Week 13 (August 15th to August 22th)[edit]

  1. Used BCP tool to compress Boost to ~50MB (svn version is ~ 100MB)
  2. Wrote a script that compiles the bundled libraries (except LLVM, which we assume to be already instlled)
  3. Tests made on different machine (Ubuntu 10.04) is crashing when running the oslc application from OSL.

Week 12 (August 8th to August 15th)[edit]

  1. Started adding osl dependencies to a separated svn repository.
  2. Added ilmbase, openexr, oiio and osl, though boost and llvm libraries are too big to be added (~500MB and ~250MB respectively).

Week 11 (August 1st to August 8th)[edit]

  1. Implemented a multi-sample mode (similar to increasing mode, but it keeps the average of the sampled pixels and always uses full resolution)
  2. Implemented a random shoot mode, but it seems there's a bug with rendering in unbuffered mode
  3. Investigated the bug with the unbuffered mode, still without success.

Week 10 (July 25th to August 1st)[edit]

  1. Studied OSL code in order to discover how to use it with a ray tracer.
  2. Implemented a simple mode for ray tracer, but I don't know how to add reflection and refraction to it. I intend to come back to it in the future.
  3. Studied BRL-CAD code to design the new framebuffer mode (similar to incremental).

Week 9 (July 18th to July 25)[edit]

  1. [19/07] Implemented a basic texture shader written in OSL.
  2. [20/07] Stopped using the hypersampling option (-H) and added a loop in the sh_osl to avoid shooting the same ray many times.

Week 8 (July 11th to July 18th)[edit]

  1. Implemented a thread-safe version of sh_osl, though no advantages were being taken from parallelism.
  2. Studied some OSL code to figure out how to use it with multiple threads
  3. [07/18] Implemented a new thread-safe version of sh_osl with less blocking.

Week 7 (July 4th to July 11th)[edit]

  1. Wrote a small tutorial on how to use OSL in BRL-CAD (link)
  2. Ported BRL-CAD cloud shader to OSL (screenshot)
  3. Ported BRL-CAD checker shader to OSL (screenshot)
  4. Implemented support for group of shaders

Week 6 (June 27th to July 4th)[edit]

  1. Implemented support for reflection and refraction in the OSL Shader that is used by rt
  2. Implemented support for refraction in the stand alone application that uses BRL-CAD shooting system to render BRL-CAD scenes with osl shaders [screenshot]
  3. Started getting errors due to different version of libpng. BRL-CAD uses 1.4 and OIIO used 1.2. Recompiled OIIO to use 1.4 and the problem was solved.

Week 5 (June 20th to June 27th)[edit]

  1. Implemented an osl shader, which seems not to give correct results
  2. Implemented a stand-alone application to render BRL-CAD scenes with OSL shaders [screenshot]

Week 4 (June 13th to June 20th)[edit]

  1. Discovered that the crashing was due to multi-threaded issues. I'm currently using -P 1 on my tests.
  2. Re-defined the OSLRenderer interface, so that it considers recursion rays needed for reflection and transmission
  3. Start studying how the rt application works.

Week 3 (June 6th to June 13th)[edit]

  1. Adapted the OSL raytracer so that it can be called by the osl shader.
  2. The rt application is currently crashing. Inspecting with valgrind I suspect that it is due to memory leaks.

Week 2 (May 30th to June 6th)[edit]

  1. Not very productive week. With the help of the mentors I managed to solve the compile error.

Week 1 (May 23th to May 30th )[edit]

  1. Implemented a (poor) polka dot for BRL-CAD, to get familiarized with the shaders syntax. [polka dot goblet]
  2. Trying to compile the OSL raytracer outside OSL build, but getting compile errors, that I have no idea on how to solve. I've emailed the dev list asking for help.

Week 0 (Community Bonding)[edit]

  1. Compiled OSL sources both in Ubuntu 10.10 and Mac OS X Snow Leopard. (had trouble with 11.04, mainly due to LLVM)
  2. Found a raytracer developed by Erich Ocean and Brecht [1] that uses OSL-written shaders. It currently renders refraction wrongly. [screenshot]
  3. Discovered the cause of the above error: I was using the wrong shader for glass. Now, the image is correctly rendered. [screenshot]