Open JayFoxRox opened 2 years ago
I finally found a function which uses these instructions, but still decompiles correctly. Ghidra decompiler currently displays the instruction name like a function; that's acceptable and probably better than inlining what these instructions would do (which I proposed).
Feel free to close the issue if the current behaviour is intended.
I finally found a function which uses these instructions, but still decompiles correctly. Ghidra decompiler currently displays the instruction name like a function; that's acceptable and probably better than inlining what these instructions would do (which I proposed).
Feel free to close the issue if the current behaviour is intended.
There really isn't a good way to model hardware random number generation. This would be best left as it is.
The instructions are declared, but don't seem to be injected:
https://github.com/beardypig/ghidra-emotionengine/blob/9e55ddaf47377b1fc5e0cea4f4abb4fb246cfda3/data/languages/vuupper.sinc#L39-L41 https://github.com/beardypig/ghidra-emotionengine/blob/9e55ddaf47377b1fc5e0cea4f4abb4fb246cfda3/data/languages/vuupper.sinc#L43-L43
Implementations can be found in PCSX2:
(In a function which uses these instructions, the decompiler also complains about "Low-level Error: Overlapping input varnodes", and I suspect it could be related)