<br>
<div><span class="gmail_quote">2008/3/6, Thilo Schulz <a href="mailto:arny@ats.s.bawue.de">arny@ats.s.bawue.de</a>:</span></div>
<div> </div>
<div>> By the way, I got a few things on my mind still:<br><br>> Firstly, if you did not change too much in the renderer, you should be able to<br>> use the built-in /video command in ioquake3 to record videos. In more recent<br>
> revisions, a bug in sound recording was fixed, so if you use that one and<br>> don't use OpenAL you can even record sound. You can record videos with higher<br>> resolutions, as the engine will record a constant 25fps and will slow down if<br>
> the performance does not suffice such that the time scale is correct in the<br>> video.</div>
<div>Might be worth a try, thanks for the hint!<br><br>> Secondly, I was curious whether the rendering in your ioquake3 version is<br>> already multicore capable.</div>
<div>Yep, it spawns a thread for each core.<br><br>> And thirdly, complexity in ray-tracing is probably also dependent on the<br>> number of triangles in a scene. I also had the impression that the videos<br>> showed maps with not that many polygons.</div>
<div>I thought Breach was one of the more complex maps, hmm. Suggestions?<br><br>> Fourthly, what are windows users going to do? g++ is not common there as you<br>> probably know.<br>I think we should investigate using the Tiny C Compiler (<a href="http://www.landley.net/code/tinycc/">http://www.landley.net/code/tinycc/</a>) for this. It can be compiled as a library and embedded into the ioquake 3 engine. Rewriting the shader transcoder to emit C code instead of C++ should be too hard.</div>
<div> </div>
<div>Stephan</div>