Simpler side_dimension_for_vectorized_dimension

[blegat]

As a double check, there does not seem to be any difference until 100 millions.

findall(m -> isqrt(2m) != div(isqrt(1 + 8m), 2), 1:100000000)
Int64[]

[chriscoey]

Fascinating

[blegat]

It came up when trying to understand the link between isqrt(2m) which is used by SDPLR and the Barvinok-Pataki bound in jump-dev/SDPLR.jl#11. Once I had this 2-line arguments, it clarified things a lot 😄

[odow]

I guess, just as proof, because @blegat's original example has a typo:

julia> findall(n -> isqrt(2n) != div(isqrt(1 + 8n), 2), 1:100_000_000)
Int64[]