Advanced compilers for Ogre's scripts

Post by **Praetor** » Sat Aug 25, 2007 1:21 am

I've now tried CelShading and Compositor demos to check other demos based around shaders. Both work. So, it either has something to do with shadows, something special is happening in that shadow demo's program definitions, or the shadowing system itself in HEAD is going through a hiccup. I'll tried disabling the new compilers and re-running all the demos again. In the mean time, anyone have any input?

Post by **sinbad** » Sat Aug 25, 2007 12:03 pm

Just looking at those shots, standard texture shadows is incorrect too, since the Athene model is missing the albedo texture (she's far too smooth looking - there's supposed to be a subtle stoney texture on her). HEAD shadow demo is fine here.

Here's what I would suggest - re-serialise the material using MaterialSerializer and see what it looks like. You might want to do the same from HEAD so that any differences just from the serialization process are normalised out. It will be much easier to compare them side by side that way and pick out what's different about what's parsed in.

My guess is that it's to do with multiple passes or even techniques, since neither CelShading or the compositor shaders use multiple passes.

Post by **Praetor** » Sat Aug 25, 2007 4:01 pm

When I went back to the old material compiler there were problems with the shadows demo as well. I'll try reserializing from both compilers. Maybe there's something wrong somewhere else.

Post by **Praetor** » Sun Sep 02, 2007 8:33 pm

After hunting down a small bug in the grammar I have committed a completed (an minimally tested) compositor compiler. It is integrated with the CompositorManager in the same way as the particle and material compilers have been (OGRE_USE_NEW_COMPILERS switch).

I will be running some preliminary performance tests this week. I'll let you all know how that goes and what, if anything, can be done to speed these things up.

Post by **Kencho** » Sun Sep 02, 2007 8:57 pm

Cool!

By the way, an off-topic suggestion. As the GSoC has ended, what about moving this thread to another section? (obviously, leaving a shadow thread here) Developer talk maybe?

Post by **Praetor** » Sun Sep 02, 2007 9:04 pm

That would be fine with me. I'll start a new thread with a link back when I get information from the performance tests.

Post by **sinbad** » Mon Sep 03, 2007 2:46 pm

Sounds great. As things are starting to settle down more, it's probably time to ask people to try it out in more varied platforms / compilers too.