irmatden/pprt

Automatically exported from code.google.com/p/pprt

pprt's Issues

Artifacts on reflection surfaces' edges

Appeared when geometries' transform matrices were added (revision 2262505252f4)

Original issue reported on code.google.com by [email protected] on 22 Apr 2011 at 2:08

Slow ray/sphere intersection

Too much:
movaps [mem], xmmn
movss xmmn, [mem]

conversions to double

Original issue reported on code.google.com by [email protected] on 23 Apr 2011 at 2:28

Re-organize rib parser's actors

Awfull compile time induced!

Original issue reported on code.google.com by [email protected] on 23 Apr 2011 at 9:26

Separate SL VM from shader instance

CompiledShader is already quite bloated and it doesn't even provide 10% of what 
the spec requests...

1. Create a ShaderCompiler class which maps mnemonics to bytecode and output a 
CompiledShader
2. Create a ShaderVM class which only knows how to execute SL code in a given 
context
3. Create a ShaderContext to hold an instance of runtime variables, along with 
a pointer to the actual bytecode of the shader contained in a CompiledShader

Original issue reported on code.google.com by [email protected] on 20 Apr 2011 at 3:38

[deleted issue]

[deleted issue]

Deduce required VM stack size for a shader

This can easily be done while mapping mnemonics to bytecode since stacked types 
is already tracked

Original issue reported on code.google.com by [email protected] on 20 Apr 2011 at 3:55

Optimize mesh's polygon intersection routine

* 2nd term of each cot can be precomputed
* could use some simd-ing

Original issue reported on code.google.com by [email protected] on 25 Apr 2011 at 8:39