Difference between revisions of "User:Marco-domingues/GSoC17/Log"
|Line 241:||Line 241:|
* Working on a solution to iterate only over partitions evaluated in the shading process
* Working on a solution to iterate only over partitions evaluated in the shading process
Revision as of 17:49, 27 July 2017
- 1 Development Logs
- 2 May 30, 2017
- 3 Week 1: 30 May - 2 June
- 4 Week 2:
- 5 5 June
- 6 6 June
- 7 7-8 June
- 8 9 June
- 9 Week 3
- 10 12-14 June
- 11 15 June
- 12 16-17 June
- 13 Week 4
- 14 19 June
- 15 21 June
- 16 22 June
- 17 23 June
- 18 24 - 26 June
- 19 27 June
- 20 28 June
- 21 29-30 June
- 22 3 July
- 23 4-5 July
- 24 6 July
- 25 7 July
- 26 10 July
- 27 11 July
- 28 12 - 14 July
- 29 17 July
- 30 18 July
- 31 19 July
- 32 20-21 July
- 33 24 July
- 34 25 July
- 35 26 July
- 36 27 July
Community Bonding Period
- During the community bonding period, I read the code and asked some questions about the ANSI C boolean evaluation algorithm to better understand some concepts and to plan the necessary things to port to OpenCL.
- I started to write some code in order to get an advance on my project. This consisted in creating the kernels in OpenCl to weave the segments and to evaluate the partitions, and setting the arguments for those kernels. Ended with an initial version of the "weave_segs" kernel.
May 30, 2017
- Coding Period officially begins!!
- Because I misunderstood an important concept about the weave of segments, the resulting number of partitions with the initial version of the 'weave_segs' kernel was incorrect. Started to implement a new version of the code
Week 1: 30 May - 2 June
- Finalized the "weave_segs" kernel and the the solution seems ok. When testing the code with a simple CSG scene (two spheres intersecting), the number of partitions and the number of segments in each partition is correct. 1, 2 or 3 partitions per ray at most, and the rays with 3 partitions have only 2 segments in the second partition, which is supposed according to the testing example.
- After evaluating the partitions, the results also seemed correct. Every partition evaluated with "TRUE" value is from a ray with 3 partitions, and only
the middle partition of the ray has the "TRUE" value.
- Planning to shade the segments from the evaluated partitions next week.
- Worked on an initial solution to shade the segments resulting from boolean evaluation. The resulting image differs from the expected result, so i will have to get back to this and figure out what is missing
- After suggestion from my mentor, implemented a white/black shading of the resulting partitions. From the resulting image, I could figure out that a large number of partitions that should be evaluated with 'true' were missing
- After debugging the 'weave_segs' and the code to evaluate partitions, found some issues on the 'weave_segs' kernel that caused the missing resulting partitions. After fixing the bugs on the code, the total number of partitions evaluated with true by the OCL code matched with the total number of partitions evaluated by the Ansi C code and the resulting white/black image was correct!
- Adaptaded the code of the 'shade_segs' kernel to shade the segments of the resulting partitions. There are still some problems in the resulting image.
- The next figure contains images with the results of the OCL code and a comparison with the Ansi C code.
- Planning to refractor the 'weave_segs' kernel next week to a solution that does not use a bounded array for the segments in each partition and to clean up the code in order to submit an initial patch.
- Implemented the 'weave_segs' kernel without using a bounded array to store the segments in each partition. The number of evaluated partitions matches the number of evaluated partitions in the ansi c code and the resulting image is correct.
- Tomorrow I will create some more CSG scenes and will test the code with those new scenes to make sure everything is okay before submitting the patch to weave segments into partitions.
- Reviewed the code with my mentor and will make some improvements based on the received feedback
- Fixed an issue on the code that would occur when inserting new partitions on the partitions list
- Implemented a bitvector to represent the segments in each partition
- Fixed a bug on the code that would cause some missing partitions when running the OCL code in the GPU
- Refactored the code and submitted an initial patch to weave the segments into partitions (https://sourceforge.net/p/brlcad/patches/468/)
- Added a lightmodel to the ansi code to perform white shading, in order to have a more exact comparision between the Ansi C and the OpenCL code
|Ansi C||OCL - Intel i5 4690k||OCL - Nvidia GTX 970|
|Union||0,314 sec||0,098 sec||0,173 sec|
|Difference||0,318 sec||0,098 sec||0,175 sec|
|Intersection||0,301 sec||0,097 sec||0,164 sec|
- Removed pointers from struct cl_partition and updated patch (https://sourceforge.net/p/brlcad/patches/468/)
- shade_segs kernel shading only segments from evaluated partitions
- optimized struct cl_partition to store indexes to segments instead of cl_segs, making it more memory efficient
24 - 26 June
- working on a list structure to store the partitions instead of an array of partitions and implementing a dynamic bit vector to represent the segments in each partition. Still some issues to fix
- Fixed an issue that would occur when duplicating partitions.
- Inverted the loop used to perform boolean evaluation with the bitvector of segments, which should reduce the total number of iterations.
- Code refactored and cleaned up
- Updated the patch with the weave_segs (https://sourceforge.net/p/brlcad/patches/468/). The patch contains code to perform the weave of segs using a dynamic bitarray to represent the segments in each partition and the partitions represented in a list structure, which should optimize the insertion of partitions
- GSOC 2017 Phase 1 Evaluations period
- Started to port rt_boolfinal into OpenCl
- Prepared and sent the boolean regions to the 'eval_partitions' kernel, instead of sending only the boolean trees
- Working on an algorithm to create the list with all the boolean regions involved in the partition. This requires iterating over all the segments of the partition and over all of the regions, which may not be the best solution yet.
- Created a regiontable bitvector to hold the boolean regions involved in each partition. The regiontable is necessary to evaluate the partitions only against all the regions involved, and to resolve overlapping partitions.
- Found and fixed a bug in the insertion of partitions that would generate partitions with wrong 'back_pp/forw_pp' pointers. Now is possible to iterate over the partitions list in a sequential manner in the 'eval_partitions' kernel, and test the partitions against additional conditions before evaluation.
- Removed the pointers from the bool_region structure. Now, I created an OpenCl buffer that contains boolean trees from all the regions, and I use other buffer to know the offset for each region.
- Worked on the shading of evaluated partitions. Before, I was wrongly shading segments from the evaluated partitions, but now I shade with the information of the 'inhit' and the results seem to match with the ansi c results. There are some slightly differences in the illumination, but those differences were already there before.
- Debugged the code trying to identify the cause for the wrong results when evaluating partitions in the share/db/operators.g scene, since the regiontable seems to be correctly built. By comparing the output produced by the ansi c version and the OpenCl version, could identify several occurrences of segments in partitions with wrong 'seg_sti', which would explain the wrong results when evaluating partitions. Despite the wrong 'seg_sti', the segments of the partitions have the correct values (hit_dist values of the segments match the values of the ansi c partitions).
- After further investigation on the 'seg_sti' issue, it appears that the boolean trees have the bits of the ordered primitives, and the 'seg_sti' calculated by the 'store_segs' kernel are the same as the 'rtip->rti_Solids->st_bit'. To confirm this, I translated the bits of the boolean trees to the 'rtip->rti_Solids' bits before sending the boolean trees to OpenCL, and was able to get the following results:
- Hopefully this help us find the cause for the 'seg_sti' mismatch between the segments and the bits in the boolean tree.
12 - 14 July
- Cleaning and refactoring the code in order to submit the patch against the opencl branch in svn.
- While testing the code against some more complex scenes, found and fixed the bug that was causing some of those scenes to take more time than expected to render.
- Found and fixed the bug that was causing some 'holes' in some views. Here is an example of a view that was having this problem, now fixed:
- Will finish the rt_boolfinal kernel, by implementing the handler for overlapping partitions before submiting the patch in the svn.
- Implemented an overlap handler to resolve the overlapping of partitions
- Completed the rt_boolfinal kernel. There are still some bugs with more complex scenes, that I believe are caused by some partitions reporting overlap when they shouldn't (by comparison with the ansi c code), but still have to investigate.
- Cleaned and refactored the code and submitted the patch in the svn: https://sourceforge.net/p/brlcad/patches/472/ (This was created against the opencl code branch)
- Planning to optimize the code next, starting by implementing a list for the regiontable, instead of using a dynamic bitvector.
- Debugged the code and found the cause for weird results when drawing geometry with the command 'e *'. It comes from a bug in the overlap handler, that might be hard to resolve using bitarrays.
- Working on a solution that uses an array to represent the regiontable, instead of a bitarray
- Adjacent partitions should no longer write to the same memory positions of the bitarray. This was causing CSG scenes with more than 32 regions and/or more than 32 segments per ray to produce wrong results.
- Optimized the code to iterate over bitarrays, now iterating only over set bits in the bitarrays.
- Working on a fix to the overlap handler problem
- Resolved the weird results when drawing geometry with the command "e *"
- Changed the code to use only 1 regiontable per ray, instead of 1 regiontable per partition, which reduces the necessary memory.
- Prepared the code and submitted the patch in the svn (https://sourceforge.net/p/brlcad/patches/472/)
- Made some changes on the 'shade_segs' kernel to replicate the "Surface normals" lighting model. It looks like there is an issue with the interior normals, but I am not sure if this comes from the known issue of the primitives having the ids in a different order or not.
- Collected some time measurements over some scenes in the 'share/db' directory. Wasn't quite lucky with the profiling tools, so ended doing this with the elapsed time given from the ray tracing command.
- Tomorrow will add some more measurements and will share the document with the results.
- Did some further testing over some scenes in 'share/db' to gather more statistics to optimize the code
- Fixed the issue that was causing some interior normals to be wrongly represented
- Working on a solution to iterate only over partitions evaluated in the shading process
- Optimized the process of building the regiontable by precomputing a table with all the regions involved in each primitive instead of doing this for each partition in the rt_boolfinal kernel.
- This change on the code resulted in a significant improvement on the performance, as you can see in the following table: