Don't compute diff unnecessarily in simple update

[kshyatt]

Right now this compare_weights is being run every iteration, even when completely unused. We can move it down into the showinfo block to avoid wasting some time.

[lkdvos]

Nice catch! Small extra suggestion but otherwise happy to merge

[kshyatt]

OK all fixed :)

[Yue-Zhengyuan]

My original intention is to check weight difference after each step of SU, so we can stop immediately when it converges, even when the interval to output this information is usually larger. For cases where convergence is not checked (finite-T and real time) it is unnecessary; I just kept weight diff there in case I want to have a look. Even when convergence is checked, I guess we can afford to run a few more steps to reach one full check_interval.

[kshyatt]

Maybe then we can change the behavior depending on a flag or kwarg?

[Yue-Zhengyuan]

How about calculating weight difference after each step only for ground state evolution, and every check_interval for other cases?

[kshyatt]

Sure, that also works for me. Feel free to push to this branch.

[Yue-Zhengyuan]

Wait a few hours before I get back to my laptop

[kshyatt]

Yeah of course, no one is about to die from over comparison of weights 😹

[Yue-Zhengyuan]

Turns out this is already achieved 😂 The ground state evolution is dispatched to time_evolve(it, H; ...) that further needs a Hamiltonian H, and it still has the old behavior that weight diff is calculated in each iteration. Other purposes use time_evolve(it; ...), which is already fixed.

[kshyatt]

The GPU stuff is irrelevant here so I'll just merge, then. Thanks everyone! <3

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/algorithms/time_evolution/simpleupdate.jl	97.08% <100.00%> (-0.03%)	⬇️

... and 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.