|
|
2nd version of the patch with reduced aliasing at the cost of slightly increased CPU usage. Implementation is finished, which mostly leaves the libopenmpt up for discussion.
The options are:
- Leave
render.resampler.emulate_amiga as a boolean setting, and add render.resampler.emulate_amiga_type as a new string setting with values "a500", "a1200", "unfiltered", "auto" (default). Making A500 the default for "auto" would be compatible with the current implementation, however A1200 definitely has the better sound so it would be a better default value. Plus, A1200 will be the default in OpenMPT 1.29. This has the advantage that the setting can simply be toggled on and off, but for anyone caring about choosing the filter, two library calls are involved. This the variant currently implemented in the patch.
- Turn
render.resampler.emulate_amiga into a string setting with values "0", "1", "a500", "a1200", "unfiltered". This has the advantage that it's easier to write code that is backwards-compatible with libopenmpt 0.4 (no #ifdefs necessary). People who don't care about the exact filter being emulated can keep using "0" and "1" (equal to "auto" in variant 1) as they have in previous versions of the library. The library user only has to remember a single setting. This is what OpenMPT also does internally.
|
