Few small questions - in context of latest oggenc2. When I do 3 simple tests:
oggenc2 -q8 test.wav
oggdec test.ogg -w check.wav
Then P4 and x64 give the same check.wav, but the result differs from generic version.
I presume icl version uses sse2 instructions, while generic one relies on x87 code (besides other optimizations potentially influencing the ogg encode) - so this is kinda expected.
Similary, rhe results are also different from vorbis-tools' oggenc (built against the same libs, but compiled with gcc on linux host).
This is I guess normal and nothing to worry about ?
Correct. Theres no exception that different compiles of the same code will give the same results, nor that the same binary on different processors will give identical output. Floating point calculations are only approximate.