Michael Lindon

I was using :vectorz implementation when this error message was produced.

I'd be more than happy with just real eigenvalues implemented for now :) That's all I need for my purposes

Fish is cool, especially with oh-my-fish :) I'm running fedora 22 with vim 7.4.827. I can confirm the problem does not arise when bash is my default shell but switch...

That worked :) thank you

I'm not sure its exactly what I am looking for. I have n 2 vectors and n 2x2 matrices. At the moment I am storing them as Arrays like: m=Array{FixedSizeArrays.Vec{2,Float64},1}()...

@c42f it seems I missed your point about fslice, but does it work for matrices? ``` v = [Mat(rand(2,2)) for i=1:10] ``` ``` julia> @fslice v[1:2,1:2,:] ERROR: MethodError: `size` has...

@c42f Thanks!! I've been able to implement it now as I had hoped, preallocating everything in advance, but sadly it still runs 1 second slower than using native linear algebra...

_**Native Julia:**_ ``` function FFBS(y,x,t,m,M,A,AMAQ,μ,σ²,ł,ρ²) n=length(y) resize!(m,n) resize!(M,n) resize!(x,n) Cₛ=statcorr(ł) m[1]=reshape(ρ²*Cₛ[:,1]*(y[1]-μ)/(σ²+ρ²*Cₛ[1,1]),d,1) M[1]=ρ²*Cₛ-ρ²*Cₛ[:,1]*Cₛ[1,:]*ρ²/(σ²+ρ²*Cₛ[1,1]) resize!(AMAQ,n-1) resize!(A,n-1) Q=Array{Float64}(d,d) for i=2:n Δ=t[i]-t[i-1] if(isdefined(A,i-1)) transition!(A[i-1],Δ,ł) else A[i-1]=transition(Δ,ł) end innovation!(Q,Δ,ł) broadcast!(+,AMAQ[i-1],A[i-1]*(M[i-1]*A[i-1]'),ρ²*Q) BLAS.gemv!('N', 1.0, A[i-1], vec(m[i-1]), 0.0,...

Hi Everyone, thanks for your comments. I profiled my code and it is spending the majority of time in constructors.jl:  The other blocks on...

@timholy I just noticed something from your post, you specify the return type. I just tried the following: ``` julia> @time for i=1:1000000 A=Mat(rand(2,2)) end 11.588897 seconds (56.00 M allocations:...