That's super-useful, thanks. I'm still not sure how to tell which 
functions are predicted though: presumably it's those called from a 
usercmd, but is there a proper channel for checking exactly which those are?

(I noticed FTYPEDESC_NOERRORCHECK...Is that the equivalent of "not 
FTYPEDESC_INSENDTABLE"? Certainly seems that way to me.)

Yahn Bernier wrote:
> In table, predicted forward, rolled back
>   In table and FTYPEDESC_INSENDTABLE -- controlled by server value which has 
> been networked, error checked
>   In table and not FTYPEDESC_INSENDTABLE -- rolled forward and back, but not 
> received from server, so no error checking
>   In table, but FTYPEDESC_PRIVATE -- introspectable (cl_pdump) but not rolled 
> forward/back or error checked (may or may not be networked) -- not part of 
> prediction really
>
> Not in table
>   Unrelated to prediction, not rolled forward back, etc. (rendering 
> variables, e.g.)
>
> All of the logic for rewind/prediction is in client .dll in prediction.cpp 
> and in prediction_copy.cpp and c_baseentity.cpp, etc...
>
> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Tom Edwards
> Sent: Wednesday, August 12, 2009 2:07 PM
> To: hlcoders
> Subject: [hlcoders] Prediction table questions
>
> I'm unclear on what the prediction table actually achieves. Here is what 
> I think might be happening:
>
> *In table:*
>
>     * Prediction error if too far out when next update arrives
>     * Memorised when the prediction system rewinds, and restored afterwards
>
> *Not in table: *
>
>     * Not checked for prediction accuracy
>     * Can be operated on multiple times as the prediction system rewinds
>           o Test prediction->IsFirstTimePredicted() to avoid this
>
> Is this at all accurate? Is there anything else? And:
>
>     * What does it mean to have a var that is the network table but not
>       the prediction table?
>     * What decides which functions get rewound?
>
>
>
> _______________________________________________
> To unsubscribe, edit your list preferences, or view the list archives, please 
> visit:
> http://list.valvesoftware.com/mailman/listinfo/hlcoders
>
>
> _______________________________________________
> To unsubscribe, edit your list preferences, or view the list archives, please 
> visit:
> http://list.valvesoftware.com/mailman/listinfo/hlcoders
>
>
>   


_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlcoders

Reply via email to