Memory usage with @benchmarkable

julia> A = randn(10000,10000);
julia> b = @benchmarkable println(size($A)) samples=1;
julia> A = [];
julia> varinfo()
A    40 bytes 0-element Array{Any,1}
b    64 bytes BenchmarkTools.Benchmark{Symbol("##benchmark#363")}
julia> run(b);
(10000, 10000)

Where is the matrix stored at this point?

If I have a large suite of benchmarks in a BenchmarkGroup, like in the example in the manual, won’t it end up holding all of the data needed to run all benchmarks in memory at the same time?