SVector construction not working when the input is Int32

[ducanh-le]

Hello,
When trying to improve the performance of my program, I changed all of my variables and input data from Int to Int32. After doing so, I encountered an error with the SVector construction method:

neededResource = usedResource .+ SVector{md.nbM, Int32}(md.r[md.ct[v]][a])

The error description:

ERROR: TypeError: in typeassert, expected Tuple{Vararg{Union{StaticArraysCore.Dynamic, Int64}}}, got a value of type Tuple{Int32}
Stacktrace:
[1] Size
@ ~/.julia/packages/StaticArraysCore/U2Z1K/src/StaticArraysCore.jl:464 [inlined]
[2] Size(s::Type{Tuple{40}})
@ StaticArraysCore ~/.julia/packages/StaticArraysCore/U2Z1K/src/StaticArraysCore.jl:470
[3] Size(#unused#::Type{SVector{40, Int32}})
@ StaticArraysCore ~/.julia/packages/StaticArraysCore/U2Z1K/src/StaticArraysCore.jl:492
[4] length(a::Type{SVector{40, Int32}})
@ StaticArrays ~/.julia/packages/StaticArrays/9KYrc/src/abstractarray.jl:2
[5] convert(#unused#::Type{SVector{40, Int32}}, a::Vector{Int32})
@ StaticArrays ~/.julia/packages/StaticArrays/9KYrc/src/convert.jl:200
[6] StaticArray
@ ~/.julia/packages/StaticArrays/9KYrc/src/convert.jl:174 [inlined]
[7] resourceAvailable(v::Int32, a::Int32, s::Solution, t::Int32, md::ModelData)
@ Main ~/stage-ONERA/model-metaheuristic/GreedyHeuristic.jl:23

How to get rid of this error when using Int32 please?

[mateuszbaran]

Hi,
AFAIK sizes need to be Int64. It shouldn't be too much of a problem though because they are used primarily at compilation time.

[mateuszbaran]

This: SVector{md.nbM, Int32}(md.r[md.ct[v]][a]) is likely suboptimal. If md.nbM can't be constant-propagated (and almost surely it can't), there will be a lot of dynamic conversions and dispatch. You need to carefully profile it to see if this isn't slower than just using Vector.

[ducanh-le]

Hi, AFAIK sizes need to be Int64. It shouldn't be too much of a problem though because they are used primarily at compilation time.

You are right, this solved the problem:

nbM::Int = md.nbM
SVector{nbM, Int32}(md.r[md.ct[v]][a])