[Bf-committers] Qdune code
Jonathan Merritt
j.merritt at pgrad.unimelb.edu.au
Thu Nov 8 07:30:15 CET 2007
On 08/11/2007, at 5:03 PM, Dietrich Bollmann wrote:
>> I challenge your assertions on the basis of the results of the Aqsis
>> test suite, which runs by default on Pixie, Aqsis, PRMan, 3Delight,
>> etc. One quick run through the output of those tests hilights
>> numerous bugs in Pixie which have remained un-addressed through the
>> addition of many snazzy new features, while Aqsis aims to be far more
>> feature-complete. The tests are also timed, and your assertions
>> about
>> rendering speed are way off for most geometric primitive types. The
>> data is there, if you'd care to run the tests on your platform and
>> confirm the results...
>
> To me it sounds strange to use a test suite of one of the renderers to
> evaluate the others. Wouldn't it be strange if the renderer the test
> suite belongs to wouldn't win the tests it was developed with??
This would be true if the tests were renderer-specific (i.e. using
features of one renderer which another one doesn't implement). But
there is a RenderMan specification, and renderers claiming to
implement specific features should follow the spec for those features
(or the behavior of PRMan in cases where it's hazy :-). Certainly
3Delight, Air and RenderDotC do very well.
> How about writing a Blender test suite testing those things which are
> important for Blender?
This is certainly a good idea. Those main features would probably be
polygons and subdiv surfaces, both of which are particularly weak in
Aqsis. However, the best way to generate these tests is probably to
have RenderMan export from Blender working.
The point of my original complaints was that QD had achieved a higher
status by being incorporated into Blender directly. If it's so
fantastic, why not let it compete with the other renderers directly by
using the Render API, rather than giving it preferential status? The
same goes for Aqsis, Pixie, et al.
I stand by my original comments. Please use these renderers, and test
out their features. It will help everyone! As actual users, you may
have a different outlook on the "which-renderer-is-my-baby" question
than people who simply read feature-lists and browse online galleries
looking for the sexiest pictures to decide which is best.
Jonathan Merritt.
More information about the Bf-committers
mailing list