[Nelson Castillo Di 24. November 2009]: > On Tue, Nov 24, 2009 at 10:43 AM, Werner Almesberger > <[email protected]> wrote: > > Nelson Castillo wrote: > >> What I understood from this thread is that filtering has to be done in > >> user-space and that drivers should send unfiltered sets of points at a > >> higher rate. > > > > ... or that you (or someone) has to prove clearly that this is one > > of the things you can't do properly in user space, e.g., because > > the processing required would exceed the systems's capabilities. >
To me this seems to be a clear case for a kernel driver task. AIUI we a) want to have kernel hardware interfaces (devices) that are as untroubled as feasible by particular hw device properties and peculiarities (a ts is a ts is a ts [is a pointerdevice]), and b) the whole filtering is to *fix* a *bug* in ts-hw, as jitter is no generic or even desired property of a touchscreen and userland never will be interested in jitter. So if userland ever would need to tweak kernel filtering of ts, this is solely due to the fact the kernels job of filtering isn't performing good enough to filter really just jitter and nothing else. If we want to have a most uniform userland across all platforms, this is a kernel task. An even stronger point in a broader view is userland has no idea of the actual cause of jitter, so it never can perform as good as kernel may. Who knows if we might find tomorrow jitter can be fixed by disabling the backlight converter for the 5 ms of ts-A/D-conversion - or any other means userland has no idea how to ever accomplish it. cheers jOERG
signature.asc
Description: This is a digitally signed message part.
