I'm looking for feedback on an idea for a Google Summer of Code project.
I looked at bugzilla and found that the D3DX9 versions of Float16To32Array and Float32To16Array were needed for some games to run. I thought that implementing those functions would make a good Google Summer of Code project, but later discovered that Misha Koshelev had already implemented them in a patch that wasn't committed yet.
In looking up documentation on half floats and these functions gave me ideas for other improvements. For instance, the functions may be implemented differently in D3DX10. A post in this thread: http://www.devmaster.net/forums/showthread.php?t=10924 says that D3DX10 uses different rounding rules and can store NaNs and infinities. I have access to computers running Vista and 7 so I would be able to test how the D3DX10 versions of the functions convert floating point numbers. After doing so, I could make the D3DX10 versions of the functions by modifying the D3DX9 patch's code. I could also write a conformance test to make sure that the functions are implemented correctly in D3DX10.
Joshua Beck