From: Felix <fx...@gm...> - 2004-04-06 10:20:24
|
On Tue, 06 Apr 2004 12:24:09 +0200 Paul Heldens <phe...@ti...> wrote: > On Tue, 2004-04-06 at 00:33 +0200, Felix K=FChling wrote: >=20 > > I can't test it right now, but IIRC the radeon drivers even limit the > > size of antialiased points to 1. This seems to be compliant with the > > specs, but wouldn't it be nicer to draw antialiased points as software > > fallbacks and not limit their size (beyond the mesa limits, if there are > > any), if the hardware doesn't support point antialiasing? >=20 > specs of what? the hardware? Any kind of workaround would be a great man glget: ... GL_POINT_SIZE_RANGE params returns two values: the smallest = and largest supported sizes for antialia= sed points. The smallest size must be at most= 1, and the largest size must be at least 1. = See glPointSize. ... So limiting the size of antialiased points seems to be ok, just not very ni= ce. > feature indeed! if it's possible in software that would be ok to for me > and all other users now unable to properly use a few very nice model > apps. If not the DRI people, who can I contact about this for a > workaround? the mesa developer? can I configure mesa in some way to skip > hw acceleration for points? would this theoretically even work? >=20 > And yes I understand the fault is not dri's and not the application > developer, but the hardware. (correct this if I'm wrong) I don't think there is any hardware around that can accelerate all that is specified in OpenGL. So drivers usually have to workaround hardware limitations or fall back to software in some cases. In the case of the point size the spec leaves a lot of room for interpretation. The way the driver acts now seems to be counter-intuitive for many OpenGL programmers. I don't think it's really anyone's fault. It's a choice someone made, and others are not happy with. ;-) Felix |