Issue with DSP.jl and FFTW.jl using "MKL" provider
- Dominant language
- Julia
- Stars
- 300
- Forks
- 61
- PR merge metrics
- No merged PRs in 30d
Description
Hi, I'm experiencing a problem using DSP.jl and FFTW.jl together. Digging into it, I was able to locate part of the problem from the build of FFTW.jl. The problem arises with `ENV["JULIA_FFTW_PROVIDER"] = "MKL"` (and Pkg.build("FFTW") afterwards) and the order of which packages are imported.
This works (with provider "MKL" and "FFTW")
```
using FFTW, DSP
a = rand(100000)
y1 = fft(a)
y2 = fft(a)
y1 == y2 # true
```
This does not work (with "MKL")
```
using DSP, FFTW # <- importing DSP first triggers error
a = rand(100000)
y1 = fft(a)
y2 = fft(a)
y1 == y2 # false
```
Also, for shorter length vectors `a` the error is not triggered.
I'm not sure how to debug further. Any suggestions?
This is not a big problem since I can just use provider "FFTW".
`versioninfo()`:
```
Julia Version 1.3.1
Commit 2d5741174c (2019-12-30 21:36 UTC)
Platform Info:
OS: macOS (x86_64-apple-darwin18.6.0)
CPU: Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-6.0.1 (ORCJIT, skylake)
```
```
(v1.3) pkg> st
Status `~/.julia/environments/v1.3/Project.toml`
[d9c5613a] AbstractOperators v0.2.2
[537997a7] AbstractPlotting v0.9.25
[03ede2b8] AeroAcoustics v0.1.0 [`~/.julia/dev/AeroAcoustics`]
[c7e460c6] ArgParse v1.1.0
[c52e3926] Atom v0.12.7
[6e4b80f9] BenchmarkTools v0.5.0
[b99e7846] BinaryProvider v0.5.8
[336ed68f] CSV v0.5.26
[5d742f6a] CSVFiles v1.0.0
[35d6a980] ColorSchemes v3.6.0
[5ae59095] Colors v0.11.2
[f65535da] Convex v0.13.0
[150eb455] CoordinateTransformations v0.5.0
[a2441757] Coverage v1.0.0
[717857b8] DSP v0.6.3
[a93c6f00] DataFrames v0.20.2
[864edb3b] DataStructures v0.17.10
[b4f34e82] Distances v0.8.2
[31c24e10] Distributions v0.22.5
[e30172f5] Documenter v0.24.5
[8f5d6c58] EzXML v1.1.0
[7a1cc6ca] FFTW v1.2.0
[5789e2e9] FileIO v1.2.2
[f6369f11] ForwardDiff v0.10.9
[e9467ef8] GLMakie v0.0.16
[28b8d3ca] GR v0.47.0
[f67ccb44] HDF5 v0.12.5
[7073ff75] IJulia v1.21.1
[6218d12a] ImageMagick v1.1.2
[3570752a] InsertionLoss v0.1.0 [`../../dev/InsertionLoss`]
[c601a237] Interact v0.10.3
[e5e0dc1b] Juno v0.8.0
[b964fa9f] LaTeXStrings v1.1.0
[266f59ce] LaTeXTabulars v0.1.1
[5078a376] LazyArrays v0.15.0
[9c8b4983] LightXML v0.8.1
[79261aaa] LocalCoverage v0.1.0 #master (https://github.com/tpapp/LocalCoverage.jl.git)
[2fda8390] LsqFit v0.10.0
[23992714] MAT v0.7.0
[6e857e4b] MCMCDiagnostics v0.3.0
[ee78f7c6] Makie v0.9.5
[eff96d63] Measurements v2.2.0
[7269a6da] MeshIO v0.3.1
[0987c9cc] MonteCarloMeasurements v0.8.3
[80a8ab25] MultipleScattering v0.1.2 #master (https://github.com/jondea/MultipleScattering.jl.git)
[2774e3e8] NLsolve v4.3.0
[15e1cf62] NPZ v0.4.0
[429524aa] Optim v0.20.1
[3b7a836e] PGFPlots v3.2.0
[8314cec4] PGFPlotsX v1.2.4
[bef90210] PLCT v0.1.0 [`../../dev/PLCT`]
[9b87118b] PackageCompiler v1.0.2
[d96e819e] Parameters v0.12.0
[91a5bcdd] Plots v0.27.1
[140ffc9f] ProximalAlgorithms v0.3.1
[a725b495] ProximalOperators v0.10.3
[438e738f] PyCall v1.91.4
[d330b81b] PyPlot v2.8.2
[dca85d43] QuartzImageIO v0.7.0
[295af30f] Revise v2.5.3
[f2b01f46] Roots v1.0.0
[6038ab10] Rotations v0.13.0
[c946c3f1] SCS v0.6.6
[6b88af47] ShadowMaps v0.0.0 #master (.)
[928aab9d] SpecialMatrices v1.0.0
[65254759] StatsMakie v0.1.3
[f3b207a7] StatsPlots v0.12.0
[5e0ebb24] Strided v0.3.3
[46cd3e9d] StructuredOptimization v0.2.3
[bd369af6] Tables v1.0.3
[5e47fb64] TestImages v1.0.1
[ebf5ac4f] TexTables v0.1.0
[c751599d] ToeplitzMatrices v0.6.1
[592b5752] Trapz v1.1.0
[fce5fe82] Turing v0.8.3
[b8865327] UnicodePlots v1.2.0
[8149f6b0] WAV v1.0.3
[fdbf4ff8] XLSX v0.6.0
[e88e6eb3] Zygote v0.4.9
```
EDIT: Had pinned DSP@0.6.2 in above status, issue persist for DSP@0.6.3. Corrected in status print
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.