Summary:
1. Proposal 1: allow for a GPU operating mode where a single index array is used for vertices, normals and texture coordinates encoding.
Solution: design a GPU-friendly profile where we constrain the existence of a unique index table shared by coordinates and attributes. Having one index table is already supported by the specification.
2. Proposal 2: allow for an embedded GPU operating mode where only triangles are encoded and supported.
Solution: design a GPU-friendly profile where we constrain the facets to be only triangles.
3. Proposal 3: allow for a GPU operating mode where several triangle series can be represented at encoding time in a single vertex array, using the concept of primitive restart as described in OpenGL ES 3.0.
Solution: this feature requires changing of the bitstream syntax and semantics.
4. Proposal 4: allow for a GPU operating mode where the model is encoded as already flattened according to its texture UV map.
Solution: This proposal is already included in Proposal 1.
5. Proposal 5: allow for a GPU operating mode where texture formats such as ASTC and ETC2 are natively supported at encoder and decoder sides. By doing so the loading time decreases by 10-20 times for typical models with large textures.
Solution: Add support for ETC2 and ASTC in MP4 File Format (AMD of MPEG-4 Part1). Check the usage conditions of these texture formats (with Khronos?!, others).
6. Proposal 6: allow for a GPU operating mode where the mipmap texture cascade can be described and formatted by the encoder as a single image track.
Solution: this can be done by supporting KTX or a similar wrapper that will refer to ETC2 or ASTC (AMD of MPEG-4 Part1). Check the usage conditions of KTX.
7. Proposal 7: allow for a GPU operating mode where shading techniques can be described at high level by the encoder for efficient execution by GPUs with different capabilities.
Solution: Support for Shaders should be added in MPEG-4 BIFS, same level as for COLLADA FX and X3D Shaders. Start an AMD of MPEG-4 Part 11.
8. Proposal 8: allow for an embedded GPU operating mode where the total number of bones and the number of bones associated to a single vertex is limited during encoding to a specification-defined limit.
Solution: define several levels wrt the total of bones in the model and the maximum number of vertices connected to a bone. Numbers from OpenGL ES2.0 and ES3.0 can be used.
9. Proposal 9: a cleared integrated reference model should be provided with representative examples and documentation.
Solution: a unique place with the latest version of the code, examples, and documentation more precise that the specification. 3DG website is necessary more than ever.
10. Proposal 10: a better distinction between the notions of bones and joints should be made into documentation.
See previous solution.
11. Proposal 11: work on detailed description of BBA decompression with complexity measurements.
Solution: Revise the implementation of the BBA decoder.
12. Proposal 12: investigate an embedded GPU operating mode usage of transformation matrices at 16 bit floating point precision and related coding.
Solution: this should be investigated (during the BBA revision)
AP1: start a Profile and Level for 1, 2 and 8 (TFAN and BBA)
AP2: start a AMD of Part 11 for shaders (point 7)
AP3: start AMD of Part 1 for textures (point 5 and 6)
AP4: website (9, 10)
AP5: BBA v2 (11, 12)
AP6: Corr/AMD of Part 16 (point 3)
|