fix VPMASKMOV and VMASKMOVP

[ye-yeshun]

Just like 0x8c, I've made changes to the relevant code to avoid accessing out-of-bounds memory when a portion of the mask is 0.
I added the test case but commented it out, I don't think I should tamper with ref30.txt and the binary file test30.

[ptitSeb]

You need to check if ed is x3 after the geted(...) so you don't need to restore "ed" to it's initial value. The SUB(ed, ed, 24) and SUB(ed, ed, 8) can also be removed after that.

[ptitSeb]

Side question: did you encounter program/games that needs this?

[ye-yeshun]

Side question: did you encounter program/games that needs this?

Sure. Opening the email page of WXWork software and switching emails several times will cause an exception.
After debugging and analysing, I found that it was the MASKMOV instruction accessing an unallocated memory page at a memory page boundary.

[ye-yeshun]

You need to check if ed is x3 after the geted(...) so you don't need to restore "ed" to it's initial value. The SUB(ed, ed, 24) and SUB(ed, ed, 8) can also be removed after that.

Due to the number of changes, I used a forced push, please re-review

[ptitSeb]

Looks better now. Some VEORQ can be moved inside the "else" prt of the if(MODREG) but I'll do that later, it's a minor change.

[ptitSeb]

It seems this PR has broken "The Talos Principle 2" (an UE5+ game). I'll debug tomorrow.

[ye-yeshun]

It seems this PR has broken "The Talos Principle 2" (an UE5+ game). I'll debug tomorrow.

OK, thank you. If you get anything after debugging, please comment here.

[ptitSeb]

It seems this PR has broken "The Talos Principle 2" (an UE5+ game). I'll debug tomorrow.

OK, thank you. If you get anything after debugging, please comment here.

Yes, the issue was that "ed" was not check for being "x3". It's fixed.