Dávid Balatoni

> Looks good. Happy to merge this now and do the other comparisons in a follow-up if you feel like it, but also ok with awaiting those I won't take...

> Maybe the latter is a bit more readable? Yes, especially if considering that I'll have to add more.

a < b and b > a generate the exact same code, I wonder which one I should go for.

Hmm, if you look at a > 0x1F) + (arg0 >> 0x1FU) + M2C_CARRY) (((s32) arg0 >> 0x1F) + ((u32) arg1 >> 0x1FU) + M2C_CARRY) Do you think we can...

Oh wow, that generates code that uses an instruction that's not even implemented yet: orc. I'll take a look at the patterns from the old comparison test and leave the...

Actually, let's look at the patterns from the old comparison test some other time. Looking forward to your feedback to merge this :)

Thanks! I just had this idea: if we know that a certain pattern is only generated if the args are u32 for example, can't we use this information to guess...

May I ask how to do that? :D Though the problem is that in the a < b case we don't know which one from the two is u32. It...

I feel like `arg0 >> 31` and `(arg0 >> 31) ^ 1` might be too short to not create false positives. On the other hand: should I implement these three?:...

I ran this through the whole Mario Party 4 decomp and these new patterns caused 30 files to be changed. Only one of them is a false positive, but that...