, MKL_jll.libmkl_rt), Nothing, (Int, Ptr{$tin}, Ptr{$tin}, Ptr{$tout}), length(A), A, B, out)
             vml_check_error()
             return out
         end

src/IntelVectorMath.jl

@@ -3,6 +3,12 @@
 [[Base64]]
 uuid = "2a0f44e3-6c83-55bd-87e4-b1978d98bd5f"
 
+[[CompilerSupportLibraries_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "b57c5d019367c90f234a7bc7e24ff0a84971da5d"
+uuid = "e66e0078-7015-5450-92f7-15fbd957f2ae"
+version = "0.2.0+1"
+
 [[CpuId]]
 deps = ["Markdown", "Test"]
 git-tree-sha1 = "f0464e499ab9973b43c20f8216d088b61fda80c6"
@@ -17,6 +23,12 @@ uuid = "ade2ca70-3891-5945-98fb-dc099432e06a"
 deps = ["Random", "Serialization", "Sockets"]
 uuid = "8ba89e20-285c-5b6f-9357-94700520ee1b"
 
+[[IntelOpenMP_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "fb8e1c7a5594ba56f9011310790e03b5384998d6"
+uuid = "1d5cc7b8-4909-519e-a0f8-d0f5ad9712d0"
+version = "2018.0.3+0"
+
 [[InteractiveUtils]]
 deps = ["Markdown"]
 uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
@@ -30,15 +42,21 @@ uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
 [[Logging]]
 uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
 
+[[MKL_jll]]
+deps = ["IntelOpenMP_jll", "Libdl", "Pkg"]
+git-tree-sha1 = "7f7034f27491a8bb955a9e4097e4fb5c2167bda5"
+uuid = "856f044c-d86e-5d09-b602-aeab76dc8ba7"
+version = "2020.0.166+0"
+
 [[Markdown]]
 deps = ["Base64"]
 uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
 
 [[OpenSpecFun_jll]]
-deps = ["Libdl", "Pkg"]
-git-tree-sha1 = "65f672edebf3f4e613ddf37db9dcbd7a407e5e90"
+deps = ["CompilerSupportLibraries_jll", "Libdl", "Pkg"]
+git-tree-sha1 = "d110040968b9afe95c6bd9c6233570b0fe8abd22"
 uuid = "efe28fd5-8261-553b-a9e1-b2916fc3738e"
-version = "0.5.3+1"
+version = "0.5.3+2"
 
 [[Pkg]]
 deps = ["Dates", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Test", "UUIDs"]

src/setup.jl

@@ -1,5 +1,5 @@
 [deps]
 CpuId = "adafc99b-e345-5852-983c-f28acb93d879"
-Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+MKL_jll = "856f044c-d86e-5d09-b602-aeab76dc8ba7"
 SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

test/Manifest.toml

@@ -1,3 +1,4 @@
+import MKL_jll
 using Test
 using IntelVectorMath