kicad-developers team mailing list archive

[Lorenzo Marcantonio <l.marcantonio@xxxxxxxxxxxx>]

On Sun, Aug 31, 2014 at 07:03:20PM +0200, Maciej Sumiński wrote:
> wheel, the zoom changes depend on the scrolling speed (to be precise -
> intervals between consecutive scroll events).

Never looked at that behaviour closely.

> I am wondering - is the step 4 really necessary? The current scrolling
> behaviour does not warp the mouse cursor (as advised in UI handbooks)
> and still zooms in the object under the cursor. Does it matter if it
> is perfectly in the screen center or offsetted a bit?

Well the idea is to avoid panning... and if you want to, say, plan
a route going right to a part which is on the right even when scrolling
back...:P yes, you could zoom back *even more* until you get it
sufficiently centered. I just find the current warping behaviour more
useful (and anyway there is an option to disable it).

> It should not be very hard to adjust border panning speed in the code,
> but there is still no option UI to do so. It would be better not to
> add GAL-specific settings now, as they may confuse users if they do
> not apply to the default renderer.

It was just a consideration, it would be faster to explicitly drag with
the middle button anyway.

> I am wondering if the Cairo backend would not work better for you.
> Long time ago I was experimenting to enhance the performance [1]. To
> gain even more speed, you could also disable antialiasing as I know
> you hate it.

AFAIK without hw acceleration antialiased cairo is *fully* unusable. And
in gerbv results are plain wrong (also in gerbv it can't do the xor
composition which is one of the most important thing while checking
gerbers...)

IMHO the real solution would be to derive a GAL renderer which simply
uses the stock xlib/wxDC/whatever calls. These still are the most
responsive and clearest to see for long routing session (as in, do board
for 8 hours a day).

Sadly I know that you can't optimize for the general case :( even in
opengl code for a full screen triangle filler like the current GPUs
can't be used efficiently for a tile based rendered like the (old)
PowerVR (no idea about the current ones...). Too bad for the drop in
acceleration for the conventional 2D primitive (look at windows which
essentially deprecated the GDI and the bugged ATI driver under Linux),
there is no common performance solution between current GPUs and the
slightly older GMA ones (essentially everything before the i3, more or
less... I *hope* that Intel HD would be sufficient for the task).

-- 
Lorenzo Marcantonio
Logos Srl