If you need to work around that, you can write a fairly simple vertex
shader to make quads behave as point sprites that should still get
very good performance.
If you make a vertex buffer with each point repeated 4 times with
different texcoords.
ie: x y z 0 0, x y z 1 0, x y z 1 1, x y z 0 1, ...
pass the x and y rows of the view matrix into the vertex shader as constants,
in the shader modify the position as something like
pos.xyz=pos.xyz + scale * viewx * (texcoord.x-0.5)
+ scale * viewy * (texcoord.y-0.5);
then you just render your vertex array as quads
it is a bit more of a hassle and uses 4 times as much memory for
verts. But speed should be comparable, and it is more flexible if you
are interested in adding something like randomized rotations on
sprites or non-square sprites
/<eith
On 10/30/07, Richard S. Wright Jr. <email@hidden> wrote:
> I didn't go into specifics as I was looking for a more general answer. These
> issues have all been brought up before (re my search of the archives), and
> radars filed. The problems are trivial to reproduce and lots of people seem
> to have been complaining about this. What I did not see in the archives was
> something like your response (which actually sufficiently answers my
> question).
>
> It kinda sux that Apple is tying my hands in order to get customers to
> upgrade their hardware/software. I have a fall back, and I can just test for
> Tiger/Leopard at runtime. I was hoping there was a well known work around
> (like, turn on flat shading or something). I guess not, and if this is the
> will of "The Jobs", then so be it. Reason #458322 why Leopard runs faster
> than Tiger ;-)
>
> The Apple guys on this list have always been great and very helpful. I may
> yet hear from them, but I can imagine that no one wants their name at the
> end of a message that states "we abandoned OpenGL on PPC under Tiger" ;-)
> Not that any Apple OpenGL developers did, I'm sure this was a policy
> decision, and a couple of the guys who haunt this list probably spoke up at
> a meeting some months ago, but got shot down.
>
>
> Richard
>
>
> On Oct 30, 2007, at 1:39 PM, John Stiles wrote:
>
>
>
> On PowerPC machines running Tiger, OpenGL didn't receive any updates since
> 10.4.3 so there were a lot of long-standing bugs that didn't get addressed
> until Leopard. If you can't require Leopard, you are basically on your own
> to work around the issues. Intel was a little more lucky and got several
> updates during 10.4's lifespan, though whether it will see any future
> updates is an open quesion.
>
> You should probably be more specific about which configurations work and
> which do not... that might help narrow down the issues more specifically. OS
> version, Intel-or-PowerPC and GPU type are the most important variables.
>
> ________________________________
>
> From:
> mac-opengl-bounces+jstiles=email@hidden on
> behalf of Richard S. Wright Jr.
> Sent: Tue 10/30/2007 10:09 AM
> To: OpenGL Mailing List List
> Subject: What's the poop on point sprites?
>
>
>
>
>
> I'm using point sprites and a custom shader to render a star field in
> an astronomy program. Switching to this method makes my star fields
> almost "free", so it's a very desirable optimization (they are
> rendered via a VBO). I have four test machines, and depending on how I
> move the shells around (a metaphor for you non-american readers) I get
> a variety of results:
>
> Newer machines running Tiger or Leopard... wow, that's fast and pretty.
>
> Older machines, depending on the order of operations, or changes to
> the shader... locks up machine completely with a test shader (OS X
> 10.4.11), or stars are always green (same machine), or points do not
> obey GL_VERTEX_PROGRAM_POINT_SIZE.
> I've never seen a single OpenGL feature behave so many different ways
> on different machines/configurations. Point sprites really don't seem
> to me to be _that_ hard to implement...
>
> A search of the archives reveals that all these problems have been
> reported before, but no one from Apple seems to have any response (or
> if so, I have missed them). Older ATI hardware with 9600 based
> chipsets... is this EVER going to be fixed? I'm running the newest
> 10.4.11 PPC build and my whole machine just freak'n hangs. I would be
> happy to at least hear "this is a hardware limitation and we just
> can't fix it". At least that would show up in the archives and all us
> point sprite fan boys could give up our false hopes ;-)
>
> Richard
>
>
> _______________________________________________
> Do not post admin requests to the list. They will be ignored.
> Mac-opengl mailing list (email@hidden)
> Help/Unsubscribe/Update your Subscription:
> http://lists.apple.com/mailman/options/mac-opengl/email@hidden
>
> This email sent to email@hidden
>
> _______________________________________________
> Do not post admin requests to the list. They will be ignored.
> Mac-opengl mailing list (email@hidden)
> Help/Unsubscribe/Update your Subscription:
> http://lists.apple.com/mailman/options/mac-opengl/email@hidden
>
> This email sent to email@hidden
>
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Mac-opengl mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/mac-opengl/email@hidden
This email sent to email@hidden
