SigmoidNumbers and sparse Matrix-Vector-Multiplication

[milankl]

I want to use posits in sparse Matrix-Vector multiplications. If I do something like

using SigmoidNumbers
P = Posit{8,0}

i = [1,2,3,4,5]
j = [2,3,4,5,1]
v1 = P.([-1.2,1.1,-1.,2.,3.])

S = sparse(i,j,v1)
x = S*v1

I get the error ERROR: MethodError: no method matching zero(::Type{Any}). Would it be possible to implement the zero function for posits?

[ityonemo]

the zero function exists for posits. https://github.com/interplanetary-robot/SigmoidNumbers/blob/master/src/Math/constants.jl

…

On Thu, Mar 1, 2018 at 9:56 AM, Milan K ***@***.***> wrote: I want to use posits in sparse Matrix-Vector multiplications. If I do something like using SigmoidNumbers P = Posit{8,0} i = [1,2,3,4,5] j = [2,3,4,5,1] v1 = P.([-1.2,1.1,-1.,2.,3.]) S = sparse(i,j,v1) x = S*v1 I get the error ERROR: MethodError: no method matching zero(::Type{Any}). Would it be possible to implement the zero function for posits? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABKaARs-GWHlAPadfXSKot83-kKpO3Tsks5taDZQgaJpZM4SYu8k> .

[milankl]

Is this than an issue of the sparse library?

[ityonemo]

possibly, I'll take a quick look and make sure it's not something I can fix on my end.

…

On Thu, Mar 1, 2018 at 10:10 AM, Milan K ***@***.***> wrote: Is this than an issue of the sparse library? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABKaAbgF-_EGyV8CQQoS0Ekdhdeh3kLsks5taDmTgaJpZM4SYu8k> .

[milankl]

Apparently, the following works though

using SigmoidNumbers
P = Posit{8,0}

i = [1,2,3,4,5]
j = [2,3,4,5,1]
v1 = P.([-1.2,1.1,-1.,2.,3.])
x = zeros(v1)

S = sparse(i,j,v1)
A_mul_B!(x,S,v1)

looking into the SparseMatrices.jl library

function (*){T,S}(A::SparseMatrixCSR{T}, x::StridedVector{S})
    TS = promote_type(Base.LinAlg.arithtype(T), Base.LinAlg.arithtype(S))
    A_mul_B!(similar(x, TS, size(A, 1)), A, convert(AbstractVector{TS}, x))
end

I assume the error occurs in the promote_type function...