pjgaudre / desinceig.jl Goto Github PK

Prior to today, the second order differentiation matrix's infinity-norms of the error on the sinc points was:

julia> using SincFun # checkout commit 4a656dc29a34247c65388d11d513a5982b2f256d

julia> k=collect(-500.0:500.0);

julia> norm(sinc(2,k)-convert(Vector{Float64},sinc(2,convert(Vector{BigFloat},k))),Inf)
1.734723475976807e-18

julia> norm(sinc(2,k)./convert(Vector{Float64},sinc(2,convert(Vector{BigFloat},k)))-1,Inf)
3.3306690738754696e-16

Now, I updated the code to make it essentially allocation-free, and the infinity-norm is also smaller:

julia> using SincFun # checkout master

julia> k=collect(-500.0:500.0);

julia> norm(sinc(2,k)-convert(Vector{Float64},sinc(2,convert(Vector{BigFloat},k))),Inf)
4.336808689942018e-19

julia> norm(sinc(2,k)./convert(Vector{Float64},sinc(2,convert(Vector{BigFloat},k)))-1,Inf)
2.220446049250313e-16

The tests fail because it's not clear how many eigenvalues will be returned, just the ones below the threshold tolerance. And somehow the results are worse in that there are fewer eigenvalues returned than before?

I can't seem to get my code for centrosymmetry to work. When I run my code, I obtain the following error:

using DESincEig, SincFun
(RESULTS, All_Abs_Error_Approx , hoptimal , n, MatrixSizes)=SincEigen(x->x.^2+x.^4,ones,infinite2{Float64}(), [0.125,0.125] , [2.0,2.0] , pi/4,Centro=true)

WARNING: [a,b] concatenation is deprecated; use [a;b] instead
in depwarn at deprecated.jl:73
in oldstyle_vcat_warning at abstractarray.jl:29
in vect at abstractarray.jl:38
in SincEigen at C:\Users\Philippe.julia\v0.4\DESincEig\src\DESincEig.jl:228
while loading no file, in expression starting on line 0
ERROR: MethodError: call has no method matching call(::Float64, ::Array{Float6
4,1})
Closest candidates are:
BoundsError()
BoundsError(::Any...)
DivideError()
...
in qtilde at C:\Users\Philippe.julia\v0.4\DESincEig\src\DESincEig.jl:158
in SincEigen at C:\Users\Philippe.julia\v0.4\DESincEig\src\DESincEig.jl:222

It appears to be in the function qtilde, however, when I run my code without any centrosymmety, (Centro =false), there is no problem....

A small test suite of core functions would help development.

to be consistent with Julia package names. See any of the registered packages.

This line declares that Range is a Vector{Int64}, which is odd. Depending on the use, is Range better used as a Julia Range?

typeof(1:100)
subtypes(Range)
super(Range)