[Bf-committers] Qdune code

[Joe Eagar]

Jonathan Merritt wrote:
> On 08/11/2007, at 4:26 PM, Joe Eagar wrote:
>
>   
>> Ray tracing especially is important for some things.
>>     
>
> Yes, but so is actually having a renderer that can manage half of the  
> Ri-Spec.
>
> Have you actually tried to use Pixie on a project?  I evaluated Pixie  
> alongside Aqsis for my equine forelimb movie, which was shown at the  
> Blender conference a couple of years back.  Here's a still frame:
>     http://www.aqsis.org/xoops/modules/xcgal/displayimage.php?pid=196&fullsize=1
>
> I wanted to use Pixie's AO solutions, but I simply couldn't get it to  
> work properly with the (very basic, well set-up) subdiv surfaces that  
> I exported from Aqsis.  I had numerous problems with coordinate system  
> transformations, where Pixie would fall-over as soon as I scaled by -1  
> along any axis (which I was using to set up my coordinate system and  
> do the RH->LH transformation).  Don't even start me on the problems I  
> encountered with motion blur!!!
>
> Having done all of this, it seems ludicrous to me to find somebody  
> claiming Pixie is "better" in a discussion of features.  Are you  
> simply believing the feature list of Pixie, or are you really somebody  
> who's been-there, done-that?
>
>
>   
Eh well actually yes I was believing the feature list. . .

Scaling by -1 seems bad in general to me and hard to get working right, 
but I guess that's nitpicking since it sounds like
aqsis does perform well with it.  Anyway, from a web site point of view, 
Pixie's site presents a far more complete
renderer then aqsis.  They even have DSMs.  I certainly have nothing 
other then the web sites to compare 'em.  Then I,
um, read eh I think it was paul in the aqsis IRC channel mentioning how 
aqsis rendered slower then both pixie *and* jrman. . .which kindof 
"confirmed" it for me.

Though reading between the lines it sounds like aqsis has gotten much 
faster since then, if the testsuite clearly has it on
par with other renderers.

Joe