Closed Mc-muffin closed 9 months ago
I think I've fixed it. The problem with the vclip instruction was my SLEIGH code was assuming it was always a "xyz" variant (the manual implies this), and the problem with the vsqrt instruction was it was checking that unused fields were always zero. I've gone through all the instructions for which these problems are relevant and made the disassembly much more permissive.
Could you test the latest unstable build for me before I push out a new version? https://github.com/chaoticgd/ghidra-emotionengine-reloaded/releases/download/unstable/ghidra_10.3.3_PUBLIC_20230915_ghidra-emotionengine-reloaded.zip
Had a power outage today, but now I can confirm the new build correctly disassembles the problematic code sections now, thanks for the quick actions on your part :)
Okay great, I've put out a new release.
When loading Kingdom Hearts (SLPS_25105) there are some VU macro mode opcodes that aren't recognized, disassembling stops and a bookmark is added with the text
according to ee-objdump the unrecognized instruction is
Error [Bad Instruction]: Unable to resolve constructor at (address) low from (other address)
for example:vsqrt Q,vf5x
(bytesBD 03 25 4A
)