Hi, "Matthieu Fertré" <matthieu.fer...@free.fr> a écrit : > You can find some explanations here for instance [1].
Or here, a much-referred paper: http://docs.sun.com/source/806-3568/ncg_goldberg.html But in essence, yes, absolutely. Some projects even have code paths using fixed-point (not only for FPU-less machines, because most FP operations are often as fast as integer ones on for instance x86 line of CPU) to validate that their output is deterministic across several archs. > b/ do not use floating points operation at all. That means relying of > fixed point integers or using a deterministic floating point library > such as streflop [3] As much as I prefer that solution in the infinitely long term (ie there is enough will to implement it), we can also fall victim to overflow and other side effects. > From my point of view, we have to re-add synchronizations, at least > in release mode. Anyway, what has been done for 0.9.0 was useful > and we must find way to not hide bugs with synchronization. You mention "release", so that actual users aren't impacted by non-deterministic operations? Because another use would be in debug mode to signal discrepancies between local computation and synchronization data, so as to track the less deterministic stuff (and possibly correct it). But that would only be helpful if instead b/ was chosen. Best regards, Kurosu _______________________________________________ Wormux-dev mailing list Wormux-dev@gna.org https://mail.gna.org/listinfo/wormux-dev
