So I was trying to write an ASM code today. One of the lines was something like
stw r4,0(r0)
The code was freezing the game. So I tried stepping through, and it appears that the PPC is trapping the instruction and halting execution. Every time I hit "Step" the instructions never progresses forward...it's the same symptom as trying to use a null pointer for stw/lwz.
I switched the the ASM code to use
stw r4,0(r3)
and it started working fine.
Has anyone else had this happen to them?
EDIT: by the way, just to be clear, lwz r4,0(r0) works. stw r4,0(r0) doesn't.
Post Merge: February 21, 2010, 05:07:22 PM
Heh, so after looking at the PPC ASM reference, I noticed that you can't actually use r0 as the pointer for stw.
http://pds.twi.tudelft.nl/vakken/in1200/labcourse/instruction-set/stw.html
If the bits for rA are 0, then instead of getting the pointer in r0 and adding d to it, it just uses d!
So stw r4,0(r0). Here, rA is r0, and d is 0. Since rA is 0, it just uses d to compute the address for the stw, and since d is 0, we were actually dereferencing a null pointer, wihch was causing the trap.
Post Merge: February 21, 2010, 04:11:48 PM
And it looks like lwz actually has the same problem as stw? If rA = r0 then it ignores the address in r0 too.
http://pds.twi.tudelft.nl/vakken/in1200/labcourse/instruction-set/lwz.html
There are certain registers that you don't want to use because they cause problems like this.
Well yeah, you don't want to use r1 because you'd mess up the stack, or r2, and you have to be careful that you don't overwrite a register that's in use.
But my problem wasn't any of that. The problem is that stw and lwz behave differently when rA = r0. Other instructions, like addi (http://pds.twi.tudelft.nl/vakken/in1200/labcourse/instruction-set/addi.html), have the same flaw.
But not all instructions are like this. ori (http://pds.twi.tudelft.nl/vakken/in1200/labcourse/instruction-set/ori.html) doesn't have this quirk.