Crossing survival curves

[takra]

Hello!
I have 2 groups with treatment and placebo and their survival curves cross (PH assessment is violated), My aim is to estimate effect of treatment. As far as I know LR test or Cox model has low power in such conditions. Are there any test or method which deals with crossing survival curves?

[takra]

1, One idea is to use one of the weighted version of logrank test, because early events are more important. But I don't understand how to estimate power in such test in case H0 is not declined.
2. Also we can use KS test for censored data (i haven't find such function anywhere except R). But interpretation of result becomes hard in case H0 declined. In this case S1not equal S2, but my question is which one is better.