Closed PFedak closed 3 years ago
This actually came up in the same function that led me to notice the issues in the previous PR, truly a goldmine.
I'm not sure this is exhaustive, I just did a search for a regex pattern of a write to RD128
followed by a read from the same bit range in RS
or RT
. pcpyh
fits this pattern, too, but is correct as written.
I had wanted to just use a single 128-bit temp register for some of these to avoid hurting code readability, but sleigh complained. Is there some way around that I'm missing?
Unfortunately Sleigh doesn't support 128 bit very well. (NationalSecurityAgency/ghidra#206)
Unfortunately Sleigh doesn't support 128 bit very well. (NationalSecurityAgency/ghidra#206)
It doesn't support a varnode larger than 64 bits at all.
Were there any changes you wanted to see in this PR?
Were there any changes you wanted to see in this PR?
I haven't actually taken a look yet. Can do tomorrow.
If these are just reordered then it is ok with me.
These commands are commonly called with the same register as rd and either rt or rs. Written naively, bits to be read will be clobbered by writes. The assignments can simply be reordered in cases where the collapsed command has no cycle.
Also add a missing swap in pexch.