(07-09-2019, 08:34 PM)Vega Wrote: It appears the floating point RAM changes in NTSC-J causes a bad floating value, which causes an invalid instruction somewhere, then causing a crash and you end in an infinite loop (i know the one you are talking about). There's a few things to try.Okay! I will let you know shortly what happens.
Some options....
1. Change the address of the 1st ASM, you can use the Millisecond Display Modifer address so the actual floating point changes to RAM don't start til the timer starts. If it still freezes at track load on JAPAN, then its something with your nop. Ofc, you are nopping some branch, follow the branch again, and double check what you are skipping/not skipping.
2. Try a different set of floating values, if it works on Japan. Then change them for other regions and retest U, P, and K. If all good, it's just something you will have to accept, idk how much time you want to dig into this rabbit hole.
3. Try #1 and #2 with old source first!! Just in case I botched the new source.
--
Also here's a side note when dealing with offset values greater than 0x7FFF when you wanna load from a memory address set by lis...
Let's say you want to load a word value from 0x8045C21C. To load this without using an ori instruction, just increase the lis value by 1, then sign extend the offset value. Like this...
lis r12, 0x8046 #8045 plus 1
lwz r12, 0xFFFFC21C (r12) #0xC21C sign extended
Try that out for future codes.
Okay it is ready to be put in the Codes Thread!
Super Mario Eclipse, what Super Mario Sunshine could've been.